JPS63239398A - Learning control method for centrifugal compressor - Google Patents

Abstract

PURPOSE:To enable the changes of conditions and secular changes to be automatically followed up, by correcting the parameter value on a characteristic function, when the difference between the predicted value of flow rate, efficiency and the actual value is greater than a prescribed value, so that the predicted value can approach the actual value. CONSTITUTION:Based on the characteristic function in which flow rates and efficiencies are predetermined, a control unit 28 determines respective vane angle operating quantities for inlet guide vanes 11 through 14, and diffuser vanes 15 through 18, and based on these operating quantities a centrifugal compressor 1 is controlled. In this case, after parameter values for setting these characteristic functions have been set, flow rate and efficiency are actually measured and predicted. And when the difference between the predicted value and the actually measured value is greater than a prescribed value, the parameter value is corrected on the basis of the calculated parameter correcting value, so as to bring the predicted value close to the actual value, and respective characteristic functions are changed and settings are revised. Thus, the changes of conditions or secular changes are detected as the changes of flow rate and efficiency, and the changes can be automatically followed up.

Description

Detailed Description of the Invention [Field of Industrial Application] The present invention relates to centrifugal compressors such as raw air compressors, factory air source compressors, and gas compressors for chemical plants used in oxygen production plants and various plants. The present invention relates to a learning control method for controlling the capacity and pressure of .

[Prior Art] Generally, a centrifugal compressor having a multi-stage configuration is used in an oxygen production plant or various plants. In such a multistage centrifugal compressor, as shown in Fig. 8,
The centrifugal compressor 1 includes a first stage compressor 4, which is driven by a power transmission gear 3 that speeds up rotation from a drive machine 2. 2nd stage compressor 5° 3rd stage compressor 6 and 4th stage compressor 7
In addition, an intercooler 8 is installed between the compressors 4 and 5.

An intercooler 9 is provided between the compressors 5 and 6, and an intercooler 10 is provided between the compressors 6 and 7.

Note that the compressors 4 and 5 and the compressors 6 and 7 are respectively overhanged on the same shaft end.

In such centrifugal compression [1], the air sucked into the first stage compressor 4 is sequentially compressed and cooled by each compressor 5 to 7 and intercooler 8 to 10, and is then compressed and cooled by the fourth stage compressor 4. From the compressor 7 it is sent to the process.

Variable angle inlet guide vanes (GV) 11 to 14 are provided on the inlet side of the compressors 4 to 7 in each stage, and by adjusting the angles of these inlet guide vanes 11 to 14, each The air capacity flowing into the compressors 4 to 7 can be adjusted.

Further, diffuser vanes (DV) 15 to 18 are provided on the outlet side of the compressors 4 to 7 in each stage.

By adjusting the angles of these diffuser vanes 15 to 18, the volume of air flowing out from each compressor 4 to 7 can be adjusted.

The angles of these inlet guide vanes 11 to 14 and diffuser vanes 15 to 18 are each adjusted to arbitrary values by a drive device 19.

Furthermore, this centrifugal compressor 1 as a whole or each stage compressor 4
Operating state quantities 1 to 7, for example, air flow rate, temperature, pressure, etc., are determined by the flow rate sensor 20. Temperature sensor 2
1. It is detected by a detection means such as the pressure sensor 22. A control device 23 is provided between each of the sensors 20 to 22 and the drive device 19.

Conventionally, a control means as shown in FIG. 9 has been disclosed in order to control the multi-stage centrifugal compressor as described above so that a predetermined air capacity (flow rate) can always be obtained with optimum operating efficiency according to various operating conditions. (Japanese Unexamined Patent Publication No. 55-60692).

In this control means, as a manipulated variable for realizing an optimal operating state for various operating states expressed by air flow rate, temperature, pressure, etc. in the entire centrifugal compressor 1 or each stage of compressors 4 to 7, Entrance guide vane 11 for each stage
14 and the angles of the diffuser vanes 15 to 18 are programmed and stored in advance in the storage section of the control device 23 (step SO).

As shown in FIG. 9, when the control bag [1123 receives the state detection values from the sensors 20 to 22, it calculates and monitors the current operating state of the multistage centrifugal compressor from the detected values (step SL ), and corresponds to its operating condition.
A predetermined operation program (combined values of vane angles) is checked to find a manipulated variable that can achieve optimal operating efficiency, and this manipulated variable is output to the drive device 19 (step S2).

Thereafter, it is determined whether the operating state calculated based on the state detection values from the sensors 20 to 22 is the optimum operating efficiency state determined in advance (step S3). If it is determined that the operating efficiency state is optimal, control is terminated at that point and the vane angle in the selected operating program is maintained. Control output (vane angle) by operation program
correct and re-output (Step S4), and determine whether the efficiency has improved from the detected state value (Step 85).
.

In this way, the operating status of each stage is fed back, and the inlet guide vanes 11 of each stage are programmed in advance.
14 and diffuser vanes 15 to 18 is optimal or not, and adjust the vanes in the operation program so that the optimal operating efficiency can be obtained at all times in response to changes in operating conditions such as aging and performance changes. Control is performed to correct the combination of angles (step S6).

By the way, in order to control the multistage centrifugal compressor so that a predetermined air capacity can always be obtained with optimum operating efficiency according to various operating conditions, the inlet guide vanes 11 to 14 are used as described above.
In addition to means for controlling the angles of the diffuser vanes 15 to 18, a control means as shown in FIG. 10 has also been disclosed (Japanese Unexamined Patent Publication No. 56-66490). This control means, instead of controlling the angles of the inlet guide vanes 11-14 and the diffuser vanes 15-18,
Angle of inlet guide vane 11 (see Figure 8) and driver 2
As shown in FIG. 10, there is a drive device 19A that adjusts the angle of only the inlet guide vane 11, and a drive device 19A that controls the rotation speed of the drive machine 2 (see FIG. 8). A rotation speed sensor 24 and a drive machine control device 25 for controlling the rotation speed of the drive machine 2 are provided. Then, the inlet guide vane 11 is used as a manipulated variable to realize the optimum operating state for various operating states expressed by air flow rate, temperature, pressure, etc. in the entire centrifugal compressor 1 or the compressors 4 to 7 in each stage. An operation table for the optimal combination of the angle of ) in the control bag E.
! It is programmed and stored in the storage section 26 in 23A.

As shown in FIG. 10, when the control device 23A receives the state detection values from the sensors 20 to 22.24, the control device 23A determines the current operating state of the multistage centrifugal compressor (operating conditions, suction flow rate, etc.) based on the detected values. Calculate and monitor (step Tl)
In addition, in response to the operating state, especially when the suction flow rate changes due to a change in operating conditions, an operating amount that can realize R-optimal operating efficiency is determined based on the operation table in the storage unit 26, and the operating amount is is output to the drive device L9A and the drive machine control bag [25.

That is, the operating states of the drive machine 2 and the inlet guide vane 11 at the present time are compared with the operating table in the storage unit 26 (step T2), and it is determined from the comparison result whether or not the operating efficiency state is optimal ( Step T3).

At this time, if it is determined that the operating efficiency is optimal, the control is terminated at that point and the operating state (vane angle and rotation speed) is maintained; however, if it is determined that the operating efficiency is not optimal, the inlet guide Angle of vane 11 and driver 2
After the rotation speed is corrected based on the operation table, the obtained operation amount is output to the drive device 19A and the drive machine control device 25 (step T4).

[Problems to be Solved by the Invention] However, in the conventional centrifugal compressor control means as described above, the combination of the angles of the inlet guide vanes 11 to 14 and the angles of the diffuser vanes 15 to 18, or the inlet guide The combination of the angle of the vane 11 and the rotational speed of the drive machine 2 must be set in advance to set such a combination.

A considerable trial run period is required before or at the initial stage of operation of the centrifugal compressor, and in particular, setting the optimum combination requires a huge amount of data, which is considered impossible to set in reality.

Furthermore, if the optimum operation amount (optimum combination) changes due to changes in factors other than the detectable operating state, such as temperature and pressure, it is not possible to cope with the change.

Furthermore, if the characteristics of the centrifugal compressor change due to changes over time, the above combinations programmed in advance will no longer be able to achieve the optimal operating conditions, so it is necessary to perform a trial run as described above again and set a new combination. There must be.

The present invention aims to solve these problems, and makes it possible to automatically follow changes in conditions and changes over time without requiring a test run period or a huge amount of data as described above. The purpose of the present invention is to provide a learning control method for a centrifugal compressor that can always maintain optimal operating conditions.

[Means for Solving the Problems] Therefore, the learning control method for a centrifugal compressor of the present invention is such that the angles of the inlet guide vane and the diffuser vane are adjusted on the vane angle plane determined by the angles of the inlet guide vane and the diffuser vane. The vane angle operation amount for each of the above-mentioned inlet guide vane and diffuser vane is determined based on the flow rate characteristic function and efficiency characteristic function that are set in advance to give the flow rate and efficiency according to these angles. In the method of controlling a centrifugal compressor based on the vane angle operation amount, the parameter values for setting the above-mentioned flow rate characteristic function and efficiency characteristic function are given in advance, and then the flow rate and efficiency in the above centrifugal compressor are actually measured. and predict based on the above flow rate characteristic function and efficiency characteristic function,
Compare the actual measured values and predicted values obtained for each of flow rate and efficiency, and if the difference between these actual measured values and predicted values is greater than a predetermined value, select a parameter that will bring the predicted value closer to the actual measured value. After calculating the correction amount for each of the above-mentioned flow rate characteristic function and efficiency characteristic function, the above-mentioned parameter values are corrected based on the above-mentioned parameter correction amount, and the above-mentioned flow rate characteristic function and efficiency characteristic function are respectively changed and set again. be.

[Embodiments of the Invention] Hereinafter, a method for controlling a centrifugal compressor as an embodiment of the present invention will be explained with reference to the drawings. FIG. 1 is a flowchart thereof. First, before explaining the method according to the present embodiment, The configuration of a centrifugal compressor to which the method of this embodiment is applied will be explained with reference to FIG. Incidentally, in FIG. 7, the same reference numerals as in FIG. 8 indicate substantially the same parts, so a description thereof will be omitted. However, the seventh
In the multi-stage centrifugal compressor to which the control device of this embodiment shown in the figure is applied, the multi-stage centrifugal compressor shown in FIG. Although different from the compressor, the method of the present invention can also be applied to the multistage centrifugal compressor shown in FIG.

As shown in FIG. 7, in the multi-stage centrifugal compressor in this embodiment, inlet guide vanes (GV) 11 to 14 are driven by inlet guide vane drive devices W 19 a to 19 d, respectively, and diffuser vane (DV) 15 ~
18 are diffuser vane drive devices 19a to 1, respectively.
9h. Further, as a sensor, a flow rate sensor 20. Temperature sensor 21. In addition to the pressure sensor 22, a rotation speed sensor 24 for detecting the compressor rotation speed and a humidity sensor 27 are provided.

The detection signals from the sensors 20 to 22, 24, and 27 are all input to a control device 28, which controls the inlet guide vane in order to control the flow rate in the centrifugal compressor 1. In order to adjust the angles of the diffuser vanes 11 to 14 and the diffuser vanes 15 to 18, appropriate control signals can be calculated and output to the respective drive devices 19a to 19h upon receiving each detection signal.

Next, the case where the learning control method of this embodiment is applied to the above-mentioned centrifugal compressor will be explained using FIGS. 1 to 6. First, before you start controlling.

Inlet guide vane 11 is stored in the storage section of control bag ff128.
~14 and the angle of the diffuser vanes 15-18 (α
, β), the flow rate characteristic function Q = which is a function of the vane angles (α, β) and gives the flow rate Q and efficiency η according to these vane angles (α, β). f□(α.

β) and efficiency characteristic function η=f2(α, β) are each mathematically modeled and set in advance, and the surging line SL (shaded area in Fig. 4) defines the surging area (shaded area in Fig. 4). (see Figure 4) also g = G (α, β)
It has been mathematically modeled and set in advance as 20. Note that the flow rate characteristic function f□ and the efficiency characteristic function f2 are as follows:
These are shown by dotted lines and solid lines in FIG. 4, respectively, and can be easily determined by selecting appropriate samples and approximating them based on short-term actual machine operation data.

In the learning control method in this embodiment, as shown in FIG.
The flow rate of the centrifugal compressor 1 is controlled by the control device 28 by implementing a control method called the hill-climbing method. Note that, as described later, steps A19 to A21 are for carrying out the surging elimination method, and step A
22 to A24 implement a surging avoidance method.

When controlling the flow rate of the centrifugal compressor 1, first, a plurality of parameter values P are given to set the flow rate characteristic function f1 and the efficiency characteristic function f2.

In this embodiment, the efficiency characteristic function f2 will be explained below, but the flow rate characteristic function f0 will be treated in the same way. That is, the efficiency η is set as f, (α, β, P). Here, the shape of the efficiency characteristic function f2 is approximated as an ellipse with equal efficiency lines, for example, as shown in FIG. 2, and the efficiency characteristic function f2 is expressed by a quadratic expression of α and β. In this case, the parameter value P is
As shown in FIG. 2, α. ,β. , σ□, σ2. θ,C
There are six [P=(α., β., σ,, σ2.θ c
) T co, by these the position of the efficiency characteristic function f2,
size.

Shape etc. are represented.

In this way, the flow rate characteristic function f□ and the efficiency characteristic function f
After giving the parameter value P for 2 in advance, the target flow rate QFN is set (step Al), and learning control of the flow rate of the centrifugal compressor 1 is started. After this, first, the flow rate sensor 20. The flow rate Q and efficiency η are actually measured using the temperature sensor 21, pressure sensor 22, etc. (Step A2
), the predicted values Q and η of the flow rate and efficiency are determined from the current angles (α, β) of the inlet guide vanes 11 to 14 and the diffuser vanes 15 to 18, respectively, based on the above flow rate characteristic function f1 and efficiency characteristic function f2 ( Step A3
).

Then, in step A, compare the actual measured values Q, η obtained for the flow rate and efficiency with the predicted values 2, θ.
4. ), the difference +q−c>+ or 1η−1 between the measured values Q, η and the predicted values 6, δ is the predetermined value ε1.
It is determined whether ε2 or more (step A5). As a result of the determination, if it is smaller than the predetermined value L1T F2, the following steps 86 to A8 are skipped and the process moves to the mountain climbing method, while
If the predetermined value fil F2 or more, steps 86 to A8
Therefore, (:)>Q.

When 〉η, the parameter value P in each characteristic function f□tL is corrected so that the predicted value Qrn becomes small, and Q
When < Q r q < η, the predicted value is 6. The parameter value P is corrected so that the number of times is increased.

That is, first, in step 86, the parameter value P
The influence coefficient on the predicted value when changing is af, C
It is obtained by calculating α, β, P)/aP. In this example, for example, af2/aα. 〉0, then α
. The predicted value can be increased by increasing . Therefore, it is possible to determine in which direction the parameter value P should be modified to reduce the difference between the predicted value and the actual value η, using the influence coefficient af27aP.

Next, the difference between the predicted value and the measured value η and the influence coefficient af,
/aP, the parameter correction amount ΔF is determined as shown in the following equation (1) (step A7).

Here, K is a gain matrix for correction, which is most simply given in advance as a 6×6 diagonal matrix whose diagonal components are positive numbers □~.

Then, the efficiency characteristic function fi (the same applies to the flow rate characteristic function f1) is changed and reset by correcting the parameter value P to P+ΔF based on the parameter correction amount ΔP determined by the above formula (step A8).

In addition, in order to always keep the shape of the efficiency characteristic function f2 in a physically valid shape, each parameter α. ,β. ,σ1゜σ2T”
Set a tolerance range for lc, and set the parameter value P
may not exceed that range. For example, for parameter C, c win≦C≦a max
If the above-mentioned modification of the parameter value P results in c)cmax, then C is set as QU
Correct it to aaX so that it does not exceed the set range.

After learning the state and changing the flow rate characteristic function f1 and the efficiency characteristic function f2 as described above, vane angle control is performed using the hill climbing method. ~
18A current angle (α.

β) is measured (step A9). Then, from the measured vane angle (α, β), it is determined whether or not this current angle is within a preset surging region (Step A10). If it is within the surging region, [G(α, β)
<When O], the process moves to step A19, which will be described later. On the other hand, if it is not within the surging area, the next step All
Move to. At this time, the current flow rate QNow at the vane angle (α, β) is determined by a preset flow rate characteristic function f□
Predict by

If the current vane angle (α, β) is not within the surging region, first, the first vane angle operation amount vector (Δα1.Δβ,) toward the target flow rate QFIN is determined as follows. In other words, the constant flow rate curve Q(
α, β) - QNow, the target flow rate QFIN in the normal direction of this equal flow rate curve (af, /aα, af□/aβ)
If the vane angle (α, β) is driven in the direction toward the side, the shortest operation path can be obtained, so this normal direction vector, which is directed toward the target flow rate QFIN, is the first vane angle operation amount vector (Δα1 .Δβ1).

Here, the vane angle operation amount vector (Δα□, Δβ□)
Letting the flow rate corrected based on ΔQ, ΔQ=(af,/aα)Aa, +(af,/aβ)A1
3, ... (2), and the vane angle operation amount Δα1. Since Δβ1 is proportional to the difference between the target flow rate Q FIN and the current flow rate Q Now, considering the above normal direction, ...(3) The first vane angle operation amount vector (Δα,

Δβ, ) can be obtained. Note that K means a tuning parameter.

In this way, after the first vane angle operation amount vector (Δα0.Δβ,) is obtained in step All, the current vane angle (α, β) is in the vicinity of the surging line SL shown in FIG. It is determined whether or not (step A12). Then, if it is in the vicinity of the surging line SL [when G (α, β) <ea; however, eQ is a positive value close to 0], the process moves to step A22, which will be described later, while if it is not in the vicinity of the surging line SL Then, the process moves to the next step A13.

In step A13, the efficiency is increased without changing the flow rate with respect to the first vane angle operation amount vector (Δα□, Δβ,) determined to approach the target flow rate QF engineering N by the shortest path. The second vane angle operation amount vector (Δα2°Δβ2) is determined so that In other words,
As in the case of flow rate control in step All described above,
The current vane angle (
Regarding the iso-efficiency curve η(α, β)=ηNOW in α, β), the normal direction of this iso-efficiency curve (af, /aα, a
f2/aβ) and the vane angle (α, β) is driven in a direction in which the improved efficiency Δη becomes positive, the efficiency η is improved.

Here, the second vane angle operation amount vector (Δα2゜Δ
The efficiency Δη modified based on β2) is Δη=(c'
f z / acc ) A αz + (δf, /
aβ)Δβ2...(4) It is expressed as follows. At this time, in order to prevent the flow rate from changing due to efficiency improvement operations, the vane angle operation amount Δα2
．． In order to make the flow rate change ΔQ due to Δβ2 zero, from equation (2), (af , /aa)A a, +caf , /aβ)Δβ
,=0...(5) The vane angle operation amount Δα2. Determine Δβ2. That is, the tangential direction vector of the above-mentioned constant flow rate curve is such that the improvement efficiency Δη>0 is given by (4).

When calculated from equation (5), the second vane angle operation amount vector (Δα2゜Δ
β2) can be obtained. Note that K2 indicates a tuning parameter.

Then, the first vane angle operation amount vector;~le(Δα□,
) and the second vane angle operation amount vector (Δα
2. By adding Δβ2), the flow rate becomes the target flow rate Q
A third vane angle operation amount vector (Δα3.Δβ3) that can improve efficiency while approaching the F-engine N is determined and output as a vane angle operation amount vector (Δα, Δβ) (step A14).

That is, calculations based on the following equation (7) are performed.

After this, each vane angle operation amount Δα is determined.

If Δβ exceeds the maximum allowable operating amount of the inlet guide vanes 11 to 14 and the diffuser vanes 15 to 18, that is, the angle that can be driven at one time may be exceeded, so the vane angle operating amount Δα or Δβ , the fifth
Substitute it into the limiter function FL(X) as shown in the figure, and reduce the vane angle operation amount Δα or Δβ if it exceeds the maximum allowable operation amount (step A1
5).

Furthermore, when the vane angle (α, β) is driven based on the vane angle operation amount vector (Δα, Δβ) obtained up to step A15, the new vane angle ( α+Δα.

When it is determined that β+Δβ) falls within the surging region, this vane angle operation amount vector (Δα.

Δβ) is reduced and output (step A16)
.

The vane angle operation amount vector 1- (
The drive bag [19
Output a control signal to a to 19h and inlet guide vane 11
14 and diffuser vanes 15 to 18 are driven and controlled (step A17).

Then, it is determined whether the flow rate changed by one or more vane drive controls has reached the vicinity of the target flow rate QFN (Step A18), and if it has not reached it, the process returns to Step A2 again and the learning control and hill climbing method are performed. While the vane angle control is continued, if it is determined that the flow rate has reached the vicinity of the target flow rate QFIN, the flow rate control is terminated at that point. Here, the determination method in step A18 is as follows.

(i) Based on the current vane angle (α, β) obtained by the vane drive control described above, the current flow rate is predicted from a preset flow rate characteristic function f, and this predicted value Q and the target flow rate QFIN are If the comparison shows that the difference between the two is within a certain value Ea, that is, IQ-QrN1≦ε3, it is determined that the target flow rate has reached the vicinity of QFN.

(n) The vane angle operation amounts Δα and Δβ are 1Δα1≦ε.

And when 1Δβ1≦ε (when the operation is no longer performed), it is determined that the target flow rate QF engineering N has been reached.

(ni) The vane angle operation amounts Δα and Δβ are Δα + (Δ
β) 2≦ε.

When this happens, it is determined that the target flow rate QFIN has been reached.

As described above, the starting point of the vane angle (α
. , β, ), the vane angle control is performed so that the flow rate reaches near the target flow rate QFN and achieves high efficiency.

Next, the surging elimination method in steps A19 to A21 and the surging avoidance method in steps A22 to A24 will be explained. First, in step AIO, the current vane angle (α.

When β0) is within the surging region, CG(α, β) is 0
If it is determined that , the shortest operation route to eliminate surging is naturally the surge line S L [g = G (α, β) = O], as shown in Fig. 6 (a). Since the normal direction Ca g /aα, ag/aβ) is the direction toward the outside of the surging region, this normal direction vector, which is toward the outside of the surging region, is the vane angle operation amount vector for surging cancellation (Δα4.Δβ , ). Here, the amount of operation is proportional to the difference between the boundary line G (α, β)=0 and the current surging function value G (α., β.).

...(8) As the vane angle operation amount vector for eliminating surging (
Δα1. Δβ4) is obtained (step A19) and output as a vane angle operation amount vector (Δα, Δβ) (step A20).

Then, as in the case of step A15, the obtained vane angle operation amounts Δα and Δβ are
14 and diffuser vanes 15 to 18, the vane angle operation amount Δα or Δ
While reducing β.

Output a control signal to the drive devices 19a to 19h.

The inlet guide vanes 11 to 14 and the diffuser vanes 15 to 18 are driven and controlled (step A21). After this, the process returns to step A9 again, and in step AIO it is determined whether or not the current position is within the surging area, and the vane angle (
The above-described steps A19 to A21 are repeated until α, β) are outside the surging region.

Further, in step A12, the current vane angle (α
, β) is near the surging line SL, [G(α
, β) <ea; where eQ is a positive value close to O], first, the first vane angle operation amount vector (Δα1.Δβ□
) is heading into the surging region as shown in FIG. Skipping steps A13 and A14, the first vane angle operation amount vector (Δα,.

Δβ, ) is directly converted into the vane angle manipulated vector (Δα,
Δβ) Step A24) Step A15
By moving to , it is possible to quickly avoid the vicinity of the surging line SL without considering efficiency improvement.

On the other hand, in step A22, if it is determined that the first vane angle operation amount vector (Δα□, Δβ1) is heading into the surging region as shown in FIG. By removing the component toward the surging region from the first vane angle operation amount vector (Δα4.Δβ1) in this way, the vane angle operation amount vector for surging avoidance along the surging line (Δα5.Δβ
5) In search of.

Avoid surging. Here, the unit vector in the normal direction of the surging line SL heading into the surging area is n
Let's say Q.

...(9), and using this equation (9), the surging avoidance vane angle operation amount vector ( Δα
5. Δβ, ) is calculated.

...(10) , and are tuning parameters.

In this way, the first vane angle operation amount vector (Δ
α1. Δβi) to the surging avoidance vane angle operation vector (Δα1.
Δβ, ) is determined (step A23), and this vane angle operation amount vector (Δα6.Δβ,) is output as a vane angle operation amount vector (Δα, Δβ) (step A24).
By moving to step A15, the flow rate of the centrifugal compressor 1 is controlled without considering efficiency improvement and while avoiding entering the surging region.

Incidentally, since the procedure after step A15 is as described above, the explanation thereof will be omitted.

As described above, while eliminating or avoiding surging, according to the map of the flow rate characteristic function fi and the efficiency characteristic function f2 on the vane angle plane corrected in step A2-88 by the hill climbing method, Target flow rate QPIN
The flow rate is controlled.

As described above, according to this embodiment, if the flow rate characteristic function f□ and the efficiency characteristic function f2 are given at the start of operation, thereafter, there is no need to perform a trial run etc. when a change occurs as in the conventional case. , each sensor 2 detects state changes or changes over time.
Flow rate by 0-22 etc.

Since changes in efficiency can be detected and automatically followed, optimal operating conditions can always be maintained using the hill-climbing method. In addition, if we can actually measure the flow rate and efficiency, we can overcome the effects of conditions that cannot be detected online.

Changes in the flow rate characteristic function f1 and the efficiency characteristic function f2 can be corrected, and optimum operating conditions can be maintained at all times.

For example, when the efficiency characteristic function f2 set in advance at the start of operation is as shown in Fig. 3(a), the temperature drops and the actual centrifugal compressor characteristics become as shown in Fig. 3(c). According to this embodiment, step A2
By learning through A8, the efficiency characteristic function f2 can be modified as shown in FIG. 3(b).

In addition, when implementing the control method of the above embodiment, the inlet guide vanes 11 to 14 and the diffuser vanes 15 to 18 of all stages may be controlled separately according to the above control method for each stage, or the control method of all stages may be separately controlled. The inlet guide vanes 11 to 14 and the diffuser vanes 15 to 18 may be driven and controlled at the same angle. In addition, the entrance guide vane 11~
14 and the angles of the diffuser vanes 15 to 18 are set using a set of dimensionless inputs, logarithmic vane angles, and vane angles such that the operating flow rate of the centrifugal compressors 4 to 7 in each stage has a similar operating flow rate at the same ratio to the design flow rate. The angles of the inlet guide vanes 11 to 14 and the diffuser vanes 15 to 18 of each stage may be expressed as a dimensional diffuser vane angle, and the vane angle may be controlled by regarding the angles of the inlet guide vanes 11 to 14 and the diffuser vanes 15 to 18 as one set.

In addition, if the operating conditions such as temperature, pressure, humidity, etc. or centrifugal compressor characteristics are thought to be changing, the parameter value P is corrected during operation using the learning control method in steps A2 to A8 of the above embodiment. However, if the characteristics of the centrifugal compressor are sufficiently stable and sufficient measured values under stable conditions have been obtained, it is better to perform control using the least squares method etc. based on the accumulated data. The parameter value P may be newly calculated by using the parameter value P.

Furthermore, to avoid erroneous learning due to abnormal data,
If the difference between the predicted value and the actual value becomes extremely large,
The data may not be used, or it may be determined that an abnormality has occurred in the device.

Furthermore, although the above embodiment shows the case where the method of the present invention is applied to a multi-stage centrifugal compressor, the method of the present embodiment can also be applied to a single-stage centrifugal compressor in the same way. It can be applied not only to flow control but also to pressure control.

[Effects of the Invention] As described above, according to the present invention, the flow rate and 9h rate predicted by the flow rate characteristic function and the efficiency characteristic function are compared with the actually measured flow rate and efficiency, respectively, and these predicted values are calculated. If the difference between the actual value and the measured value is greater than or equal to a predetermined value, the parameter values for the flow rate characteristic function and efficiency characteristic function are corrected so that the predicted value approaches the actual measured value. Once the flow rate characteristic function and efficiency characteristic function are given, the system automatically detects state changes or secular changes as changes in flow rate and efficiency, without having to perform test runs when changes occur as in the past. It is possible to follow these changes and has the effect of always maintaining optimal operating conditions.

[Brief explanation of drawings]

Figures 1 to 7 show a learning control method for a centrifugal compressor as an embodiment of the present invention. Figure 1 is a flowchart, and Figure 2 is a diagram for explaining parameter values for the efficiency characteristic function. Graphs, FIGS. 3(a) to (Q) are graphs showing numerical calculation examples for explaining the effects of the method of the present invention, and FIG. 4 is a graph showing preset flow rate characteristic functions, efficiency characteristic functions, and surging line Graph showing model example, 5th
The figure is a graph showing an example of the limiter function, Figure 6 (a)
is a graph for explaining a method for eliminating surging, and FIG. 6(b) is a graph for explaining a method for avoiding surging.
FIG. 7 is a block diagram showing a centrifugal compressor to which the method of this embodiment is applied, FIG. 8 is a block diagram showing a conventional general multi-stage centrifugal compressor, and FIG. 9 is a block diagram showing a conventional centrifugal compressor. Flowchart for explaining control means. FIG. 10 is a flowchart for explaining control means for another conventional centrifugal compressor. In the figure, 1-centrifugal compressor, 11-14--population guide vane, 15-18-diffuser vane, 19a-
19d-Population guide vane drive device, 19f-19h-
Diffuser vane drive. 20-flow rate sensor, 21.-temperature sensor, 22.-pressure sensor, 24-rotation speed sensor, 27.-humidity sensor, 2
8-Control device. Fig. 2 f (praise β) = (C (-((・・β−β・)(to) 爲こ Beast Magazine Sleeping)・CN4 Fig. 5 Fig. 6 (G) β (b) 71st!O tu 12 kQi 2 now 2' SI!I 9th cause

Claims (1)

[Claims] For a centrifugal compressor that has variable-angle inlet guide vanes and diffuser vanes on the inlet and outlet sides, respectively, the angles of these inlet guide vanes and diffuser vanes on the vane angle plane determined by the above angles of the inlet guide vanes and diffuser vanes. Determining the vane angle operation amount for each of the above-mentioned inlet guide vane and diffuser vane based on a flow rate characteristic function and an efficiency characteristic function that are a function of angle and are preset to give flow rates and efficiency according to these angles, In the method of controlling the centrifugal compressor based on the determined vane angle operation amount, parameter values for setting the flow rate characteristic function and efficiency characteristic function are given in advance, and then the flow rate and efficiency in the centrifugal compressor are controlled. are measured and predicted based on the above flow rate characteristic function and efficiency characteristic function, and the actual and predicted values obtained for flow rate and efficiency are compared to determine the difference between these measured values and predicted values. If the value is greater than or equal to the predetermined value, the parameter correction amount that brings the predicted value closer to the actual measured value is calculated for each of the flow rate characteristic function and the efficiency characteristic function, and then the parameter value is corrected based on the parameter correction amount. A learning control method for a centrifugal compressor, characterized in that the above-mentioned flow rate characteristic function and efficiency characteristic function are respectively changed and reset.