JPH06332501A - Feedback controller and incinerator using the feedback controller - Google Patents

Abstract

PURPOSE:To attain the otimum control even for a controlled ssystem that has the disturbance which can not be expressed in a numerical expression. CONSTITUTION:The refuse is blown up by the primary air A1 and then star to burn by a prescribed ignition device in an incinerator 1. The oxygen density S1 and the NOx density S2 measured in the exhaust air are supplied to a PID controller 3, and the disturbance parameters P such as the internal pressure of the incinerator 1, the amount of refuse thrown in the incinerator 1, etc., are supplied to a neural network controller 4. The controller 3 calculates the deviation between the density S1 and S2, and a prescribed set value respectively and then calculates the manipulated variable C1 of the secondary air A2 in order to minimize the preceding deviation. Meanwhile, the controller 4 calculates the manipulated variable C2 of the air A2 based on those density S1, S2 and parameters P of the incinerator 1 and through a network which is previously obtained by learning. Then, the amount of the air A2 to be supplied to the incinerator 1 is controlled based on the manipulated variable C3, i.e., the deviation between both variables C1 and C2.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an incinerator for incinerating refuse and the like, and a feedback control device suitable for use as a control device for controlling combustion by adjusting the amount of air, and an incinerator using the control device. Regarding

[0002]

2. Description of the Related Art Conventionally, there has been known feedback control for controlling an operation amount for a controlled object so as to bring the value closer to a target value according to the controlled variable output by the controlled object. The above feedback control device is used, for example, in a fluidized bed type waste incinerator that combusts waste in a fluidized state by blowing up the waste put in the incinerator with primary air. In such a refuse incinerator, the amount of unburned gas (NOx, dioxin) in the exhaust is minimized,
Secondary air is supplied to adjust the amount of oxygen in the incinerator in order to completely burn the waste, and the amount of oxygen and carbon dioxide contained in the exhaust gas are detected, and these amounts are optimized. The amount of secondary air is controlled by PID so that Generally, if the amount of oxygen in the exhaust gas is controlled to be about 6%,
It is empirically known that the amount of unburned gas remaining is minimized.

[0003]

By the way, in the above-mentioned incinerator, the combustion state in the incinerator changes due to changes in the amount of refuse input, the type of waste, or the internal pressure of the incinerator. Since such a change (disturbance) is difficult to formulate, sufficient combustion control cannot be performed simply by controlling the amount of secondary air according to the amount of oxygen in the exhaust gas as in conventional PID control. There is a problem that unburned gas such as NOx and dioxin is generated. As described above, the normal feedback control device has a problem that it is difficult to control the disturbance.

The present invention has been made in view of the above-mentioned circumstances, and an object of the present invention is to provide a feedback control device which can perform optimum control even for a disturbance that cannot be mathematically expressed, and an incinerator using the feedback control device. .

[0005]

In order to solve the above-mentioned problems, in the invention according to claim 1, the first control for calculating the first manipulated variable in accordance with the parameter of the controlled object which can be mathematically expressed. Means and a neural network, and a second control unit according to the non-matrixable parameters of the controlled object.
And a second control means for calculating the operation amount of No. 3, and an addition means for adding the first operation amount and the second operation amount to calculate a third operation amount. The control target is controlled based on an operation amount.

[0006] In the invention according to claim 2, the incinerator for incinerating the material to be burned and the amount of air supplied to the incinerator according to the components of the exhaust gas discharged from the incinerator are controlled. A first control means for calculating the operation amount of No. 1 and a neural network, and a second operation unit for operating the amount of air supplied to the incinerator according to the disturbance in the incinerator in addition to the components of the exhaust gas. And a second control means for calculating the operation amount of, and controlling the air amount according to the first and second operation amounts.

[0007]

According to the first aspect of the present invention, the first control amount is calculated by the first control means in accordance with the parameter of the controlled object that can be mathematically expressed. On the other hand, the second manipulated variable is calculated by the neural network of the second control means in accordance with the parameter of the controlled object that cannot be mathematically expressed. Further, the third control amount is calculated by adding the first operation amount and the second operation amount by the addition means. Then, finally, the controlled object is controlled according to the third operation amount. In this way, the controlled object is controlled by the second control means in addition to the first control means, which is a normal control system, according to the parameters that are disturbances and cannot be mathematically expressed. According to the invention of claim 2, the first control means controls the first operation amount for operating the amount of air supplied to the incinerator according to the components of the exhaust gas discharged from the incinerator. It is calculated. On the other hand, in addition to the components of the exhaust gas, a second operation amount for operating the amount of air supplied to the incinerator is calculated by the neural network of the second control means according to the disturbance in the incinerator. And finally, the amount of air supplied to the incinerator is the same as the first
And is controlled according to the second operation amount. in this way,
The air amount in the incinerator is controlled not only by the first control means, which is a normal control system, but also by the second control means according to the disturbance.

[0008]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described with reference to the drawings. FIG. 1 is a block diagram showing the configuration of a combustion control system of a refuse incinerator using the feedback control device according to the present invention. Reference numeral 1 denotes a refuse incinerator, into which refuse is introduced every predetermined period. Primary air A1 for blowing up the dust is supplied from below, and secondary air A2 for adjusting the oxygen concentration in the refuse incinerator 1 is also supplied. The gas components (for example, oxygen concentration S1 and NOx concentration S2) in the exhaust gas of the refuse incinerator 1 are detected by a predetermined sensor 2 and the PID controller 3
And the neural network controller 4.
The PID controller 3 is an ordinary control device, and supplies the manipulated variable C1 to the adder 5 in accordance with the oxygen concentration S1 and the NOx concentration S2 in the exhaust gas so that the oxygen concentration and the NOx concentration become optimum values. .

In addition to the oxygen concentration S1 and the NOx concentration S2, the neural network controller 4 may also be equipped with, for example, the internal pressure of the refuse incinerator 1 and the amount of dust (time-varying parameter that is difficult to quantify). Other parameters P such as
Is being supplied. The neural network controller 4
Has a learning function, and has the oxygen concentration S1 and NOx.
According to the density S2 and the parameter P, the manipulated variable C2 is supplied to the adder 5 so as to take an optimum value. The details of the neural network controller 4 will be described later. In the adder 5, the manipulated variable C1 and the manipulated variable C2 are added to obtain the difference between the two, and this is used as the manipulated variable C3 of the secondary air amount to the final refuse incinerator, and the secondary air not shown is shown. It is supplied to the supply device. The secondary air supply device adjusts the supply amount of the secondary air A2 to the refuse incinerator according to the operation amount C3.

Here, the above-mentioned neural net controller 4 will be described. Generally, in a neural network, each constituent element has a function similar to that of a human nerve cell. FIG. 2 is a conceptual diagram showing the basic configuration of the neural network. In the figure, the neural network comprises an input layer L1, an intermediate layer L2 and an output layer L3. Each layer L1, L2, L3 is composed of a plurality of units, and in this example, each of the three units U11, U1.
2, U13, U21, U22, U23, U31, U3
2, U33. Each unit U11, U
The input / output characteristics of 12, ..., U32, U33 are characterized by being non-linear. Also, each layer is sequentially coupled to the input layer L1, the intermediate layer L2, and the output layer L3, and there is no coupling in the same layer. Also, to determine the strength of each bond,
A variable weighting coefficient (not shown) is supplied to the input ends of the units U11, U12, ..., U32, U33.

In the learning in the above-mentioned neural network, first, each unit U11, U12 of the input layer L1.
And input data IND to U13. In this embodiment, the oxygen concentrations S1 and N contained in the exhaust gas of the refuse incinerator 1 are
The Ox concentration S2 and other parameters P (internal pressure, internal temperature, CO; hereinafter referred to as disturbance parameter P) are input data. These signals are sent to each unit U11, U12.
And U13, and the converted data is supplied to the intermediate layer L2 and further supplied to the output layer L3 and output.
Learning is performed by comparing the output value OUTD with a desired output value (target value) and changing the coupling strength (weighting coefficient) of each unit so as to reduce the difference. With such a neural network controller, for example, a phenomenon that is difficult to model with a mathematical expression such as the amount of dust to be thrown into the refuse incinerator can be acquired and controlled by learning.

According to the above-mentioned structure, a proper amount of dust is first thrown into the incinerator 1. The dust is blown up by the primary air A1 and the combustion is started by a predetermined ignition device. The sensor 2 detects the oxygen concentration S1, NOx concentration S2 in the exhaust gas and the disturbance parameter P at all times or at predetermined sampling intervals, and supplies them to the PID controller 3 and the neural network controller 4. Well-known control is performed in the PID controller 3,
Deviations between the supplied oxygen concentration S1 and NOx concentration S2 and a predetermined set value are obtained, and the manipulated variable C1 of the secondary air is supplied to the adder 5 so as to minimize the deviations. In the neural net controller 4, the refuse incinerator 1 supplied by the network obtained by learning in advance.
An output value that matches the target value is obtained according to the oxygen concentration S1, the NOx concentration S2, and the disturbance parameter P, and is supplied to the adder 5 as the manipulated variable C2 of the secondary air A2. The adder 5 takes the deviation between the manipulated variable C1 and the manipulated variable C2, and supplies this as a final manipulated variable C3 of the secondary air A2 to the refuse incinerator 1 to a secondary air supply device (not shown). . The secondary air supply device adjusts the supply amount of the secondary air A2 to the refuse incinerator 1 according to the operation amount C3. Even if dust is thrown in, the amount of secondary air is controlled more optimally by the PID controller 3 and the neural net controller 4.

Next, a second embodiment of the present invention will be described with reference to FIG.
Will be described with reference to. In the second embodiment, a neuro prediction model 20 is added to the neural net controller 4 described above. In the figure, a neuro prediction model 20 is a model of the refuse incinerator 1, and is composed of a neural network like the neural network controller 4 described above. When the secondary air A2 is controlled by the final operation amount C6 according to the final operation amount C6 supplied to the secondary air supply device (not shown) of the refuse incinerator 1, the neuro prediction model 20 is:
The output of the refuse incinerator 1 (for example, oxygen concentration S1, NOx concentration S2) is predicted, and the predicted output of the refuse incinerator 1 is added to the neural network controller 4 as a predicted value K1. Supply to the container 6. Also,
Oxygen concentration S which is the actual output parameter of the refuse incinerator 1
1, the NOx concentration S2, the internal pressure in the adder 7,
It is added to the disturbance parameter P of the dust input amount and supplied to the adder 6 as a parameter PP. The adder 6 takes the deviation D1 between the two by adding the predicted value K1 and the parameter PP, and feeds this back to the adder 8. The adder 8 receives the deviation D1 and the target value TV
And the difference D2 between the calculated value and the difference is supplied to the adder 9.
The adder 9 takes the difference D3 between the two by adding the difference D2 and the parameter PP, and supplies this to the PID controller 3. The PID controller 3 is a normal control device, as in the above-described embodiment, finds the manipulated variable C4 according to the difference D3, and supplies it to the adder 10. On the other hand, the neural network controller 4 obtains the manipulated variable C5 such that the predicted value K1 becomes the target value TV according to the predicted value K1 and the target value TV, and supplies it to the adder 10. The adder 10 adds the manipulated variables C4 and C5 and supplies the final manipulated variable C6 of the secondary air amount to the refuse incinerator to a secondary air supply device (not shown). The secondary air supply device is
Secondary air A2 to the refuse incinerator 1 according to the operation amount C6
Control the supply amount of.

According to the above-mentioned structure, an appropriate amount of dust is first put into the incinerator 1. The dust is blown up by the primary air A1 and the combustion is started by a predetermined ignition device. The neural network controller 4 obtains a manipulated variable C5 such that the predicted value K1 output by the neuro prediction model 20 becomes the target value TV. On the other hand, in the PID controller 3, the actual oxygen concentration S1, N in the refuse incinerator 1
According to the Ox concentration S2, the disturbance parameter P, and further the predicted value K1 output by the neuro prediction model 20, the manipulated variable C4 is set so that these values become the target value TV.
Is required. Then, the operation amount C5 and the operation amount C
4, the supply amount of the secondary air A2 to the refuse incinerator 1 is controlled. And, even if garbage is thrown in properly,
The PID controller 3 and the neural network controller 4 more optimally control the amount of the secondary air A2.

As described above, in the second embodiment, in the feedback control system, the prediction model is formed by the neural network, and the neural network controller is provided in parallel with the conventional PID controller, so that the time-varying disturbance is prevented. There is a feature that it works very effectively even in some cases. As a result, the robustness is improved, and if the accuracy of the prediction model is improved by the learning function of the neuron, the robustness becomes even more robust. Furthermore, since the neuro prediction model is directly inserted online in the control system, learning is fast and the control system can be quickly dealt with.

[0016]

As described above, according to the first aspect of the invention, the first control amount is calculated by the first control means according to the parameter of the controlled object which can be mathematically expressed. The second manipulated variable is calculated by the neural network of the second control means according to the parameter of the controlled object that cannot be mathematically expressed, and finally the first manipulated variable and the second manipulated variable are calculated. Since the controlled object is controlled according to the third manipulated variable added by the adding means, there is an advantage that optimal control can be performed even for a disturbance that cannot be expressed by a mathematical expression.

According to the invention described in claim 2, the first
The control means calculates the first manipulated variable that controls the amount of air supplied to the incinerator according to the components of the exhaust gas discharged from the incinerator, and the second control means controls the neural network. In addition to the components of the exhaust gas,
A second operation amount for operating the amount of air supplied to the incinerator is calculated according to the disturbance in the incinerator, and finally,
Since the amount of air to be supplied to the incinerator is controlled according to the first and second manipulated variables, optimum control can be performed even for disturbances that cannot be mathematically expressed, and the incombustible gas in the exhaust can be reduced. Is obtained.

[Brief description of drawings]

FIG. 1 is a block diagram showing a configuration of a first exemplary embodiment of the present invention.

FIG. 2 is a conceptual diagram showing a basic configuration of a neural network used in the embodiment.

FIG. 3 is a block diagram showing a configuration of a second exemplary embodiment of the present invention.

[Explanation of symbols]

 1 incinerator 3 PID controller (first control means) 4 neural network controller (second control means) 5,10 adder (addition means) C1, C4 operation amount (first operation amount) C2, C5 operation amount (Second manipulated variable) S1 oxygen concentration (exhaust gas component) S2 NOx concentration (exhaust gas component) A1 primary air P disturbance parameter

─────────────────────────────────────────────────── ─── Continuation of the front page (51) Int.Cl. ⁵ Identification number Reference number within the agency FI Technical indication location G06F 15/18 8945-5L G06G 7/60 (72) Inventor Shigeki Murayama Toyosu 3-chome, Koto-ku, Tokyo No. 1-15 Ishikawajima Harima Heavy Industries Co., Ltd. Toni Technical Center (72) Inventor Hideki Kidori 3-15-15 Toyosu, Koto-ku, Tokyo Ishikawajima Harima Heavy Industries Co., Ltd. Toji Technical Center

Claims (2)

[Claims] 1. A first control means for calculating a first manipulated variable according to a parameter of a controlled object which can be mathematically expressed, and a neural network, which is configured according to a parameter of the controlled object which cannot be mathematically expressed. Second control means for calculating a second manipulated variable, the first manipulated variable and the second manipulated variable are added,
An addition means for calculating a third manipulated variable;
A feedback control device that controls the controlled object based on the operation amount of. 2. An incinerator for incinerating a material to be burned, and a first operation amount for operating an amount of air supplied to the incinerator according to a component of exhaust gas discharged from the incinerator. A second control amount configured to control the amount of air supplied to the incinerator according to a disturbance in the incinerator in addition to the components of the exhaust gas. And a control means for controlling the air amount according to the first and second operation amounts.