JPS58213285A - Method of controlling plasma - Google Patents

Abstract

(57) [Summary] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

[Detailed Description of the Invention] The present invention relates to a plasma control method in a tokamak-type nuclear fusion device, and in particular, a method for identifying a plasma dynamic characteristic model from plasma discharge data and optimizing control parameters based on this model. The present invention relates to a method for controlling a plasma by using the following method.

Conventionally, in tokamak-type fusion devices, various feedback controls such as plasma position control, plasma density control, and current distribution have been necessary for stable plasma confinement. However, some parts of the dynamic characteristics of the plasma, such as the interaction between the plasma and the limiter (zitniter), the behavior of impurities, and the heating mechanism, are still unclear, making it difficult to obtain an accurate dynamic characteristics model. The reality is that control parameters in a feedback control system cannot be uniquely determined. That is, in the past, it was necessary to conduct discharge experiments many times by trial and error in order to search for optimal control parameters.

For example, when optimizing the control parameters of 2fIl,
The discharge experiment is repeated several dozen times. Furthermore, discharge experiments are repeated every time the operating conditions of the fusion device (e.g., plasma current level) are changed, and when optimizing control parameters in this way, optimization cannot be performed economically. That is what it is. This is because, as nuclear fusion devices become larger, the cost required for one discharge experiment is becoming prohibitive. Further, when optimizing in this way, there is a problem that a lot of time is required until optimization.

Therefore, an object of the present invention is to provide a method of quickly optimizing control parameters by reducing the number of discharge experiments, and controlling plasma using control parameters other than t after optimization.

For this purpose, the present invention artificially applies various pulse disturbances to the control input to the fusion device main body, and calculates the difference between the response when the disturbance is applied and the response when no disturbance is applied. In this system, input and output discharge data are collected using the same method, and a large number of collected discharge data are combined into a series of time-series data to perform model identification, and control parameters are optimized based on this model. . In the past, V, for example, the reproducibility of discharge data was not good, as there was a variation of about 26% in the plasma position in a small tokamak type fusion device, and the discharge data was different every time there was a discharge. However, in the case of the present invention, it is possible to perform complete model identification with high accuracy.

Furthermore, in the conventional optimization of control parameters, the optimization operation necessarily involves the driver's empirical judgment and tends to take a lot of time. Tokamak-type fusion devices operate regularly, for example, by repeating electrical discharge (duration: 1 to 10 seconds) at 10-minute intervals, but optimization operations that involve the driver's empirical judgment are performed at intervals of 10 minutes. You need to make sure that it ends within. However, when the control parameters are interactively optimized based on the identified model, the optimization operation can be performed efficiently.

The present invention will be explained below with reference to FIGS. 1 to 4.

First, a plasma control system according to the present invention will be explained with reference to FIG. - As shown, the tokamak-type fusion device 5 is integrally equipped with a measurement data processing device (not shown) for controlling the plasma position, density, current distribution, etc. The data is fed back to the plasma control device [6] in a suitably processed form by the open side data processing device. Based on the fed-back measurement data, the plasma control device 6 performs feedback control of the tokamak-type fusion device 5. The plasma control system according to the present invention further includes a disturbance application device 7 and model identification/control parameter optimization. computing device 8
is now added.

Among these, the disturbance applying device 7I/Majitsu 9) is for applying various disturbances to the control input to the Kamak type nuclear fusion device 5. Further, the arithmetic device 8. Measurement data (discharge data) from the measurement data processing device Xh (result), Llh
Based on the collected discharge data, the plasma dynamic characteristic model is identified by the model identification unit 2.

The identified dynamic characteristic model is sent to the parameter optimization section 3, where the control parameters are optimized. In this case, in order to efficiently perform the optimization operation, the optimization of the control node is performed in an interactive manner using the CLT display device 4 having a keyboard manual imaging section. In this way, the #L-optimized control node ζ is set in the plasma control device gi, 6, and is used for optimal control of the plasma.

Next, the performance in the model identification section 2 will be explained as follows.

Month 1j, t-H shot of plasma (discharge 1,
The dynamic relationship between discharge data (human output data) (Jh, λ) in ) is expressed as a linear regression model as follows.

Xb'+1 = a, )'l 3 +・””・10a*Xh
'-8+ I) I Ll e -to-ten b, um
Otsu + e: ... (1) However, k is discrete time t
Discharge k at k to t. Further, ε indicates noise. By the way, the model parameter Φmo(a+,...Ha@+b+...
·, b, ) are determined so that fi& in the following equation (2) is minimized.

σt == , r, Σ [st]1...
(2)k+l However, N is the number of samples per shot, and fr-1
L indicates the number of shots.

The residual sum of squares is determined to be the minimum, and the model parameter Φ is obtained by solving the following equations (3) and (4). a kΦh) / CI
+a b Pka k)...(3) P k41 == P b -Pkak[Pbak]"/[i+aJPham)
...(4) However, a k=(XI',...
,X, LIkl...

LJbζ).

The model is identified by determining the model parameters Φ.Next, the performance performed by the parameter optimization section 3 will be explained.

In the parameter optimization section 3, two types of control parameters are optimized. - Human power - Control parameters (proportional gain, integral gain, differential gain) in PID (proportional/integral/derivative) control, which is a general control method for output, and optimization, which is a general control method for multi-input, multi-output systems. Control parameters in control (such as feedback gain matrix 1 are set to +A).

First, optimization of control parameters in P I l) control is performed by the following procedure.

1) From the identified model (formula (1)), calculate the frequency response spectrum A (f) using the following formula (5).

A(f) = Σ Cm exp(j :2r fm
)...(5) Then, C7 is the impulse response series of equation (1), υ, f
indicates the frequency.

2) Frequency response spectrum A (f) and Ito's equation (6
) of the plasma control device 6 shown in the transfer function Gc (81 and'
The spectrum of open-loop transfer numerical aperture obtained by this is G(S) = G c(S) A(81'fc C'
4 on the H and T displays. Note that in equation (6), 5=jzπf.

Gc(8) = C+p +0w/' + Go
8...(6) Accordingly, the driver uses proportional gain (P) and integral gain as the evaluation criteria for overall phase and gain margin.

(j!, which determines the differential gain Go.

Next, the procedure for optimizing control parameters in Kuyu control is as follows.

1) The model shown in equation (1) is transformed into a state equation shown as equation (7).

Y=CX X=AX+B U...(7
) However, Y, X, and IJ are each output vectors.

They are a state vector and an input vector, and X indicates the time differential of X.

2) The optimum jf value criterion 'fc shown in Equation (8) is determined by the control bar 2 notor matrix F-i Equation (9) e lWf <.

Cm (1/2) ・/; (Y TPY-t-Ll
"RU) dt...(8) Equation (8) corresponds to the most efficient control of the output to a constant value under the condition -F that the input fluctuation is as small as possible.

3) Simulate the closed-loop time response when using the optimal control parameter matrix F and calculate CI? It is displayed on the T display device 4. Based on this result, the driver adjusts the lookup matrices P, l(,) and repeats Step 2.

Now, finally, we will discuss the extent of the effects of the present invention.

First, FIG. 2 shows plasma position if measurement data (1) and vertical magnetic field coil current measurement data (n). As shown, the calculated buzzer position data (III) whose response was simulated using the model identified by equations (1) to (4) are also shown. In general, the plasma position is expressed as a horizontal shift from the vacuum center of the plasma ring, and the position is controlled to maintain zero by the vertical magnetic field generated by the current flowing in the vertical magnetic field coil. As you can see, the plasma position measurement data (1
) and the Keito Gramema position data (III) is 0.5
It can be seen that the model identification is carried out well as it is suppressed to within υ cm. In addition, the ring outer diameter and ring inner diameter of the ring-shaped vacuum container in this example are each 70 cyr+
, 30crn, and the plasma cross-sectional diameter is approximately 14
It was crri. This condition also applies to FIG. 4, which will be described later.

Figures 3(a) and (b) show the open-loop transfer function G (8) = G c (S) A (S
) is shown as an example. As mentioned above, the driving force is the proportional gain Gp, the integral gain Gl, and the differential gain G based on the overall phase, gain margin, etc. as evaluation criteria.

However, if the determination is made according to the displayed Bode diagram, the optimization operation becomes extremely easy.

Incidentally, in this example, optimization has been completed using a phase margin of 2π/3 as an evaluation criterion, and Gs=0.0. (3F = 1.
The spectrum related to 0 is the one before optimization, and Gx=
10.0 and Gp = 6.8 are those after optimization.

FIG. 4 shows the change in plasma position when a step disturbance is applied. Guzma position fM before optimization
, (IV) change as shown in the figure, but the plasma position (V) after optimization does not change in phase and the control 1 characteristic is significantly 1-
<I can see that it has been improved.

As discussed above, the present invention collects a lot of input/output discharge data by finding the difference between the response when a disturbance is applied and the response when no disturbance is applied, and then The data was combined with a series of time-series data to perform model identification, and the control parameters were optimized based on this model.The plasma in the fusion device was then feedback-controlled based on the optimized control parameters. This is what I did. In the past, the theoretical dynamic characteristics of plasma were unknown and it took a lot of time and cost to optimize control parameters as nuclear fusion devices became larger. Even if the theoretical dynamic characteristics of the platform are unknown, the control parameters can be optimized quickly and at low cost, and the plasma can be controlled well.

[Brief explanation of drawings]

Figure 1 is a diagram showing the configuration of a plasma control system according to the present invention, Figures 2 and 4 are diagrams for explaining the extent of the effects of the present invention, and Figures 3 (a) and (b). is C during the optimization operation.
FIG. 3 is a diagram illustrating an example of the frequency response of a one-loop transfer function displayed on an RT display device. DESCRIPTION OF SYMBOLS 1... Arithmetic device (for model identification/control parameter optimization), 2... Model identification section, 3... Parameter optimization section, 5... Tokamak type nuclear fusion device, 6... Plasma control Device, 7... Disturbance application device. Agent Patent attorney Masami Akimoto / ¥2 Figure $3 Eye (ku) (, b)

Claims (1)

[Claims] 1. A plasma control method that performs feedback control of the plasma in a tokamak legged fusion device is used to calculate the difference between the response when a disturbance is applied to the plasma control input and the response when no disturbance is applied. After collecting the output discharge data, identifying the plasma dynamic characteristic model by considering the data as a series of time series data, and optimizing the control parameters based on the model, the optimized city 1. A plasma control method characterized by feedback control of plasma based on several parameters.