JP2907672B2 - Process adaptive control method and process control system - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a process adaptive control method and a control system, and more particularly, to a process suitable for obtaining good control characteristics by adapting to process characteristics even when the process is a distributed constant system. The present invention relates to an adaptive control method and a control system.

[0002]

2. Description of the Related Art As a control characteristic of a process to be controlled, there is a process having a large response delay. For example, a control feature of a thermal power plant is a large delay in steam temperature response. For example, the time constant of the main steam temperature is about 10 to 20 minutes. Therefore, in the ordinary feedback control, when the load command largely changes, there is a problem that the main steam temperature greatly changes, the thermal stress of the turbine increases, and the life is shortened. For this reason, it is considered that the thermal power plant is difficult to control.

In order to cope with this, it has been proposed to incorporate a model of a thermal power plant in a control system, to predict the near future movement of the plant using this model, and to determine the operation amount based on the prediction result. Have been. In a control method that incorporates a model of a process and determines an operation amount using the model, it is necessary to construct a model and adjust parameters of the model. A method for that has been proposed. That is, (1) Y. Sato, et
al. , "Steam PredictionCont
rol for Thermal Power Pla
nt, "IEEE / PES 1984 Winter"
Meeting, Dallas, Texas. U. S.
A. January 29-February 3,1
984. , (2) Sato, et al., "Boiler Steam Temperature Predictive Control Using Kalman Filter," The 18th SICE (Society of Instrument and Control Engineers) Scientific Lecture, 1201, August 29-September 1, 1979, ( 3) Y. Sato, et a
l. , "Steam Temperature Pre
dictation Control for Therm
al Power Plant ”, IEEE Tran
s. on PowerApparatus and S
systems, Vol. PAS-103, no. 9, S
eptember (1984), pp. 146-64. 2382-23
87, there is a process adaptive control method.
In these documents, this process adaptive control method is applied to a thermal power plant. The prediction model used in these is a lumped constant model in which the characteristics of the last-stage superheater are represented by physical equations. This model predicts the outlet steam temperature of the final stage superheater, that is, the main steam temperature, with the outlet steam temperature of the previous stage superheater as a disturbance.

[0004]

By the way, recently, with the increase in power demand, the difference in power demand between day and night has been increasing. Therefore, even in a large-capacity thermal power plant, load following operation (intermediate load operation) and DSS (Daily S
(Start Stop) operation is required. With these demands, there is a demand for a thermal power plant control system to have improved control characteristics at startup and load following characteristics.

However, it is difficult for conventional predictive control to sufficiently cope with this. This is because the conventional predictive control approximates a thermal power plant with a lumped parameter model as described in the above-mentioned literature.

[0006] The thermal power plant comprises a furnace water wall, a primary superheater,
It is composed of a plurality of heat exchangers such as a secondary superheater and a tertiary superheater. While water flows from these upstream to downstream in these heat exchangers, it absorbs the energy of the combustion gas to become steam and further becomes superheated steam. That is, the thermal power plant is a distributed constant system. Nevertheless, in the prior art, the final stage superheater of the thermal power plant is approximated by a lumped parameterized model, the near-term future movement of the thermal power plant is predicted by this model, and the operation amount is determined based on the prediction result. I have decided.

Therefore, it is predicted that the steam at the outlet of the pre-heater
Predictive performance and controllability because temperature changes cannot be considered
There is a problem that the improvement of the quality cannot be expected. That is,
Is a lumped parameter for thermal power plants, which are originally distributed parameter systems.
Since the model is approximated, the characteristics of the plant can be accurately determined.
I can't simulate. Therefore, such a model
Forecasting the future movement of thermal power plants with
There is a limit to the degree of improvement, and there is also a limit to the controllability.

A first object of the present invention is to provide a process adaptive control method capable of controlling a process using a model capable of accurately simulating the characteristics of a plant composed of a distributed constant system.
Is to provide.

A second object is to form a distributed constant system.
Use a model that can accurately simulate plant characteristics.
An object of the present invention is to provide a process control system capable of controlling a process .

[0010]

Means for Solving the Problems] To achieve the above object, according to one aspect of the present invention incorporates a model of the process to obtain the predicted value of the controlled variable by using this model, on the predicted value An adaptive control method for a process that determines an operation amount based on the process, wherein the process is added to a flow of the process.
Along with two or more sub-processes
Processes are grouped individually based on physical formulas.
Consisting of medium constant model and for each sub-process
Kalman filter for the lumped parameter model
And the state quantity and
Estimate the error and use the model of the upstream sub-process.
The predicted value of the state quantity of the subprocess obtained by
At least one of the input variables of the model of the downstream subprocess
And the model for upstream sub-processes
The predicted value of the state quantity is calculated, and the predicted value of the state
Input to the model of the upstream sub-process and
The predicted value of the state quantity of the process is obtained, and this is
Until the predicted value of the state quantity of the process is obtained, and
Is the operation to obtain the predicted value of the state quantity of the upstream sub-process?
To calculate the predicted value of the state quantity of the sub-process at the most downstream position from
Repeat the series of calculations up to a predetermined number of times
The predicted state values obtained for each subprocess
State obtained at least for the lowest downstream subprocess
Determine the manipulated variable of the process based on the predicted value of the state parameter
An adaptive control method for a process is provided.

According to another aspect of the present invention, a processor
Process that determines the amount of operation of the application and controls the process
In a control system, the process is organized according to its flow.
Model of two or more subprocesses
Use these models to predict the state of a process
State prediction system, target values for processes, and
And, based on the predicted state quantity of the process,
Operation amount determination system that determines the operation amount for
The state quantity prediction system comprises:
Provided in advance as a model for each subprocess.
The predicted value of the state quantity is calculated using the physical equation
Measurement calculation means, and the centralized determination for each of the sub-processes.
It is configured for the numerical model, and the state quantity and its error
Kalman filter to be estimated and the two or more predicted values
The calculation means obtained by the predicted value calculation means on the upstream side
Calculates the predicted value of the state quantity of the sub process on the downstream side
Downstream prediction as at least part of the input variables of the instrument
The calculation is performed sequentially so that the value calculation means performs the calculation.
And a predicted value corresponding to the upstream sub-process
From the calculation for calculating the predicted value by the
For calculating the predicted value by the predicted value calculation means corresponding to the process
Is repeated a predetermined number of times.
Control to obtain the predicted value of the state quantity of the process.
Process adaptive control system characterized by having a step
Is provided.

[0012]

[0013]

[0014]

[0015]

[0016]

The process to which the present invention is applied includes, for example,
A thermal power plant can be envisaged. In this case, the model of each heat exchanger of the thermal power plant is composed of a lumped parameter model based on physical equations , and Kalman
-Since the filter is configured , the characteristics of each heat exchanger which is originally a distributed constant system can be accurately simulated. In addition, since the models of these heat exchangers are integrated and used to predict the movement of the thermal power plant in the near future, the prediction accuracy can be improved. Further, since the operation amount is determined based on the prediction result, controllability can be improved.

[0017]

[0018]

[0019]

[0020]

Embodiments of the present invention will be described below with reference to the drawings. First, an outline of a thermal power plant to be controlled in the present embodiment will be described. FIG. 2 is an explanatory diagram showing an example of a coal-fired thermal power plant.

The air from the pushing fan 101 is preheated through an air preheater 102, accelerated by a primary air fan 103, and sent to a coal mill 107. On the other hand, the coal in the coal bunker 104
Mill 10 by a coal feeder 106 driven by
7 to be transported. The coal pulverized by the coal mill 107 is supplied to the burner 12 in the boiler 126 together with the air flow.
7, where it burns.

Energy saving device (ECO) 130, furnace water cooling wall 10
8 (WW) and the water that has passed through the primary superheater (1SH) 109 are turned into steam by the combustion gas. This steam is 1
The next superheater (2S) passes through the next spray (SP1) 116.
H) passed through the secondary spray (SP2) 120 and further heated in the tertiary superheater (3SH) 111,
It enters the high-pressure turbine 122 through the main steam pipe and the main steam control valve 121. The steam exiting the high pressure turbine 122 is reheated by the primary reheater 112 and the secondary reheater 113 and sent to the medium / low pressure turbine 123.

The generator 124 is driven by the high-pressure turbine 122 and the medium / low-pressure turbine 123 to generate power. During·
The steam exiting the low-pressure turbine 123 is condensed by a condenser 125. This condensed water is sent again to the economizer 130 of the boiler 126 by the feedwater pump 117. Water supplied from the water supply pump 117 is supplied to the primary spray 11 via the primary spray control valve 115 and the secondary spray control valve 119.
6 and the secondary spray 120. In addition, the water supplied from the water supply pump 117 is sent to the reheat spray (SP3) 132 via the reheat spray control valve 131. The boiler 126 is provided with a gas recirculation fan 114 for recirculating the combustion gas. Also, in the boiler 126,
An induction fan 118 for controlling exhaust gas is provided.

The thermal power plant is provided with various sensors for detecting the state of the plant. That is, as shown in FIG. 2, a sensor S1 for measuring the main steam pressure (P _MS ) and a primary superheater outlet steam temperature (T _1SH ).
S2, which measures the amount of O ₂ in the exhaust gas (O ₂ ), sensor S4, which measures the furnace pressure (P _WW ), and measures the secondary superheater outlet steam temperature (T _2SH ). Sensor S5 for measuring the main steam temperature (T _MS ), sensors S7 and S for measuring the reheat steam temperature (T _RS ) and the primary reheater outlet steam temperature (T _1RH ).
9 and a sensor S8 for detecting the power generation (MW) of the generator 124. Although not shown, 1
Secondary superheater 109, Secondary superheater 110 and Tertiary superheater 1
11 is provided with a flow rate, pressure and temperature sensor for measuring the steam flow rate, pressure and temperature at the inlet or outlet, respectively. Similarly, the primary reheater 112 and the secondary reheater 113 are provided with flow rate, pressure, and temperature sensors for measuring steam flow rate, pressure, and temperature, respectively, at the inlet or the outlet. Each of the sensors S1-S
9 and the output signals of these flow rate, pressure, and temperature sensors are sent to a master control unit and a sub-loop control unit, which will be described later. The reheat spray 132 is for emergency use and uses the spray only when the temperature exceeds the limit value.

In this embodiment, the control of the thermal power plant is performed based on these sensor information. The control in this embodiment is functionally constituted by a master control unit and a sub-loop control unit based on the master control unit. FIG. 7 shows the outline.

As shown in FIG. 7, the master control unit 1000
A normal control system (1100 system) for controlling a fast response portion such as a main steam pressure and a predictive control system (1200 for performing a predictive control for a slow response portion such as a main steam temperature).
System). The normal control system includes a main steam pressure control unit 1
101, a gas O ₂ control unit 1102 and a furnace pressure control unit 1103, and a unit master 1104 that receives a load command input and outputs a demand operation command for turbine control and boiler control in response thereto.
The prediction control system includes a primary superheater outlet steam temperature control unit 120.
Each control processing function unit of the primary and secondary superheater outlet steam temperature control unit 1202, the main steam temperature control unit 1203, and the reheat steam temperature control unit 1204 is provided.

Further, the normal control system corrects the demand operation command of the unit master 1104 by the operation amount from the main steam pressure control unit 1101 and outputs an operation command for water supply control. A correction unit 1106 that corrects the output of the correction unit 1105 with an operation amount from the primary superheater outlet steam temperature control unit 1201 and outputs an operation command for fuel control, and an output of the correction unit 1106. About the gas O ₂ control unit 110
A correction unit 1107 that corrects with the operation amount from the control unit 2 and outputs an operation command for air control;
Output from the furnace pressure control unit 1103, and outputs an operation command for exhaust gas control. The output of the correction unit 1105 is a secondary superheater outlet steam temperature control unit. A correction unit 1109 that outputs an operation command for the primary spray control by correcting the operation amount from the operation unit 1202, and an output of the correction unit 1105 is corrected by an operation amount from the main steam temperature control unit 1203, About the correction | amendment part 1110 which outputs the operation command for secondary sprays, and the output of the said correction | amendment part 1105,
A correction unit 1111 that corrects the operation amount from the reheat steam temperature control unit 1204 and outputs an operation command for recirculation gas control is provided.

The unit master 1104 controls economic load distribution (ELD: Economic Load Disp.
Attach cntrol) and a command concerning automatic frequency control (AFC), the load change rate in the thermal power plant,
The load change width is limited and the frequency is corrected, and the corresponding operation command is calculated and output.

The primary superheater outlet steam temperature control unit 1201,
The respective processing function units of the secondary superheater outlet steam temperature control unit 1202, the main steam temperature control unit 1203, and the reheat steam temperature control unit 1204 operate the respective target values based on the information from the corresponding sensors. Calculate the quantity,
Output. In FIG. 7, a processing unit in a block indicated by a double line performs predictive control when determining a control amount.

The sub-loop control unit 2000 receives the output signal of the unit master 1104 and the signal indicating the amount of power generation from the sensor S8, and controls the main steam control valve 121, the turbine control unit 2001, and the correction unit 1105. The water supply control unit 2002 receives the output signal of the water supply pump 117, controls the fuel supply unit 2003 that receives the output signal of the correction unit 1106 and controls the coal feeder motor 105, and outputs the output signal of the correction unit 1107. Receiving push fan 10
3 and an air control unit 2004 for controlling the
8, an exhaust gas control unit 2005 that controls the induction fan 118 in response to the output signal, a primary spray control unit 2006 that controls the primary spray control valve 115 in response to the output signal of the correction unit 1109, and the correction unit 1110 A secondary spray control unit 2007 for controlling the secondary spray control valve 119 in response to the output signal of the recirculation gas control unit 20 for controlling the gas recirculation fan 114 in response to the output signal of the correction unit 1111
08.

These have, for example, a hardware system configuration as shown in FIG. That is, a first master controller 1100 that shares and processes the functions of the normal control system of the master control unit 1000, a second master controller 1200 that shares and processes the functions of the prediction control system, and a sub-loop control. The turbine control controller 2010, the water supply control controller 2020, and the fuel control controller 203 which constitute each processing unit of the unit 2000
0, air controller 2040, exhaust gas controller 2050, primary spray controller 206
0, a secondary spray controller 2070 and a recirculation gas controller 2080. These controllers are connected by a transmission network and exchange signals with each other. In the present embodiment, a model tuning system 1300 for tuning a prediction model used in predictive control is connected to the transmission network 1500. These controllers and tuning systems are each configured by a computer system. Although not shown, these computer systems have, for example, a configuration having a central processing unit, a memory, an interface, and the like.

As shown in FIG. 9, the model tuning system is connected to a second master controller 1200.
May be configured to perform the processing.

One of the control variables of the thermal power plant that is difficult to control is the steam temperature. For example, the primary superheater outlet steam temperature T _1SH , the secondary superheater outlet steam temperature T _2SH , the main steam temperature T _1SH
The four are _MS and reheat steam temperature T _RS . Also,
For controlling these steam temperatures, there are four manipulated _variables : fuel flow rate F _f , primary spray flow rate F _SP1 , secondary spray flow rate F _SP2, and recirculation gas flow rate F _grf . In the present invention, these are controlled by predictive control. At that time, a characteristic is that a new prediction model is used.

Next, an embodiment in which the predictive control of the present invention is applied to the above-described steam temperature control of a thermal power plant will be described.

FIG. 1 shows an outline of the configuration of an embodiment of the present invention. In the present embodiment, a target value prediction system 2 for obtaining a predicted value of a target steam temperature of the thermal power plant 1, a steam temperature prediction system 3 for obtaining a predicted value of steam temperature, and a predicted value of the target value and a predicted value of the steam temperature are provided. A manipulated variable determining system 4 for determining a manipulated variable based on a model, tuning, and the like for tuning model parameters for steam temperature prediction.
The system includes a system 5 and a switch 6 for selecting whether to use the predicted value of the prediction system or to directly use the control amount. The target value prediction system 2, the steam temperature prediction system 3, the manipulated variable determination system 4, and the switch 6 are realized in the second master controller 1200 that executes the processing of the prediction control system. In addition, the model tuning system 5 is realized by a computer 1300 configuring the model tuning system.

The target value predicting system 2 predicts a near future value of the target value r of the steam temperature by the equation (1). here,
The target values ＾ r ₁ , ＾ r ₂ , ＾ r _{3 of the} primary superheater outlet steam temperature T _1SH , the secondary superheater outlet steam temperature T _{2 SH} , the main steam temperature T _MS, and the reheat steam temperature T _RS , respectively. ^ to calculate the r _4.

[0037]

(Equation 1)

Here, ＾ r ₁ (k, n): At the present time k, the predicted value of the target value r ₁ of the primary superheater outlet steam temperature T _1SH at n sampling destinations ＾ r ₂ (k, n): At the current time point k, the predicted value of the target value r ₂ of the secondary superheater outlet steam temperature T _2SH at the n sampling destinations ＾ r ₃ (k, n): At the current time point k, the main steam temperature T
Predicted value of n sampling destinations of _MS target value r ₃ ＾ r ₄ (k, n): Predicted value r _i (k) of n sampling destinations of target value r ₄ of reheat steam temperature T _RS at current time k : the value at the current time k target value r _i of the steam temperature _{(i = 1~4) a i (} k): the current change rate at the time point k of the target value r _i of the steam temperature (i = 1~4) △ T : Sampling period In the present embodiment, the target value is determined by prediction in the near future, but the present invention is not limited to this.

The steam temperature prediction system 3 predicts a near future value of the steam temperature using a model of the steam temperature system.
As shown in FIG. 3, the model of the steam temperature system is configured by combining the models of the heat exchangers of the thermal power plant 1 with the dead time elements 301 and 303 and the lumped constant model 304 based on the physical equation. . Further, in this embodiment, the heat exchanger is a water / steam system, which is sequentially connected from the upstream side to the downstream side, and has a configuration in which a change in the state of water / steam is transmitted from the upstream side to the downstream side. In this embodiment, a model of each of these heat exchangers is used in combination to predict a near future value of the steam temperature.

The model of the lumped parameter system is as follows.
Represented by the energy conservation equation for steam and gas systems. For the gas temperature, a gas temperature calculation model 306 is used. This gas temperature calculation model 306
Are the fuel flow rate _Ff , the air flow rate _Fa, and the gas recirculation flow rate F
Calculate gas temperature based on _grf .

The lumped parameter model 3 based on the physical equation
For example, a Kalman filter 305 is configured for the 04. The value of the state variable is estimated by the Kalman filter 305, and the model of each heat exchanger is integrated and used based on the estimated value to predict the near future value of the steam temperature.

The lumped parameterization model (i) at the preceding stage shown in FIG.
In 304, the steam flow rate Fs is input via a _dead time element 302, the steam temperature θ _si-l is input via a _dead time element 301, and the gas temperature θ _gi-l is input via a dead time element 303. , The model (i) 304 is the steam temperature θ
Output _si . The lumped parameter model (i +
1) In 304, the steam flow rate Fs passes through the dead time element 302, the steam temperature θ _si output from the previous stage passes through the dead time element 301, and the gas temperature θ _gi passes through the dead time element 303. Each is entered. The model (i +
1) 304 outputs the steam temperature θ _{si + 1} . Note that FIG.
In the above, the heat exchanger is shown for two stages, but the present invention is not limited to this.

Dead time elements 301, 302 and 303
In this embodiment, is approximated by a third-order delay element formed by cascading first-order delay elements as shown in FIG. In the case of the present embodiment, it is assumed that T ₁ = 15 seconds for each time constant of each first-order lag element constituting the dead time element 301. Each time constant of each first-order lag element constituting the dead time element 302 is assumed to be T ₂ = 5 seconds. Further, each time constant of each primary delay element constituting the dead time element 303 is T
₃ = 30 seconds is assumed. Therefore, the dead time element 30
In the case of 1, the time constant is about 45 seconds, and in the case of the dead time element 303, the time constant is about 150 seconds. On the other hand, in the case of the dead time element 302, its time constant is
It is about 15 seconds, which is relatively small. Therefore, the dead time element 302 can be omitted.

The lumped parameter model in the heat exchanger is derived from the law of conservation of energy, assuming the model shown in FIG. In the model shown in FIG. 10, steam flows on one side of the metal constituting the tube wall of the heat exchanger, and gas flows on the other side.
It is assumed that heat is transferred from gas to steam via metal (the hatched portion in FIG. 10). Here, the gas at the inlet gas temperature θ _gini and the gas flow rate F _gi comes into contact with the metal of the heat exchanger, gives heat of the heat transfer amount Q _gmi to the metal, and flows out at the outlet temperature θ _gi . On the other hand, the steam having the inlet steam temperature θ _sini and the steam flow rate F _si comes into contact with the metal of the heat exchanger, receives heat of the heat transfer amount Q _msi from the metal, and flows out at the outlet temperature θ _si . The gas flow rate F _{g i,} ie,
The boiler gas flow rate F _gBF is a _sum of the air flow rate F _a , the fuel flow rate F _f, and the recirculation gas flow rate F _grf , as described later.

On the basis of such a physical model, the lumped parameter model formula is represented by formulas (2) and (3) derived from the law of conservation of energy as described later.

The symbols used in the following mathematical expressions are as follows.

V: volume γ: specific weight H: enthalpy F: flow rate Q: heat transfer amount M: weight C: specific heat θ: temperature P: pressure A: heat transfer area α: convective heat transfer coefficient β: radiant heat transfer coefficient The suffix is as follows.

S: water / steam g: gas m: metal gm: gas to metal ms: metal to water / steam i: i-th heat exchanger Water / steam system, ie, energy of fluid inside the pipe of the heat exchanger The conservation equation is represented by equation (2). The energy conservation equation for the tube metal system is given by equation (3).

[0049]

(Equation 2)

[0050]

(Equation 3)

Here, V _si : the volume (m ³ ) of the fluid (water / steam) inside the tube of the heat exchanger γ _si : the specific weight (kg / m ³ ) of the fluid (water / steam) inside the tube H _si : Outlet enthalpy of fluid (water / steam) inside pipe (kc
al / kg) H _sini : Inlet enthalpy (k) of pipe internal fluid (water / steam)
cal / kg) F _si : Flow rate of pipe internal fluid (water / steam) (kg / s) A _msi : Heat transfer area from pipe metal to pipe internal fluid (water / steam) (m ² ) A _gmi : Pipe external Heat transfer area from fluid (gas) to pipe metal (m ² ) α _msi : convective heat transfer coefficient from pipe metal to pipe internal fluid (water / steam) (kcal / m ² · s · ℃) α _gmi : pipe Convective heat transfer coefficient from external fluid (gas) to tube metal (kcal / m ² · s · ° C.) M _mi : weight of tube metal of heat exchanger (kg) C _mi : specific heat of tube metal (kcal / kg · ° C.) theta _mi: the tube metal temperature (° C.) theta _si: outlet temperature inside the tube fluid (water and steam) (℃) θ _gini: inlet temperature of Kangaibu fluid (gas) (℃) _{i: i-th} heat Exchanger The response of the fluid (gas) outside the pipe of the heat exchanger is much faster than the response of the fluid inside the pipe and the metal part. Therefore, here, it is assumed that the gas system statically holds the energy conservation equation. Therefore, the heat exchanger inlet gas temperature θ _gini is given by the following equation.

[0052]

(Equation 4)

[0053] Here, eta: fuel heating efficiency H _u: Fuel heating value (kcal / kg) F _f: fuel flow (kg / s) H _a: air enthalpy (kcal / kg) F _a: air flow rate (kg / s) H _grf : _{Enthalpy of} recirculated gas (kcal / kg) F _grf : Recirculated gas flow rate (kg / s) C _pg : Gas specific heat (kcal / kg · ° C) β _WW : Furnace radiant heat transfer coefficient F _gBF : Boiler Gas flow rate (kg / s) _QWW : Furnace water wall heat absorption (kcal / s) _QHEX : Total heat absorption of gas side upstream heat exchanger other than furnace water wall (kcal / s) enthalpy H _a is the specific heat of air C _pa,
When the temperature of the air and theta _a, is given by _{H a ≒ C pa · θ a} . Also, the enthalpy H of the recirculated gas
_grf is the specific heat of the gas C _pg, when the temperature in the vicinity of economizer gas to theta _ge, given by _{H grf ≒ C pg · θ ge} .

When the heat transfer of the water / steam system of the heat exchanger is approximated to the constant pressure process, the following equation is established. In the following, the number i of the heat exchanger is omitted.

[0055]

(Equation 5)

Here, C _ps : constant pressure specific heat (kcal / kg)
(° C.) H _so : Reference enthalpy (kcal / kg) In this case, equation (7) is approximated by the following equation.

C _ps = (ΔH _s / Δθ _s ) _p (8) By substituting the equations (5) and (6) into the equation (2) and rearranging,
The following equation is obtained.

[0058]

(Equation 6)

By transforming equation (3), the following equation is obtained.

[0060]

(Equation 7)

When the expressions (9) and (10) are put together, the following expression is obtained.

[0062]

(Equation 8)

[0063]

(Equation 9)

Here, A _ij : an element of the state transition matrix B _ij : an element of the driving matrix Note that x ₁ and u ₂ in the above equations (11) and (12) are measurable quantities, and x ₂ is , An amount that cannot be measured. The heat transfer coefficient α _ms from the pipe metal to water / steam and the heat transfer coefficient α _gm from the pipe external fluid (gas) to the pipe metal are approximated by the following equations.

Α _ms = f (F _s ) (23) α _gm = f (F _gBF ) (24) When the expressions (11) and (12) are expressed in discrete time, the following expression is obtained. . This is the physical equation indicating the prediction model.
In the present specification, the symbol “ _M ” indicates that the quantity is a quantity represented by a matrix.

[0066] _{X M (k) = A M} X M (k-1) + B M U M (k-1) ...... (10a) where, A _M: state transition matrix (model) B _M: drive matrix ( Model) X _M (k): State quantity at k sampling time (model) X _M (k-1): State quantity at k-1 time (model) _UM (k-1): Manipulation quantity at k-1 time (Model) Next, estimation of a state quantity by a Kalman filter and estimation of an error will be described. Here, assuming that the estimated value of the state quantity at the time of k sampling is ＸX _M (k) and the model error is ε (k), the most likely estimated value of the state quantity at the time of k sampling ＾ X
_{Find M} (k).

[0067] _{^ X M (k) = A} M X M (k-1) + B M U M (k-1) + ε (k) ... (10b) ε (k) = K {X M (k) -~ _{X M (k)} ...... (} 10c) ~X M (k) = A M ^ X M (k-1) + B M U M (k-1) ...... (10d) where, K: Kalman gain FIG. 4 shows a block diagram of a model of each heat exchanger alone using the model formula. That is, the heat exchanger alone model, the fuel flow rate F _f, air flow F _a and the gas recirculation flow F _grf and, furnace waterwall heat absorption Q
A gas temperature calculation model 306 for calculating the gas temperature θ _gin based on _WW and the total heat absorption Q _HEX of the gas side upstream heat exchanger other than the furnace water wall, A conversion unit 307 for converting the water (steam) into an entrance enthal H _sin , and dead time elements 301, 302, and 303 for approximately giving a third-order delay to the enthal H _sin , the steam flow rate F _s, and the gas temperature θ _gin , respectively. , A lumped constant model 304 and a Kalman filter 305.

The overall structure of the prediction model is as shown in FIG. In FIG. 5, the water sent from the feed water pump 117 passes through the economizer (ECO) 130 and the furnace water wall (WW)
Heated at 108, primary superheater 109, primary spray 1
16, the secondary superheater 110, the secondary spray 120, the tertiary superheater 111 and the main steam control valve 121 are sent to the high-pressure turbine 122. Here, each superheater 109, 110
In (111) and (111), the gas temperatures θ _g1SH , θ _g2SH, and θ _g3SH obtained by the gas temperature calculation are given, and the respective outlet steam temperatures are calculated by the above-mentioned physical formula. In addition, the steam exiting the high-pressure turbine 122 is supplied to the primary reheater 11
2. It is sent to the low pressure turbine 123 via the reheat spray 132 and the secondary reheater 113. Here, the primary reheater 11
In the secondary and secondary reheaters 113, the gas temperatures θ _g1RH and θ _{g2 RH} determined by the gas temperature calculation are given, and the respective outlet steam temperatures are calculated by the above-mentioned physical equations.

Here, the prediction is performed by calculating the control amount ＾ X _M (k, n) at the n-th sampling destination at the time of the k-th sampling by using the equation (10b).

＾ X _M (k, 1) = A _M ＾ X _M (k) + B _{MU M} (k) + ε (k) ＾ X _M (k, 2) = A _M ＾ X _M (k, 1) _{+ B M U M (k)} + ε (k) · · · ^ X M (k, n) = a M ^ X M (k, n-1) + B M U M (k) + ε (k) in this way The prediction characteristics to be performed will be examined. Here, a comparison is made between a case where a lumped parameter model is constructed and prediction is performed only for the last-stage superheater and a case where prediction is performed using a combination of the dead time element and the lumped parameter model of the present embodiment. . FIG. 11A shows a response characteristic measured for the same operation amount. As is clear from the figure, when the prediction is performed using only the lumped parameter model, the error ε increases because the response is faster than the response of the plant. Therefore, the retraction effect due to the error ε also increases, and the prediction accuracy deteriorates (FIG. 9B). On the other hand, in the combination of the dead time element and the lumped parameter model, since the response of the control amount is close to the response characteristic of the plant, the error ε is small, the pullback effect is small, and the prediction accuracy is improved. (Figure (c)).

In the present embodiment, the prediction control and the normal control are juxtaposed, and the prediction control is performed only when the error of the prediction value falls within a predetermined range. The switch 6 of FIG. 1 is provided for this purpose. When the error is large, normal PI (proportional / integral) control is performed, and when the error is small, the control is switched to the prediction control of the present invention.

Note that this embodiment is executed by being calculated by a computer. Therefore, the model formula is (10
a) It is represented in discrete time as shown in the equation. Therefore,
The case where the model formula is expressed in discrete time will be described in some detail.

The following equations are obtained by expressing the heat transfer equation models (11) and (12) derived from the energy conservation equation in discrete time and expressing them in matrix elements.

[0074]

(Equation 10)

Next, the case where the state variable is estimated by the Kalman filter 305 using the equation (25) will be described. In order to predict the steam temperature using the expression (25), which is a discrete-time expression of the heat transfer model, the values at the start of the prediction of the steam temperature θ _s and the pipe metal temperature θ _m which are the state variables are necessary. . However, the pipe metal temperature θ _m is
Since measurement cannot be performed, an estimated value is used. In this embodiment, the estimation of the tube metal temperature theta _m, applying the Kalman filter.

To apply the Kalman filter, (2
Expression 5) is expressed by the following expression.

X _M (k) = Φ _M (k−1) · X _M (k−1) −H _M (k−1) · _UM (k−1) (26)

[0078]

[Equation 11]

Here, X _M (k): a state variable vector at the time of k samplings (steam temperature θ _s , pipe metal temperature θ _m ) Φ _M (k-1): a state transition matrix H _M at the time of k-1 sampling (K-1): drive matrix at the time of k-1 sampling _UM (k-1): manipulated variable at the time of k-1 sampling Further, the observation process of the system represented by the equation (26) is as follows.
It is assumed to be represented by the following equation.

Y _M (k) = C _M × X _M (k) + V _M (k) (31) where, Y _M (i): an observation vector at the time of k sampling (corresponding to the vapor temperature θ _s ) X _M (k): State variable vector (steam temperature θ _s , pipe metal temperature θ _m ) V _M (k): Observation noise vector C _M : Observation matrix The Kalman filter of the equations (26) and (31) is , And is constituted by the following equation. Here, 'means a transposed matrix.

{X _M (k) = 〜X _M (k) + P _M (k) · _CM ′ · W _M ⁻¹ {Y _M (k) −C _M · 〜X _M (k)} _{32) ~X M (k) =} Φ M (k-1) · ^ X M (k-1) + H M (k-1) · U M (k-1) ...... (33) P M (k) ^{_{= {M M ~ 1 (k}} ) + C M '· W M ~ 1 · C M} ~ 1 ...... (34) M M (k) = Φ M (k-1) · P M (k-1) · _{Φ M '(k-1)} + H M (k-1) · U M (k-1) · H M' (k-1) ... (35) here, ^ X _M (k): the k sampling point in time Control amount X _M
Maximum likelihood estimated value of (k) ~ X _M (k): Estimated value of control amount X _M (k) at the time of k samplings Next, the manipulated variable determination system 4 will be described. The manipulated variable determination system 4 determines the manipulated variable based on the near future predicted value of the target value and the near future predicted value of the steam temperature. In this embodiment, PI control is performed. The algorithm is shown in Equation 14.

[0082]

(Equation 12)

[0083] Here, k _pi: proportional gain k _Ii: integral gain △ u _i (k): The change in model tuning system 5 of the operation amount u _i at the current time point k, the characteristics of the model of the steam temperature system Model to match the characteristics of the thermal power plant
Tune parameters. FIG. 6 shows the configuration of the model tuning system 5. Next, the tuning algorithm will be described with reference to FIG.

First, the model parameters are tuned by the hill-climbing method so that the error between the control amount of the thermal power plant and the estimated value of the control amount estimated by the prediction model, that is, the model error is reduced, and the tuning results are collected and stored. I do.

Next, a model tuning rule is learned by a neural network (neuro) by using the model tuning result by the hill climbing method as teacher data. After learning the model tuning rules by this learning, the parameters of the prediction model are tuned by the neural network.

Similarly, a model tuning rule is extracted as a fuzzy rule based on the result of the model tuning by the hill climbing method, and the parameters of the prediction model are tuned by the extracted fuzzy rule.

Note that a neural network that has learned model tuning rules can also be used to create the fuzzy rules. That is, a fuzzy rule can be created from the input / output characteristics of the learned neural network.

In the model tuning, time response data of a plant (or a simulator) may be fetched, and the static characteristics may be tuned based on the time response data.
Tune the dynamic characteristics. This tuning
For example, the processing can be performed using the master controller, but may be performed on another computer system.

The tuning of static characteristics is performed for each load level by changing the load level by a fixed width. As shown in FIG. 15, first, the current load level is fetched (step 1501). Capture the process amount (step 150
2). Here, for each heat exchanger, F _s , θ _s , P _s ,
F _g, P _g, captures the θ _g and the like. Based on these quantities,
An enthalpy (H _s ) distribution of water / steam is determined (step 1503). That is, the enthalpy (H _s ) of water / steam at the inlet and outlet of each heat exchanger is determined. Further, the distribution of the absorbed heat quantity (Q _ms , Q _gm ) of water / steam is estimated (step 1504). Q _ms and Q _gm are equal when the system is balanced.

Next, the enthalpy (H _g ) distribution of the gas is estimated. This is determined by using the temperature at the portion where the temperature can be measured and the specific heat of the gas (step 1505). The portion where the temperature can be measured is, for example, near the economizer. Further, the distribution of the gas temperature (θ _g ) is estimated (step 1506). This is estimated using the enthalpy distribution and the specific heat. Further, the temperature (θ _m ) distribution of the metal is estimated (step 1507). Finally, the heat transfer coefficients (α _ms , α _gm , β _gm ) are estimated (step 150).
8).

Next, tuning of dynamic characteristics will be described. As shown in FIG. 16, the plant (simulator)
, A step input, a ramp input, and the like are given, and the time response data of the plant corresponding thereto is taken in (step 1601). As an input, for example, a command to increase the fuel flow rate is given, and data on the change in the steam temperature at the outlet of the primary superheater is taken in. For the model, for example, initial parameters determined by design are set (step 1602). Then, the operation amount and the control amount of the plant are given to the model, and the initial state of the model is estimated (step 1603). Thereafter, a step input, a ramp input, and the like are given to the model, and a time response to the input is calculated (step 1604). The error of the time response between the plant and the model is calculated (step 1605).
Then, it is determined whether or not this error has converged to a predetermined range (step 1606). If not converged, the model parameters are modified (step 1607). This modification can be performed using, for example, a hill-climbing method, fuzzy inference, a neural network, or the like. Then, the procedure from step 1603 is repeated, and when the error converges to a certain range, the tuning is ended.

Next, the correction of the model parameters by the hill-climbing method will be described. 17 and 18 show an example thereof. FIG. 17 is an example of a configuration of a system for tuning a model. In this example, the operation amount _UM is input to the thermal power plant (may be a simulator) 1 and the prediction system 3 using the prediction model of the present invention. Correspondingly, the state quantity X _M from the thermal power plant 1 and the prediction system (model: indicated by m suffix)
3 outputs a predicted state quantity X _mM . The model tuning system 5 takes in these errors e _M , integrates them according to the following equation, and evaluates them according to the magnitude of the integrated value. Then, the parameters are corrected by the hill-climbing method so that the integrated value is minimized.

[0093]

(Equation 13)

FIG. 18 shows the principle of the complex method in which the simplex method is extended so that it can be applied when there are constraints. This complex method uses a set of (n + 1) points in an n-dimensional space to form a simplex (si
plex), of which 1 is the largest of the function values.
For each point, take a mirror image of the hyperplane spanned by the remaining points, form a new simplex, perform the same operation in this new simplex,
This is a method of repeatedly forming a new simplex to reach the minimum point of the function value. In FIG. 18, the magnitude of the evaluation value is represented by a contour line. In this example, since the minimum value is obtained, the contour lines represent valleys, not peaks. Also, in the same figure, the hatched portions represent the constraint conditions. If a point falls under a constraint, it is decided to remove it.

The example shown in FIG. 18 has two parameters P1
This is an example of executing correction by the hill-climbing method for P2 and P2. That is, first, initial values of the parameters P1 and P2 are given. For this purpose, for example, a design value can be used. Alternatively, a random number may be generated and given. This point is designated as coordinate point 1 in FIG. Next, coordinate points 2 and 3 are generated by random numbers. Then, these three parameters are set in the simulator of the thermal power plant and the model 3, respectively, and the errors of the respective state quantities are obtained, and the integration of the equation (36) is performed. Of these points, for the point having the largest function value (here, 1), the points 2 and 3 are on a plane, and the mirror image point of the point 1 is obtained.
Here, this is set to 4. And points 2, 3 and 4
Constitutes a new simplex. Then, the integration for point 4 is performed, the integrated value is compared with points 2 and 3, and the same operation as described above is repeated. If point 4 does not satisfy the constraint condition, a point shifted to the near side is newly set as point 4. In this manner, in this embodiment, a combination of parameters having a high evaluation value and a small error is sequentially searched.

The details of the hill climbing method are described in "M.
J. Box, A New Method of Con
strained Optimization wit
hOther Methods, The computer
er J. et al. Vol8, No1 (1985), pp. 146-64. 42
-52 "and" Kiyoshi Shimizu, System Control and Mathematical Programming (Published February 10, 1971), pp. 76-79 ".

In this way, an optimal parameter can be found. In the present embodiment, the parameter requiring tuning includes a time constant of a dead time element. In this example, the parameters of the model are adjusted by the hill-climbing method using the operation data of the plant, so that the parameter adjustment can be automated and the adjustment time can be reduced.

A method for constructing a fuzzy rule and making a correction in order to make a more efficient correction using the correction data by the hill-climbing method will be described. FIG.
FIG. 9 shows an example.

FIG. 19 shows an example of the configuration of a fuzzy inference system. This system can be mounted on the model tuning system 5. This system includes a fuzzy rule unit 1901 that describes a fuzzy rule,
Membership function unit 19 for storing membership functions for determining the degree of conformity of the antecedent and consequent parts of each rule
02, and a fuzzy inference engine unit 1903 that performs inference by fuzzy rules for input variables based on the fitness determined by the membership function.

In this embodiment, as shown in FIG. 20 (A), a model for a rise time T (shown by a solid line) at which the steam temperature of the primary superheater of the plant reaches 62% of a target value with respect to an increase in fuel flow rate. Rise time Tm (shown by broken line)
Focusing on the ratio (Tm / T), a rule based on the ratio is described, and the ratio (Tm / T) obtained by the test run and the simulation is described in advance in this rule as shown in FIG. The rules are applied, fuzzy inference is performed, and parameters are tuned according to the determined membership function fitness.

In the fuzzy rule section 1901, for example,
The following rules are stored. Here, an example is shown in which the time constant T ₃ of the dead time element 303 is corrected using the correction coefficient C ₃ . Rule 1: If (Tm / T) is small, the correction factor C ₃ of T ₃
To increase. Rule 2: If (Tm / T) is appropriate, correction coefficient C ₃ of T ₃ is appropriate. Rule 3: If (Tm / T) is large, the correction factor C ₃ of T ₃
Smaller.

In the fuzzy inference engine unit 1903, as shown in FIG. 21, the input (Tm / T) is
The fitness is determined by applying to the (Tm / T) membership function shown in FIG. Then, the fitness obtained, by applying membership functions of the correction factor C _3, the parts area included in the fit, determine the centroid, determining a correction factor corresponding to the position of its center of gravity. In the example shown in the figure, since rules 1 and 2 have a degree of conformity, the center of gravity is calculated for the entire area of the region included in the degree of conformity of each consequent part.

Here, only the primary superheater is shown, but this can be similarly applied to other superheaters. In that case, the membership function and the rule may be created according to each.

The fuzzy inference model is created, for example, as shown in FIG. That is, first,
Collect data about the target (step 220)
1). Then, a qualitative relationship between input and output variables is extracted (step 2202). Here, the choice of which variable to choose is also made. Further, a fuzzy rule and a membership function are created (step 2203). As described above, the model structure is determined. Then, the model parameters are identified by adjusting the fuzzy rules and the membership functions (step 2204).

As described above, according to the present embodiment, since the coefficient for correcting the parameter is determined, the parameter can be automatically corrected in accordance with the determined coefficient. Further, in this example, the parameters of the model are adjusted by fuzzy inference using the operation data of the plant, so that the parameter adjustment can be automated and the adjustment time can be reduced. Compared with the hill-climbing method, the fuzzy inference can further reduce the adjustment time, similarly to the neural network described later.

The modification of the model parameters can be performed using a neural network in addition to the fuzzy method. In this case, the rule is created by learning past data.

Next, the correction of parameters by the neural network will be described. In a neural network, parameter correction can be performed by learning in advance the past data on parameter correction. FIG. 23 shows the configuration. In the example shown in FIG. 23, the rise time ratios (T _m1SH / T _1SH , T
_m2SH / _T2SH , _Tm3SH / _T3SH ), and the corresponding correction coefficients _C31SH , _C32SH and _C33SH are obtained. FIG. 24 shows the relationship between the rise time ratios.

A neural network model can be created, for example, by the procedure shown in FIG. That is, data is collected (step 2501), and input / output variables are determined, including selection of variables (step 2501).
2). Then, the number of layers and the number of units of the neural network are determined to determine the structure (step 250).
2). Thus, the model structure is determined. Next, learning is performed by back propagation to adjust the weight coefficient of each neuron of the neural network (step 2504).

In this example, the parameters of the model are adjusted by the neural network using the operation data of the plant, so that the parameter adjustment can be automated and the adjustment time can be shortened. It should be noted that compared to the hill climbing method, the neural network can further reduce the adjustment time.

In the embodiment described above, a process in which a plurality of heat exchangers are formed as sub-processes is a combination of a dead time element and a lumped constant model by a physical equation as a model of each sub-process. Is used. It goes without saying that the model of the present invention can be applied to a process having only one process.

Further, when the present invention is applied to a process including a plurality of sub-processes, a configuration in which a dead time element is omitted can be adopted. Further, at this time, each sub-process can be constituted by a plurality of lumped constant models.

A simulation result is shown to show the effect of the present invention. FIG. 12 shows an example of the configuration, and FIG.
FIG. 3 shows a simulation result.

FIG. 12 shows the configuration of this embodiment applied to a variable-pressure once-through boiler. The present embodiment has a prediction model 601 in which a primary superheater, a secondary superheater, and a tertiary superheater are each a sub-process, and a lumped constant model and a dead time element corresponding to them are connected and incorporated in a cascade. Things. The model 601 predicts near-term future values of the primary superheater outlet steam temperature, the secondary superheater outlet steam temperature, and the main steam temperature. Operate fuel flow rate and secondary spray quantity.

FIG. 13 shows the evaluation result of the prediction model at the time of feedback control (PID control) using the current value in the embodiment shown in FIG. The evaluation shown in FIG. 13 was performed using a full-scale simulator of a variable-pressure once-through boiler. In FIG. 13, the steam temperature at the primary superheater outlet steam temperature, the steam temperature at 5 minutes ahead of the secondary superheater outlet steam temperature and the main steam temperature are indicated by broken lines, and the actual temperature is indicated by solid lines. As can be seen from this figure, the predicted value of the main steam temperature changed approximately 5 minutes before the actual temperature, and a good predicted value was obtained.

Next, FIG. 14 shows an evaluation result of the prediction control using the predicted value in the embodiment shown in FIG. The evaluation method is the same as in the case of FIG.

As can be seen from FIG. 14, in this embodiment,
The fluctuation of the main steam temperature is 1 at the time of the current value feedback control.
From / 3 to お り, it can be seen that controllability is greatly improved.

Although the above-described embodiment is an example of a thermal power plant, it goes without saying that the present invention is not limited to this. It is also applicable to nuclear power plants, chemical plants, and the like. Further, in the above-described embodiment, the prediction and the operation amount determination are performed by one controller, but may be performed by different controllers.

[0118]

According to the present invention, two or more processes are performed.
It is composed of sub-processes, and each sub-process is
Dell to input the operation result on the upstream side to the downstream side
, And perform a series of calculations and determine a series of the calculations in advance.
Since the state quantity is predicted by repeating the number of times, even if the process is a distributed constant system, its characteristics can be accurately simulated, the prediction accuracy can be improved, and the sub-process can be improved.
Kalman filter for lumped parameterized model
, The calculation load can be reduced.

[0119]

[0120]

[Brief description of the drawings]

FIG. 1 is a block diagram showing an outline of a configuration of an embodiment of the present invention.

FIG. 2 is an explanatory diagram showing an outline of a thermal power plant.

FIG. 3 is a block diagram showing details of a steam temperature prediction system according to the embodiment of the present invention.

FIG. 4 is a block diagram showing details of a prediction unit according to the embodiment of the present invention.

FIG. 5 is an explanatory diagram showing a relationship between water / steam and a gas temperature according to the embodiment of the present invention.

FIG. 6 is an explanatory diagram showing details of a tuning system constituting the embodiment of the present invention.

FIG. 7 is a block diagram showing a functional configuration of a thermal power plant control system according to an embodiment of the present invention.

FIG. 8 is a block diagram showing an example of a hardware configuration of the system shown in FIG. 7;

FIG. 9 is a block diagram showing another example of the hardware configuration of the system shown in FIG. 7;

FIG. 10 is an explanatory diagram showing the principle of a lumped parameter model of a heat exchanger.

FIG. 11 is a graph showing a comparison of response characteristics according to the present embodiment.

FIG. 12 is a block diagram showing a configuration of a steam temperature control section according to the embodiment of the present invention.

FIG. 13 is a graph showing a response when a prediction model is used to perform a control based on a current value in the system of FIG.

FIG. 14 is a graph showing simulation evaluation when control is performed using a predicted value by a prediction model in the system of FIG. 12;

FIG. 15 is a flowchart showing a process flow of tuning the static characteristics of the prediction model in the process control of the present invention.

FIG. 16 is a flowchart showing a processing flow of tuning of dynamic characteristics of a prediction model in the process control of the present invention.

FIG. 17 is a block diagram illustrating an outline of a system configuration when tuning a parameter of a prediction model in the process control of the present invention by a hill-climbing method.

FIG. 18 is an explanatory diagram showing the principle of parameter tuning by the hill-climbing method.

FIG. 19 is a block diagram showing an outline of a system configuration when tuning a parameter of a prediction model in the process control of the present invention by a fuzzy inference method.

FIG. 20 is an explanatory diagram showing the principle of tuning parameters by fuzzy inference.

FIG. 21 is an explanatory diagram showing a procedure for performing parameter tuning by fuzzy inference.

FIG. 22 is a flowchart showing a procedure for constructing a fuzzy inference model for performing parameter tuning.

FIG. 23 is a block diagram showing an outline of a system configuration when tuning a parameter of a prediction model in the process control of the present invention by a neural network method.

FIG. 24 is an explanatory diagram showing the principle of tuning parameters by the neural network method.

FIG. 25 is a flowchart showing a procedure for constructing a neural network model for performing parameter tuning.

Continued on the front page (72) Inventor Eiji Toyama 5-2-1 Omika-cho, Hitachi City, Ibaraki Prefecture Inside the Hitachi, Ltd. Omika Plant (72) Inventor Toru Kimura 5-2-1 Omika-cho, Hitachi City, Ibaraki Stock (56) References JP-A-4-205604 (JP, A) JP-A-2-301805 (JP, A) JP-A-63-128401 (JP, A) Keiji Watanabe and one other person , "Control of systems including dead time between input and output," System and Control, 1984, Vol. 28, No. 5, p. 269-277 (58) Field surveyed (Int. Cl. ⁶ , DB name) G05B 13/00-13/04 F22B 35/18 F01K 13/02 JICST file (JOIS)

Claims (4)

(57) [Claims] 1. An adaptive control method for a process in which a model of a process is incorporated, a predicted value of a control amount is obtained using the model, and an operation amount is determined based on the predicted value. along the flow, composed of two or more sub-processes, and, each sub-process, as well as constituted by lumped model <br/> based on individual physical expression for each, for each sub-process, the centralized constant
A Kalman filter for the generalized model
Estimate state quantities and their errors with Kalman filter
Then, the predicted value of the state quantity of the sub-process obtained by the model of the upstream sub-process is used as at least a part of the input variables of the model of the downstream sub-process, and the state quantity of the upstream sub-process is calculated by the model. The predicted value of the state quantity is input to the model of the downstream sub-process, and the predicted value of the state quantity of the sub-process is determined. It runs until the predicted value is obtained, and, subprocess of the upstream
From the calculation of the predicted value of the
A series of operations up to the operation to obtain the predicted value of the state quantity of the process
The number of repetitions predetermined, each sub-process
At least the lowest of the predicted values of the state quantities obtained for
An adaptive control method for a process, comprising determining an operation amount of the process based on a predicted value of a state amount obtained for a flow sub-process. 2. The adaptive control method for a process according to claim 1, wherein a model of each of the sub-processes is configured by a combination of a dead time element and a lumped constant model based on a physical equation, and an upstream sub-process. Inputting the predicted value of the state quantity obtained by the physical equation of the model to the physical equation of the model of the downstream sub-process via the dead time element of the downstream sub-process. Control method. 3. A process control system that determines a manipulated variable of a process and controls the process, including two or more sub-process models that configure the process along the flow, and using these models. A state quantity prediction system for predicting a state quantity of a process, a target value for the process, and an operation quantity determination system for determining an operation quantity for the process based on the predicted state quantity of the process, the state quantity prediction system, the provided sub-process corresponding to obtain the predicted value of the state quantity in the physical formula previously given as a model of each sub-process, and two or more predicted value calculating means, each said sub-process To the lumped parameter model
Kalman that estimates the state and its error
A filter and, for the two or more predicted value calculating means, a predicted value of the state quantity of the sub-process obtained by the upstream predicted value calculating means; In order to execute the calculation, the downstream predicted value calculation means executes the calculation sequentially , and
For calculating the predicted value by the predicted value calculation means corresponding to the process
Predicted value calculation means corresponding to the sub-process on the downstream side from
A series of calculations up to the calculation of the predicted value of
Control means for repeating a predetermined number of times to obtain a predicted value of the state quantity of the process. 4. The process control system according to claim 3 , wherein the state quantity prediction system includes a dead time element and the object as a model of each subprocess .
A built-in model constituted by a combination of a physical equation, the predicted value of the upstream side of the predicted value calculating means by the obtained state quantities, via a dead time element of the downstream sub-process equivalent
Process adaptive control system for causing the input to the predictive value calculating unit of the downstream side.