JPH1165610A - Control parameter automatic adjusting method for cascade loop - Google Patents

Abstract

PROBLEM TO BE SOLVED: To enable use of this system in a short time, without grasping the control system of a whole plant by selecting a simulation pattern equivalent to a real pattern obtained by a control time and a controlled variable, deciding the increasing and decreasing direction and increasing and decreasing amounts of plural control parameters, and adding significance for obtaining a weighted mean. SOLUTION: A simulation pattern is prepared by the simulation of an objective plant or a step response test 1 of a real machine, and automatic learning is operated by an automatic learning algorithm 3. In this case, the optimal pattern of a control time and a controlled variable has been set viously, plural control parameters ire increased or decreased so that a relation between the control time and controlled variable of the objective plant can be set as plural simulation patterns. Then, the objective plant is operated, the simulation pattern equivalent to a real pattern obtained from the relation between the control time and the controlled variable, increasing and decreasing direction and the increasing and decreasing amounts of the plural control parameters are decided from the selected simulation pattern, and significance is added for obtaining a weighted mean.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

The present invention relates to a method for automatically setting control parameters of a control controller.

[0002]

2. Description of the Related Art In a thermal power plant, functions such as high temperature and high pressure, full automation, and diversification of fuels are required to meet needs such as high efficiency, labor saving, diversification of operation, improvement of maintainability, and environmental protection. By adding it, the operation is advanced. In addition, as the number of coal-fired thermal power plants using various types of overseas coal as fuel increases, there is a need for a control system that can flexibly cope with changes in boiler characteristics caused by moisture content, crushability, ash adhesion characteristics, etc., depending on the type of coal. I have.

[0003]

In order to satisfy these demands, the control system of a thermal power plant has also been advanced,
Although advanced functions are being promoted, P
The one using an I (proportional / integral) controller is still mainstream, and the control parameters (proportional gain P and integration time I) of this PI controller have been conventionally set mainly based on the experience of the adjuster.

[0004] Therefore, there are the following problems in the trial operation adjustment in the thermal power plant. Skilled coordinators are indispensable to complete the commissioning smoothly in a short period of time.However, the shortage of skilled coordinators is increasing due to the increase in the number of plants and the increase in plant output. Operation may be affected. Plant-wide control system (APC: Automatic Pow
er Plant Control System) has become more sophisticated and sophisticated, and the coordinator cannot understand all the details. To reduce costs, there is a strong demand for shortening the trial run period.

[0005] Further, even after the trial operation adjustment, the following problems occur. Since the coal type changes in a short period (several weeks to several months) and the moisture content, crushability, ash adhesion characteristics, fuel ratio, etc. differ depending on the coal type, it is necessary to readjust the control parameters every time the coal type is switched is there. Even if the coal type is the same, the water content of the coal will change depending on the weather (rain, snow, etc.)
, Which changes the coal output characteristics. Coal operation involves not only one type of coal but also coal blending, so the properties of coal after coal blending cannot be clearly understood. Coal goes through a coal yard, bunker, belt conveyor, mill, etc., so it is not clear at what point the coal type was switched.

In order to solve the various problems described above,
There has been a strong demand for automatic setting of the control parameters of the control controller, and various means have already been proposed (for example, Japanese Patent Application Laid-Open Nos. 4-294402 and 5-5-).
No. 143113, JP-A-5-173605, JP-A-5-173605
-324011, JP-A-6-1311007, JP-A-7-268006, etc.).

However, such conventional means, for example, the "gain adjusting device for the rotational speed PID controller of an internal combustion engine" disclosed in Japanese Patent Application Laid-Open No. Hei 5-173605, uses a fuzzy logic and a learning function. Described in a formal compound proposition, each of a number of rules was created using human experience and knowledge. For this reason, conventional means have been replaced with specific plants (for example, thermal power plants).
In order to apply this, it was necessary to accurately analyze the know-how of a skilled coordinator, create a large number of logical expressions, and create appropriate instructions for each.

[0008] Therefore, in the conventional means, know-how of a skilled coordinator is indispensable, it is necessary to grasp the control system of the whole plant, and the water content, crushability, ash adhesion characteristics, I according to the characteristics such as fuel ratio
It is necessary to appropriately set the set amount in the compound propositions of the F-THEN-form, and there is a problem that a long time is required for the trial run and the readjustment.

More particularly, in a cascade loop in which a plurality of operation terminals are controlled by a plurality of measurement variables, a generator output, a main steam pressure, and a main steam temperature are changed by changing a control parameter of one operation terminal (for example, a boiler master). , reheat steam temperature, since a plurality of measurement variables, such as the exhaust gas O ₂ is affected, control characteristics of the plant becomes more complex, be brought close to the optimum pattern respectively in a short time is very difficult.

The present invention has been made to solve such a problem. That is, an object of the present invention is to provide a system which can be easily used in a short time without requiring a grasp of a control system of the entire plant, and does not require a skill level, and can bring an approximate pattern to an optimal pattern in a short control step in a short time. It is an object of the present invention to provide a method for automatically adjusting control parameters.

[0011]

According to the present invention, there is provided a method for automatically adjusting a plurality of control parameters of a control controller in a cascade loop for controlling a plurality of operation terminals by a plurality of measurement variables, comprising the steps of: For the variables, (A)
(B) setting an optimum pattern of the control time and the control amount in advance;
By increasing or decreasing the plurality of control parameters, respectively, setting the relationship between the control time and the control amount of the target plant as a plurality of simulation patterns, (C) operating the target plant,
An actual pattern is obtained from the relationship between the control time and the control amount, a simulation pattern corresponding to the actual pattern is selected, and (D) the increasing / decreasing directions and the increasing / decreasing amounts of a plurality of control parameters are determined from the selected simulation pattern. (E) A method for automatically adjusting the control parameters of a cascade loop, characterized in that the increasing direction and the decreasing amount of each control parameter for each measured variable are weighted and averaged in consideration of importance.

According to this method, for each measured variable (for example, generator output, main steam pressure, main steam temperature, reheat steam temperature, exhaust gas O ₂ ), the control pattern of (A) and the optimal pattern of the control amount are determined in advance. Just by setting, the steps (B) to (D) can be automatically set by using a computer, so that anyone can use it easily in a short time without requiring skill. Note that this optimum pattern may be common to each measurement variable. In the step (E), the weighted averaging is freely performed in consideration of the degree of importance, so that the entire measured variable can be controlled in a well-balanced manner. Further, in this method, if the simulation pattern and the actual pattern can be obtained from the target plant, it is not necessary to further grasp the control system of the entire plant. Therefore, the present invention is not limited to the above-described thermal power plant or the like, and can be applied to other general plants or apparatuses using the control regulator of the cascade loop.

According to a preferred embodiment of the present invention, the setting of the simulation pattern is based on a simulation of a target plant or a step response test of an actual machine. That is, when a high-accuracy simulation system is established, such as in a thermal power plant, a simulation pattern can be set in an extremely short time by using this system. Further, even when a simulation system has not been established, by performing a step response test using an actual machine, a simulation pattern can be accurately set without having to grasp the control system of the entire plant.

It is preferable that the time constant of the target plant is calculated, and the time constant is divided by a predetermined number to set a control time interval. The calculation of the time constant may be either a simulation or a step response test of a real machine.
By setting the control time interval according to the time constant in this manner, the same characteristics can be obtained with substantially the same number of steps without being affected by the plant characteristics.

In the simulation pattern, the area deviation, the degree of attenuation, the degree of overshoot, and the degree of vibration are set as characteristic amounts, and each characteristic amount is preferably subjected to fuzzy evaluation using a membership function. In this way, by performing fuzzy evaluation of the area deviation, the degree of attenuation, the degree of overshoot, and the degree of vibration, which are important features in the plant, using the membership function, it is possible to effectively approach the optimum pattern in a short time.

When the target plant is a thermal power plant, the control controller is a PI controller, and the plurality of measured variables are generator output, main steam pressure, main steam temperature, reheat steam temperature, exhaust gas O ₂ , for each measured variable,
(A) According to the fuzzy evaluation, evaluation values of “small”, “medium”, and “large” are determined according to the magnitudes of the three feature amounts of area deviation, attenuation degree, and overshoot, and a plurality of simulation patterns are selected. (B) The increase / decrease direction of the proportional gain P and the integration time I is determined as increasing, holding or decreasing according to each of the simulation patterns, and (C) further, “small” and “medium” according to the magnitude of the vibration. The evaluation value of “large” is determined, and the increasing / decreasing direction of the proportional gain P is determined as increasing, holding, or decreasing corresponding to the evaluation value. (D) The results of B and C are added, and the proportional gain P and the integral are integrated. The direction in which the time I is increased or decreased and its amount are determined, and then (E) the increasing and decreasing directions and the amount of each control parameter for each measured variable are weighted and averaged in consideration of the importance.

The control regulator in the thermal power plant is mainly a PI regulator, and the direction of increase or decrease of the proportional gain P and the integration time I, which are control parameters of the PI regulator, is determined from a plurality of simulation patterns. By adding and determining the increasing and decreasing directions of the proportional gain P according to the magnitude of
It is possible to approach an optimal pattern in a short time in a short control step without performing a huge and complicated control as in modern control theory.

[0018]

DESCRIPTION OF THE PREFERRED EMBODIMENTS Preferred embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram showing a learning / adjustment procedure according to the method of the present invention. As shown in this figure, the method of the present invention is roughly divided into a learning procedure and an adjustment procedure. In addition, the method of the present invention automatically adjusts a plurality of control parameters of a control controller in a cascade loop that controls a plurality of operating terminals by a plurality of measurement variables. The cascade loop in the thermal power plant includes a turbine master, a boiler master, a water-fuel ratio master, an air-fuel ratio master, reheat steam temperature control, and the like.

In FIG. 1, in the learning procedure, a simulation pattern is created / set at 2 by a simulation of a target plant or a step response test of an actual machine, and automatic learning is performed at 3. In the adjustment procedure, by calculating the time constant of 4,
The time constant of the target plant is calculated, and the time constant is divided by a predetermined number to set a control time interval. Further, in the fuzzy inference of 5, the control parameters such as the proportional gain are automatically adjusted based on the result of the automatic learning.

In this way, when the simulation pattern is set by the simulation of the target plant or the step response test of the actual machine, a high-precision simulation system such as a thermal power plant is established and used. Thus, the simulation pattern can be set in a very short time. Further, even when a simulation system has not been established, by performing a step response test using an actual machine, a simulation pattern can be accurately set without having to grasp the control system of the entire plant.

Further, by setting the control time interval according to the time constant in this manner, the same characteristics can be obtained with substantially the same number of steps without being affected by the plant characteristics. The calculation of the time constant may be either a simulation or a step response test of a real machine.

FIG. 2 is a diagram schematically showing a response pattern for learning according to the present invention. In this figure, the response pattern located at the center is a desirable optimal pattern, and the eight patterns located around it are simulation patterns when the control parameters are increased or decreased. That is, in the automatic adjustment method of control parameters of the present invention, in the automatic learning of 3, first, (A) the optimal pattern (learning pattern 5) of the control time and the control amount is set in advance, and then (B)
A plurality of control parameters (proportional gain P and integration time I in this figure) are respectively increased and decreased, and the relationship between the control time and the control amount of the target plant is set as a plurality of simulation patterns (except 5; learning patterns 1 to 9). Set.

Next, (C) the target plant is operated, an actual pattern is obtained from the relationship between the control time and the control amount, and simulation patterns (learning patterns 1 to 9) corresponding to the actual pattern are obtained.
(D), the increasing / decreasing direction of the plurality of control parameters is determined from the selected simulation pattern in a direction opposite to the increasing / decreasing direction in (B), and the increasing / decreasing amount is simultaneously determined by fuzzy evaluation. In this description, “increase / decrease” includes a case where there is no increase / decrease.

The above (A) to (D) are performed for each measured variable. Note that the optimum pattern 5 may be common to each measurement variable. Next, (E) the weighted average of the increasing and decreasing directions and the increasing and decreasing amounts of each control parameter with respect to each measured variable is taken into consideration with importance. The weighted average can be freely set in consideration of the degree of importance, so that the entire measured variable can be controlled in a well-balanced manner.

Each pattern shown in FIG. 2 is shown as a change curve of the control time (horizontal axis) and the control amount (vertical axis). In actual control, this is shown in the upper right of each figure. In addition, it is preferable to set the characteristic values of the area deviation, the degree of attenuation, the degree of overshoot, and the degree of vibration. Further, each feature value is fuzzy evaluated using a membership function. Here, the membership function is, as exemplified in FIG.
An evaluation value is given to each feature quantity, and different evaluation values are given to both of the two sections at the boundary between the sections (in this case, “small”, “medium”, and “large”).

Next, the case where the target plant is a thermal power plant will be described in more detail. In this case, the control controller is a PI controller. Further, a plurality of measurement variable In this case, the generator output, the main steam pressure, main steam temperature, reheat steam temperature, a gas O _2. First, (A) area deviation,
Evaluation values of “small”, “medium”, and “large” are determined in accordance with the three feature amounts of the degree of attenuation and overshoot, and 27 types of simulation patterns shown in Table 1 are created in advance.

[0027]

[Table 1]

Further, the increasing and decreasing directions of the proportional gain P and the integration time I are determined in advance as shown in Table 2 corresponding to the 27 kinds of simulation patterns. Note that this correspondence corresponds to N in Table 1.
o. 1 to 9, 10 to 18, and 19 to 27 correspond to Table 2, respectively, and the degree is changed depending on the magnitude of the evaluation value by the membership function.

[0029]

[Table 2]

Further, the direction of increase and decrease of the proportional gain P is determined as shown in Table 3 according to the magnitude of the vibration.

[0031]

[Table 3]

In the actual automatic adjustment, in the automatic learning described above, first, (A) an optimal pattern of the control time and the control amount (learning pattern 5) is set in advance, and then (B) a plurality of control parameters (see FIG. Then, the proportional gain P and the integration time I) are respectively increased and decreased to obtain a plurality of simulation patterns (learning patterns 1 to 9 excluding 5) for the relationship between the control time and the control amount of the target plant. The feature amounts of the deviation, the degree of attenuation, the degree of overshoot and the degree of vibration are obtained and stored.

Next, (C) the target plant is operated, an actual pattern is obtained from the relationship between the control time and the control amount, each characteristic amount is obtained from the actual pattern, and each characteristic amount is compared with Table 1. Select the corresponding pattern. Next, the increasing / decreasing direction of the proportional gain P and the integration time I is determined from Table 2 as increasing, holding, or decreasing according to the selected pattern, and further, “small”, “medium”, “large” according to the magnitude of the vibration degree. Determine the evaluation value of
Correspondingly, the direction of increase / decrease of the proportional gain P is determined from Table 3 as increase, hold or decrease, and the results of Tables 2 and 3 are added to determine the direction of increase / decrease or hold of the proportional gain P and the integration time I. Determine its amount.

The above (A) to (D) are performed separately for each measured variable, but the patterns in Tables 1 to 3 can be used in common. Next, (E) the weighted average of the increasing and decreasing directions and the increasing and decreasing amounts of each control parameter with respect to each measured variable is taken into consideration with importance.

According to the above-described method, an optimal pattern can be approximated in a short time in a short control step without requiring a huge learning process such as a neural network. Further, the patterns in Tables 1 to 3 have versatility independent of the control target and can be applied to control of general industries other than thermal power plants. Further, in this method, if the simulation pattern and the actual pattern can be obtained from the target plant, it is not necessary to further grasp the control system of the entire plant. Therefore, anyone without the need for skill,
Can be easily applied in a short time.

[0036]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS In order to verify the above-described automatic adjustment method of control parameters of the present invention, an analysis system (energy plant dynamic characteristic analysis system) for a thermal power plant that has been established as a highly accurate simulation system will be described. A simulation was performed for a 1000 MW coal-fired variable-pressure once-through boiler. Note that a boiler master PI controller was used for the cascade loop of interest.

FIGS. 4 and 5 show the obtained results.
FIG. 5 is a diagram showing an automatic adjustment trajectory of a proportional gain / integration time according to the method of the present invention, and FIG. 5 is a diagram showing the automatic adjustment process. From these results, it was confirmed that the adjustment value reached the appropriate value by about 10 to 30 adjustments, and that the adjustment value was stable thereafter.

It should be noted that the present invention is not limited to the above-described embodiment, and it is needless to say that various changes can be made without departing from the spirit of the present invention.

[0039]

As described above, the method for automatically adjusting the control parameters of the present invention does not require an understanding of the control system of the entire plant, and can be easily used in a short time by anyone without requiring skill. In addition, there is an excellent effect that an optimum pattern can be brought close to the optimum pattern in a short control step in a short time.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a learning / adjustment procedure according to the method of the present invention.

FIG. 2 is a diagram showing a response pattern for learning according to the present invention.

FIG. 3 is an explanatory diagram of a membership function.

FIG. 4 is a diagram showing a trajectory for automatically adjusting a proportional gain and an integration time according to the method of the present invention.

FIG. 5 is a diagram showing a process of automatically adjusting a proportional gain and an integration time according to the method of the present invention.

[Explanation of symbols]

 1 Step response test of simulation or actual machine 2 Create / set simulation pattern 3 Automatic learning 4 Time constant calculation 5 Fuzzy inference

 ──────────────────────────────────────────────────続 き Continuing on the front page (72) Inventor Shigeki Murayama 3-1-1-15 Toyosu, Koto-ku, Tokyo Ishikawajima-Harima Heavy Industries Co., Ltd. Tojin Technical Center (72) Inventor Toshio Inoue Toyosu 3-chome, Koto-ku, Tokyo No. 16 Ishikawajima Harima Heavy Industries, Ltd. Toyosu General Office

Claims (5)

[Claims] 1. A method for automatically adjusting a plurality of control parameters of a control controller in a cascade loop for controlling a plurality of operation terminals by a plurality of measurement variables, wherein (A) control time and control for each measurement variable The optimal pattern of the quantity is set in advance, (B) the plurality of control parameters are respectively increased / decreased, the relationship between the control time and the control quantity of the target plant is set as a plurality of simulation patterns, and (C) the target plant is operated. Determining an actual pattern from the relationship between the control time and the control amount, selecting a simulation pattern corresponding to the actual pattern, and (D) determining the increasing / decreasing directions and the increasing / decreasing amounts of a plurality of control parameters from the selected simulation pattern; Next, (E) a weighted average of the increasing and decreasing directions and the increasing and decreasing amounts of each control parameter with respect to each measured variable, taking into account the importance.
A method for automatically adjusting a control parameter of a cascade loop. 2. The method according to claim 1, wherein the setting of the simulation pattern is performed by a simulation of a target plant or a step response test of an actual machine. 3. The method according to claim 1, wherein a time constant of the target plant is calculated, and the time constant is divided by a predetermined number to set a control time interval. 4. The simulation pattern according to claim 1, wherein the area deviation, the degree of attenuation, the degree of overshoot, and the degree of vibration are set as characteristic amounts, and each characteristic amount is subjected to fuzzy evaluation using a membership function. Automatic adjustment method of the control parameter described in. 5. When the target plant is a thermal power plant, the control controller is a PI controller, and the plurality of measured variables are generator output, main steam pressure, main steam temperature, reheat steam temperature, Exhaust gas O ₂ , for each measured variable,
(A) By the fuzzy evaluation, area deviation, degree of attenuation,
Evaluation values of “small”, “medium”, and “large” are determined in accordance with the magnitudes of the three overshoot feature amounts, and a plurality of simulation patterns are selected. (B) Proportion in accordance with each of the simulation patterns The increase / decrease direction of the gain P and the integration time I is determined as increasing, holding, or decreasing. (C) Further, evaluation values of “small”, “medium”, and “large” are determined according to the magnitude of the vibration degree, and corresponding thereto. (D) The results of B and C are added to determine the direction in which the proportional gain P and the integration time I are increased or decreased and the amount thereof. Then, (E) a weighted average of the increasing and decreasing directions and the increasing and decreasing amounts of each control parameter with respect to each measured variable, taking into account the importance,
The method for automatically adjusting a control parameter according to claim 4, wherein: