JP2009128972A - Control device for plant and control device for thermal power generation plant - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a control device for a plant excellently maintaining a control property of the plant even when a property of a prediction model differs from a property of the actual plant. <P>SOLUTION: This control device for the plant has: an operation signal generation means for generating an operation signal to the plant; the model simulating the property of the plant; a learning means for generating an input signal of the model such that an output signal simulated by the model satisfies a preset target; and a learning signal generation means for calculating a learning signal according to a learning result in the learning means. The control device for the plant also has: an operation result evaluation means for calculating a first error and a second error that are errors between target values, and a first measurement signal and a second measurement signal of the plant acquired as results obtained by imparting a certain operation signal and the updated operation signal to the plant; and a correction signal generation means for generating a correction signal of the operation signal generated by the operation signal generation means when the second error is larger than the first error. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to a plant control device and a thermal power plant control device that controls a thermal power plant including a boiler.

  In general, a control device that controls a plant processes a measurement signal obtained from a plant that is a control target, calculates an operation signal given to the control target, and outputs the operation signal as a control signal.

  The plant control device is provided with an algorithm for calculating an operation signal so that the measurement signal of the plant satisfies the target value.

  As a control algorithm used for plant control, there is a PI (proportional / integral) control algorithm.

  In PI control, a value obtained by integrating the deviation with time is added to a value obtained by multiplying the deviation between the plant measurement signal and its target value by a proportional gain, and an operation signal to be given to the control target is calculated.

  The control algorithm using PI control can describe the input relationship with a block diagram, etc., so the causal relationship between input and output is easy to understand, and PI control is a stable / safe control algorithm in plant control. There are many actual application results.

  However, when the plant is operated under conditions that are not assumed in advance, such as a change in the plant operation mode or a change in the environment, an operation such as changing the control logic may be required.

  On the other hand, there is adaptive control that automatically corrects / changes the control method in response to changes in the operation mode and environment of the plant.

  As a plant control method using a learning algorithm which is one of the methods, there is a technique described in, for example, Japanese Patent Laid-Open No. 2000-35956.

  In a plant control method using a learning algorithm, which is a technique described in Japanese Patent Application Laid-Open No. 2000-35956, a model for predicting characteristics of a control target in a control device and a model input for achieving a target value of a model output A learning unit for learning the generation method is provided.

  Further, as a learning algorithm, the literature “reinforcement learning” gives a positive evaluation value when the measurement signal achieves the driving target value, and based on this evaluation value, Actor-Critic, Q learning, real-time dynamic programming A method of learning an operation signal generation method using an algorithm such as the above is described.

JP 2000-35956 A “Reinforcement Learning”, co-translation by Sadayoshi Mikami and Masaaki Minagawa, published by Morikita Publishing Co., Ltd., December 20, 2000, 142-172, 247-253

  When applying the plant control technique using the learning algorithm described in JP 2000-35956 A to plant control, the model for predicting the characteristics of the plant to be controlled does not match the characteristics of the actual plant. In this case, there is a difference between the predicted value of the model and the measured value of the plant.

  Therefore, even if the operating conditions are optimal with the predicted values of the model, they are not optimal operating conditions for the actual plant. Therefore, when the plant is controlled under these operating conditions, proper control cannot be performed. Become.

  As a method of plant control for avoiding the above phenomenon, there is a method of correcting a model using measured values of an actual plant so that the characteristics of the actual plant coincide with the characteristics of the model.

  However, in the above method, it takes a long time to accumulate actual plant measurement data necessary for correcting the model, and the expected control performance cannot be exhibited during the data accumulation period.

  Japanese Patent Application Laid-Open No. 2000-35956 does not describe any technique that suitably corresponds to the case where the model does not match the actual plant characteristics.

  An object of the present invention is to provide a plant control device and a thermal power plant control capable of maintaining good plant control characteristics even when the characteristics of a model for predicting the characteristics of the plant to be controlled are different from the actual plant characteristics. To provide an apparatus.

  The plant control device according to the present invention generates an operation signal to be transmitted to the plant in the plant control device that calculates an operation signal for controlling the plant using a measurement signal obtained by measuring the operation state of the plant. According to the learning result of the operation signal generating means, the model that simulates the characteristics of the plant, the learning means that generates the input signal of the model so that the output signal simulated by the model satisfies a preset target, and the learning result of the learning means Learning signal generation means for calculating a learning signal, and further, a first measurement signal and a second measurement signal of the plant obtained as a result of giving a certain operation signal and an updated operation signal to the plant, and a target value thereof The operation result evaluation means for calculating the first error and the second error, which are errors with respect to each other, and the second error calculated by the operation result evaluation means are the first error. Correction signal generation means for generating a correction signal for the operation signal generated by the operation signal generation means when the error is larger than the error, the correction signal generation means based on the feature quantity of the model characteristic extracted from the model The operation signal generation means calculates the operation signal for controlling the plant using at least the learning signal calculated by the learning signal generation means and the correction signal calculated by the correction signal generation means. It is characterized by having constituted so.

  The thermal power plant control device of the present invention is a thermal power plant control device that calculates an operation signal for controlling the thermal power plant using a measurement signal obtained by measuring the operating state of the thermal power plant. Of nitrogen oxide concentration, carbon monoxide concentration, carbon dioxide concentration, sulfide oxide concentration, mercury concentration, coal flow rate, mill classifier rotation speed, generator output contained in exhaust gas discharged from the plant boiler At least one measurement signal, and the operation signal includes at least one operation signal of an opening of an air damper of the boiler, an air flow rate, an air temperature, a fuel flow rate, and an exhaust gas recirculation flow rate. , Operation signal generating means for generating an operation signal to be transmitted to the thermal power plant, a model for simulating the characteristics of the thermal power plant, and an output simulated by the model Learning means for generating an input signal of the model so that the signal satisfies a preset goal, and learning signal generation means for calculating a learning signal according to a learning result in the learning means, and further, an operation signal and update A first error and a second error, which are errors between the first measurement signal and the second measurement signal of the thermal power plant acquired as a result of giving the operation signal to the plant and the target value, are calculated. An operation result evaluation unit; and a correction signal generation unit that generates a correction signal of the operation signal generated by the operation signal generation unit when the second error calculated by the operation result evaluation unit is larger than the first error. The model includes a plurality of models corresponding to a burner pattern, a load level, and a coal type of a boiler of a thermal power plant, and the operation result evaluation means is provided for a boiler mill. A function to grasp the burner pattern based on the value of the measurement signal of the coal flow rate, a function to grasp the load level based on the value of the measurement signal of the output demand or generator output, and the measurement signal of the mill classifier rotation speed A function of grasping the coal type based on the value, wherein the learning signal generating means generates a learning signal according to a result learned using a model corresponding to each condition grasped by the operation result evaluating means. And the operation signal generation means calculates the operation signal for controlling the thermal power plant using at least the learning signal calculated by the learning signal generation means and the correction signal calculated by the correction signal generation means. It is characterized by comprising.

  ADVANTAGE OF THE INVENTION According to this invention, even when the characteristic of the model which estimates the characteristic of the plant to be controlled is different from the characteristic of the actual plant, the plant control device and the control of the thermal power plant that can maintain the plant control characteristic well A device can be realized.

  Next, a plant control apparatus and a thermal power plant control apparatus according to embodiments of the present invention will be described with reference to the drawings.

  FIG. 1 is a control block diagram showing the overall configuration of a plant control apparatus according to a first embodiment of the present invention.

  In FIG. 1, the control device of the plant 100 of this embodiment is controlled by a control device 200.

  The control device 200 includes a measurement signal conversion unit 300, an operation result evaluation unit 400, a reference signal generation unit 500, a learning signal generation unit 600, a learning unit 610, a model 620, a correction signal generation unit 700, and an operation signal generation as arithmetic units. Each means 800 is provided.

  The control device 200 includes a measurement signal database 230, a model construction database 240, a learning result information database 250, a correction signal calculation database 260, and an operation signal database 270 as databases.

  The control device 200 has an external input interface 210 and an external output interface 220 as interfaces with the outside.

  The control device 200 receives the measurement signal 1 obtained by measuring various state quantities of the plant from the plant 100 via the external input interface 210, and also via the external output interface 220. For example, an operation signal 23 is sent from the control device 200 to control the flow rate of the working fluid supplied to the plant 100.

  The measurement signal 1 of various state quantities of the plant 100 taken into the control device 200 from the plant 100 is stored in the measurement signal database 230 which is a database provided in the control device 200 as the measurement signal 2 after passing through the external input interface 210. The

  The operation signal 22 generated by the operation signal generation means 800 that is an arithmetic device provided in the control device 200 is transmitted from the operation signal generation means 800 to the external output interface 220 and is an operation that is a database provided in the control device 200. It is stored in the signal database 270.

  Further, the operation result evaluation unit 400 that is an arithmetic device provided in the control device 200 calculates the flags 12, 13, 14, and 15 using the measurement signal 3 stored in the measurement signal database.

  The flags 12, 13, 14, and 15 calculated by the operation result evaluation unit 400 are the reference signal generation unit 500, the learning signal generation unit 600, the correction signal generation unit 700, and the operation signal generation which are arithmetic units provided in the control device 200. Each is transmitted to means 800.

  The reference signal generation unit 500, the learning signal generation unit 600, the correction signal generation unit 700, and the operation signal generation unit 800 determine whether or not to operate based on the transmitted values of the flags 12, 13, 14, and 15. (For example, it does not operate when the flag value is 0 and operates when the flag value is 1).

  A method for determining the flag value will be described later with reference to FIG.

  In the operation signal generating unit 800 of the arithmetic unit provided in the control device 200, the reference signal 17 generated by the reference signal generating unit 500, the learning signal 16 generated by the learning signal generating unit 600, and the correction generated by the correction signal generating unit 700. The operation signal 22 is calculated using each of the signals 18.

  The reference signal generation means 500 of the arithmetic device provided in the control device 200 calculates the reference signal 17 using the measurement signal 4 stored in the measurement signal database 230.

  The reference signal generating means 500 is composed of a PI control (proportional / integral control) circuit or the like, and is configured to calculate the reference signal 17 based on a predesigned logic.

  In the learning signal calculation means 600 of the arithmetic device provided in the control device 200, the measurement signal 4 stored in the measurement signal database 230, the operation signal 21 stored in the operation signal database 270, and the learning information database 250 are stored. A learning signal 16 is calculated using the stored learning information data 11.

  The learning information data 11 stored in the learning information database 250 is created using the learning information data 9 from the learning unit 610 and the model output 7 from the model 620.

  The arithmetic device model 620 provided in the control device 200 has a function of simulating the control characteristics of the plant 100 using a statistical model or a physical model built therein.

  The operation signal 23 generated by the control device 200 is given to the plant 100 via the external output interface 220, and the measurement signal 1 of the plant 100 as the control result is received by the control device 200 via the external input interface 210.

  By operating these conditions in combination with the learning means 610 and the model 620, the control characteristics of the plant 100 are simulated and output as the model output 7.

  That is, the model input 8 generated by the learning unit 610 is given to the model 620, and the model output 7 as the control result is received by the learning unit 610.

  The model 620 uses the model construction data 6 stored in the model construction database 240 to calculate and output a model output 7 corresponding to the model input 8 input from the learning unit 610.

  The model 620 is constructed by a statistical model such as a neural network or a physical model of the plant 100, for example.

  The model construction database 240 is used to construct the actual machine data 5 and the model 620 generated by removing noise included in the measurement signal 4 output from the measurement signal database 230 by the measurement signal conversion unit 300. Model building data 6 such as necessary model parameters is stored.

  The learning means 610 of the arithmetic device provided in the control device 200 learns how to generate the model input 8 such that the model output 7 calculated by the model 620 becomes a desired value.

  Parameters used for learning, such as a target value of the model output 7, are stored in the learning information database 250, and learning is performed by the learning means 610 using the stored learning information data 10.

  Reinforcement learning is a method for implementing the learning means 610. In this reinforcement learning, the model input 8 is generated by trial and error in the initial stage of learning.

  Thereafter, the learning means 610 can generate the model input 8 such that the model output 7 calculated by the model 620 becomes a desired value as the learning proceeds.

  As such a learning algorithm, the “reinforcement learning” in the literature gives a positive evaluation value when the measurement signal achieves the driving target value, and based on this evaluation value, Actor-Critic, Q learning, real-time dynamic A method of learning an operation signal generation method using an algorithm such as Programming is described.

  The learning means 610 can apply various optimization methods such as an evolutionary calculation method in addition to the above-described reinforcement learning method.

  Learning information data 9 that is a result of learning by the learning means 610 is stored in the learning information database 250.

  In the correction signal generation means 700 of the arithmetic unit provided in the control device 200, the operation signal 21 stored in the operation signal database 270, the measurement signal 4 stored in the measurement signal database 230, and the learning information database 250 are stored. The correction signal 18 is calculated using the learning information data 19 stored, the correction signal calculation data 20 stored in the correction signal calculation database 260, and the learning information data 19 stored in the learning information database 260.

  The correction signal generation unit 700 has a function of increasing the value of the operation signal, a function of decreasing the value of the operation signal, a function of holding the value of the operation signal at the current value, a function of returning the value of the operation signal to the original value, It is configured to have at least one function among functions for matching the value of the operation signal with a predetermined value.

  Further, the correction signal generation means 700 can extract the feature amount of the model 620 and calculate the correction signal 18. Details of the operation of the correction signal generation means 700 will be described later.

  Then, the correction signal 18 generated by the correction signal generation unit 700 is output to the operation signal generation unit 800 so as to correct the operation signal 22 generated by the operation signal generation unit 800.

  The operation signal generation unit 800 calculates and outputs the operation signal 22 based on the correction signal 18 or the like input from the correction signal generation unit 700, and the control device 200 outputs the operation signal 22 to the plant 100 via the external output interface 220. For example, the operation signal 23 for controlling the air flow rate of the boiler burner and the air port is output.

  As shown in FIG. 1, an external input device 900 including a keyboard 901 and a mouse 902, a maintenance tool 910, and an image display device 950 are installed in the vicinity of the control device 200.

  Then, an operator of the plant 100 generates the maintenance tool input signal 51 using the external input device 900 including the keyboard 901 and the mouse 902 and inputs this signal to the maintenance tool 910, thereby arranging the maintenance tool input signal 51 in the control device 200. The information of various databases being displayed can be displayed on the image display device 950.

  The maintenance tool 910 includes an external input interface 920, a data transmission / reception unit 930, and an external output interface 940.

  The maintenance tool input signal 51 generated by the external input device 900 is taken into the maintenance tool 910 via the external input interface 920.

  The data transmission / reception unit 930 of the maintenance tool 910 is configured to acquire the database information 50 from various databases arranged in the control device 200 according to the information of the maintenance tool input signal 52.

  The data transmission / reception processing unit 930 of the maintenance tool 910 transmits a maintenance tool output signal 53 obtained as a result of processing the database information 50 to the external output interface 940.

  The external output interface 940 transmits an output signal 54 based on the maintenance tool output signal 53 to the image display device 950 and displays it on the image display device 950.

  In the plant control apparatus 200 according to the above-described embodiment of the present invention, the measurement signal database 230, the model construction database 240, the learning result information database 250, and the correction signal calculation that constitute the database provided in the control apparatus 200 are described. Database 260, operation signal database 270, measurement signal conversion means 300, operation result evaluation means 400, reference signal generation means 500, learning signal generation means 600, learning means 610, model 620, correction signal generation means 700, and operation signal Although the generation unit 800 is disposed inside the control device 200, all or some of them may be disposed outside the control device 200.

  FIG. 2 is a flowchart showing a control procedure in the plant control apparatus according to the first embodiment shown in FIG.

  In FIG. 2, the control apparatus 200 of the plant 100 executes the control of the plant 100 by combining the steps 1000, 1010, 1020, 1030, 1040, 1050, and 1060 of this flowchart.

  Further, as shown in FIG. 2, the control device 200 has three operation modes of A, B, and C.

  First, in step 1000 for determining the initial mode, the initial mode is determined. In this embodiment, the initial mode is mode A.

  Then, the process proceeds to step 1010 for mode determination. In step 1010 for mode determination, mode determination is performed. If the operation mode is A mode, the process proceeds to step 1020 for generating an operation signal. If the operation mode is B mode, the operation is performed. The process proceeds to step 1030 for signal generation, and to step 1040 for operation signal generation when the operation mode is the C mode.

  In each of the operation signal generation steps 1020, 1030, and 1040 determined and advanced in the mode determination step 1010, the operation signal 22 is generated by operating the operation signal generation means 800 provided in the control device 200.

  The operation result evaluation unit 400 provided in the control device 200 is operated, and the operation signal evaluation unit 400 uses the measurement signal 3 stored in the measurement signal database to perform the reference signal generation unit 500 and learning as necessary. Flags 12, 13, and 15 for operating the signal generation unit 600 and the correction signal generation unit 700 are generated.

  When the operation mode is set to the initial mode A and the process proceeds to step 1020 for generating the operation signal, the operation signal generation unit 800 is informed using the reference signal 17 generated by the reference signal generation unit 500. The operation signal 22 is calculated based on the equation (1) provided as an arithmetic function.

  Here, So is an operation signal, and Sb is a reference signal.

So = Sb (1)
When the operation mode shifts to the B mode and the operation signal generation step 1030 is performed, the reference signal 17 generated by the reference signal generation unit 500 and the learning signal 16 generated by the learning signal generation unit 600 are obtained. The operation signal 22 is calculated based on the equation (2) provided in the operation signal generation means 800 as an arithmetic function.

  Here, S1 is a learning signal.

So = Sb + Sl (2)
Further, when the operation mode shifts to the C mode and proceeds to step 1040 of operation signal generation, the reference signal 17 generated by the reference signal generation unit 500, the learning signal 16 generated by the learning signal generation unit 600, Using the correction signal 18 generated by the correction signal generation means 700, the operation signal 22 is calculated based on Expression (3) provided as an arithmetic function in the operation signal generation means 800.
Here, Sr is a correction signal.

So = Sb + Sl + Sr (3)
In the control device 200 according to the present embodiment, the operation signal 22 is calculated using the equations (1) to (3) provided in the operation signal generation unit 800 as an arithmetic function. A function for preventing the operation signal So from changing suddenly using a rate limiter, or a function for limiting the value of the operation signal within a preset range using an upper / lower limit setting unit may be added.

  The operation signal is transmitted to the plant and the operation result is acquired after proceeding through any one of the steps 1020, 1030, and 1040 of the operation signal generation that proceeds based on the modes A, B, and C determined in the mode determination. In step 1050, the operation signal 22 from the operation signal generating means 800 generated in any one of the steps 1020, 1030, and 1040 is transmitted to the plant 100 as the operation signal 23 via the external output interface 220.

  As a result of giving the operation signal 23 to the plant 100, the measurement signal 1, which is a state quantity indicating the operation state, is acquired from the plant 100 and stored in the measurement signal database 230 of the control device 200.

  In step 1060 of the operation result evaluation and next mode determination that proceeds after the step 1050, the operation result evaluation unit 400 provided in the control device 200 uses the measurement signal 3 stored in the measurement signal database 230 to configure the plant 100. Evaluate the operation result and determine the next mode.

  As described above, the initial mode of the operation mode for controlling the plant 100 is the mode A set in step 1000 for determining the initial mode.

  Thereafter, the mode is shifted to the B mode in which the learning signal generated by the learning signal generating means 600 is used together to generate the operation signal.

  When the operation mode shifts to the B mode, the operation result evaluation unit 400 gives the plant 100 the first measurement signal Sm1 acquired as a result of giving the operation signal 22a1 to the plant 100 and the updated operation signal 22a2. As a result, the second measurement signal Sm2 acquired from the plant 100 and the target value Sa of these measurement signals Sm1 and Sm2 are used to determine the error between the first measurement signal Sm1 and the target value Sa. 1, and the second error E2 which is an error between the second measurement signal Sm2 and the target value Sa is calculated by the operation result evaluation unit 400 based on the equations (4) and (5). To calculate.

  Here, E1 is the first error, E2 is the second error, Sm1 is the first measurement value, Sm2 is the second measurement value, and Sa is the target value of the measurement signal.

  As the measurement values Sm1 and Sm2, an average value for a certain period, an instantaneous value, or the like is used.

  ABS (a) means a function for calculating the absolute value of a.

E1 = ABS (Sm1-Sa) (4)
E2 = ABS (Sm2-Sa) (5)
When the second error E2 is larger than the first error E1, the mode is shifted to mode C, and when the second error E2 is smaller than the first error E1, the mode B is set.

  That is, as a result of giving the updated operation signal to the plant 100, when the control characteristics deteriorate (the error between the measurement signal and the target value increases), the operation mode shifts to the C mode, and the operation signal generation means 800 The operation signal 22 is calculated using the correction signal 18 generated by the correction signal generation means 700.

  As will be described later, the correction signal generator 700 generates the correction signal 18 so as to maintain good control characteristics.

  As a result, an effect of improving the control performance of the plant 100 is obtained.

  In addition, in the operation signal generation step 1040, in addition to the method of calculating the operation signal 22 using the expression (3) provided in the operation signal generation means 800 as an operation function, the operation function expression (4) and the expression The operation signal 22 can also be calculated using the errors E1 and E2 calculated in (5) or based on the equations (6) to (8).

  Α and β are weight parameters, and ε is a predetermined design parameter.

So = Sb + α × Sl + β × Sr (6)
α ← α−ε (E2−E1) (7)
β ← β + ε (E2 − E1) (8)
Here, equation (6) means that ε (E2−E1) is subtracted from the previous value of α to update the value of α.

  By calculating the operation signal 22 using the equations (6) to (8) provided as the operation function in the operation signal generation means 800, the larger the value of E2-E1, the larger the β of the correction signal Sr is the weight parameter. The value increases, and the influence of the value of the correction signal 18 generated by the correction signal generation unit 700 on the operation signal 22 generated by the operation signal generation unit 800 increases.

  FIG. 3 is a setting screen for information stored in the correction signal calculation database 260 in the control apparatus 200 which is an embodiment of the plant control apparatus shown in FIG.

  The correction signal generation means 700 of the control device 200 calculates the correction signal 18 using information stored in the correction signal calculation database 260, which is various information set on the screen shown in FIG.

  By using the screen of FIG. 3, the operator of the plant 100 has a function of increasing the value of the operation signal, a function of decreasing the value of the operation signal, and a function of holding the value of the operation signal at the current value for each operation end. , The function to return the value of the operation signal to the original value, the function to match the value of the operation signal with a predetermined value, and the function to calculate based on the model feature can be arbitrarily set .

  In addition, the operation range of the operation amount can be set for each operation end.

  Examples of the model feature amount include model parameters stored in the model construction database 240 of the control apparatus 200, polarization point information of the model output characteristic curve, and the like.

  4 and 5 are diagrams for explaining each operation by the correction signal generating means 700 provided in the control device 200 which is an embodiment of the plant control device shown in FIG.

  4 (a) to 4 (d) in FIG. 4, the horizontal axis represents the operation amount to the model or the actual machine, and the vertical axis represents the controlled quantity of the model or the actual machine. An example showing model characteristics of a model 620 in 200 and a learning result thereof is illustrated.

  4 (a) to 4 (d) are displayed assuming that the operation condition that minimizes the controlled amount of the model in this embodiment is searched by the learning means 610 of the control device 200. Yes.

  In FIG. 4A, α1 on the horizontal axis indicates the current operation amount, and β1 indicates the operation amount after learning (searching).

  As shown in FIG. 4A, the controlled variable indicated by the model characteristic curve is reduced on the model 620 by updating the manipulated variable from α1 to the manipulated variable β1 after learning (searching).

  Therefore, if the model characteristic and the plant characteristic of the plant that is the actual machine match, the controlled variable of the plant 100 with respect to the manipulated variable is reduced by updating the manipulated variable from α1 to β1.

  However, if there is a model error and the model characteristics do not match the actual plant characteristics, operating the plant 100 using the learned results may increase the controlled amount of the plant 100.

  FIG. 4B shows an example in which, as a result of updating the manipulated variable from α1 to β1 after learning, the model characteristic curve rises and the controlled amount of the plant 100 increases.

  By the way, the deterioration of the control characteristics due to the model error reduces the model error by correcting the model 620 using the newly acquired measurement signal 1 of the plant 100, and then controls the corrected model 620 as a target. It can be suppressed by re-learning the operation method for the model 620 by the 200 learning means 610.

  However, when this method is used, it often takes a long time to accumulate data of a large number of measurement signals 1 for correcting the model 620, and the expected control performance cannot be exhibited during the data accumulation period. Become.

  In particular, when the plant 100 is not allowed to operate with poor control characteristics over a long period of time, it is difficult to apply the method described above.

  Therefore, the correction signal generation means 700 provided in the control device 200 of the present embodiment solves the problem caused by the above method.

  That is, in the control device 200 of the present embodiment, when the operation amount for the model 620 is updated from α1 to β1 after learning, and the controlled amount of the plant 100 to be controlled increases, the operation mode of the control device 200 is The initial mode is changed from the A mode to the C mode (see FIG. 2).

  In the changed C mode, the correction signal generating means 700 provided in the control device 200 generates the correction signal 18 for maintaining good control characteristics, and the operation signal generating means 800 based on the correction signal 18. An operation signal 22 serving as a command signal for the plant 100 to be controlled is generated.

  The correction signal generation means 700 has a function of generating the correction signal 18 using the model characteristic curve shown in FIG.

  That is, the correction signal generation unit 700 calculates the number of polarization points of the model characteristic curve based on the model characteristic curve, and calculates the correction signal 18 based on the number of polarization points.

  When the number of polarization points of the model characteristic curve is an even number, the operation amount is decreased, and when the number is an odd number, the operation amount is increased.

  The reason why the controlled amount is reduced by the above-described method and the model characteristic approaches the desired characteristic will be described with reference to FIGS. 4 (a) to 4 (f).

  First, as shown in FIG. 4A, a case where the number of polarization points of a model characteristic curve having an operation amount between α1 and β1 is an even number will be described.

  In the example of the model characteristic curve shown in FIG. 4A, the number of polarization points is 0, which is an even number.

  In this case, as shown in FIG. 4B, if the position of the polarization point in the actual machine characteristic represented by the solid line as the actual machine characteristic curve is between α1 and β1 of the manipulated variable, the controlled variable at the manipulated variable β1 is It becomes larger than the controlled amount at the time of the operation amount α1.

  Further, when the position of the polarization point in the actual machine characteristic represented by the solid line as the actual machine characteristic curve is larger than β1, the control quantity when the manipulated variable β1 is as shown in FIG. 4C is the manipulated variable α1. It becomes smaller than the controlled amount at the time.

  Therefore, when the actual machine characteristic is such that the position of the polarization point is between α1 and β1 of the manipulated variable as shown in FIG. 4B, the controlled variable is decreased by reducing the manipulated variable. Is possible.

  That is, when the number of polarization points of the model characteristic curve represented by a broken line is an even number, the controlled variable of the plant 100 can be reduced by decreasing the manipulated variable to approach α1.

  Next, as shown in FIGS. 4D to 4F, the model 620 includes models having different model characteristic curves, and the polarization of the model characteristic curve between the manipulated variables α2 and β2 is performed. A case where the number of points is an odd number will be described.

  In the example shown in FIG. 4D, the number of polarization points of the model characteristic curve is one and is an odd number.

  In this case, as shown in FIG. 4E, when the position of the polarization point in the actual machine characteristic curve represented by the solid line is closer to the operation amount β2 than the polarization point of the model characteristic curve represented by the broken line, The controlled amount becomes larger than the controlled amount when the operation amount α2.

  On the contrary, when the position of the polarization point in the actual machine characteristic curve is closer to the operation amount α2 than the polarization point of the model characteristic, the actual machine characteristic curve becomes as shown by the solid line in FIG. The controlled amount at the time is smaller than the controlled amount at the operation amount α2.

  Therefore, when the actual machine characteristic is an actual machine characteristic curve as shown in FIG. 4E, the controlled amount with respect to the manipulated variable can be decreased by increasing the manipulated variable.

  That is, when the number of polarization points of the model characteristic curve is an odd number, the controlled variable of the plant 100 to be controlled can be decreased by decreasing the manipulated variable.

  In FIG. 4, the case of FIGS. 4A to 4C where the number of polarization points of the model characteristic curve is an even number is shown on the left as pattern 1, and the number of polarization points of the model characteristic curve is an odd number. The cases of FIG. 4D to FIG. 4F are represented as pattern 2 on the right side.

  Therefore, the correction signal generation unit 700 is configured to generate the correction signal 18 for increasing or decreasing the operation amount and output it to the operation signal generation unit 800 using the information on the number of polarization points of the model characteristic curve. ing.

  Further, a graph emphasizing the position of the polarization point of the model characteristic curve as shown in FIGS. 4A and 4D, and the number of the polarization points of the model characteristic curve shown in FIG. By displaying the image on the image display device 950 of the maintenance tool 910 attached to the operator, the operator of the plant 100 can know the basis for generating the correction signal 18 generated by the correction signal generation unit 700, and the validity of the correction signal 18 Can be evaluated.

  FIG. 5 is a diagram illustrating an operation of learning an optimum operation method for the plant 100 by the correction signal generation unit 700 in the control device 200 of the present embodiment.

  The search range for learning is determined by combining learning means 610 and model 620 of control device 200.

  In FIG. 5 (a), the learning means 610 minimizes the error between the model output and the target value, as shown by a solid line in the model characteristics when the plant 100 is controlled with the manipulated variable on the horizontal axis and the controlled variable on the vertical axis. A predetermined threshold range is set as a trial and error range centering on the operating conditions.

  Next, in the correction signal generation means 700, the error between the measurement signal of the plant 100 and the target value of the measurement signal is minimized by setting the operation amount search range for changing the operation signal within the range of trial and error determined by the learning means 610. The operation signal is calculated so that

  And the learning result by the learning means 610 is preserve | saved in the learning information database 250 in the form as shown, for example in FIG.5 (b).

  In FIG. 5B, the correction signal generation means 700 generates a correction signal 18 that decreases the manipulated variable when the controlled variable is larger than γ1, and increases the manipulated variable when the controlled variable is smaller than γ1. Means.

  By displaying the correction signal graph shown in FIG. 5B on the image display device 950, the operator of the plant 100 can know the basis for generating the correction signal 18, and the validity of the correction signal 18 can be confirmed. Can be evaluated.

  FIG. 6 is a diagram for explaining an interface for changing the range of trial and error for changing the operation signal performed by the correction signal generation means 700 provided in the control device 200 of the present embodiment.

  By using various types of information shown in the screen of FIG. 6, the search range of the operation amount described in FIGS. 5A and 5B is compared with the search range 1 based on the threshold in the screen of FIG. As shown in the search range 2 based on manual correction, it can be set manually.

  As a result, an operator who is familiar with the operation of the plant 100 sets the search range of FIG. 5, thereby obtaining an effect that the plant 100 can be operated safely even during the search for the optimum operation condition.

  According to the above-described embodiment of the present invention, there is provided a plant control apparatus that can satisfactorily maintain the plant control characteristics even when the characteristics of the model for predicting the characteristics of the plant to be controlled are different from the actual plant characteristics. realizable.

  Next, a thermal power plant control apparatus according to a second embodiment of the present invention will be described with reference to the drawings.

  FIG. 7 is a control block diagram showing the overall configuration of a thermal power plant control apparatus according to the second embodiment of the present invention.

  Further, in FIG. 8 described later, as a thermal power plant to be controlled by the thermal power plant control apparatus according to the second embodiment of the present invention shown in FIG. 7, thermal power generation provided with a boiler using coal as fuel. The schematic structure of a plant is shown.

  The thermal power plant control apparatus according to the second embodiment shown in FIG. 7 has the same basic structure as the plant control apparatus according to the first embodiment shown in FIG. The description about is omitted, and only different components will be described below.

  In the control device for the thermal power plant of FIG. 7, the thermal power plant 100 a including the boiler 101 using coal as fuel is controlled by the control device 200.

  In the control device 200 of the present embodiment, measurement signals 1 which are various state quantities of the thermal power plant 100a, for example, the oxygen concentration of the combustion gas at the outlet of the boiler 101 or the carbon monoxide concentration are taken into the control device 200. In addition, an operation signal 23 for controlling, for example, the burner 102 of the boiler 101 and the air flow rate of the air port 103 is sent from the control device 200 to the thermal power plant 100a via the external output interface 220.

  By using the control device 200 of the present embodiment, at least one of the opening degree of the air damper, the air flow rate, the air temperature, the fuel flow rate, and the exhaust gas recirculation flow rate of the boiler 101 provided in the thermal power plant 100a is targeted. By controlling, it is possible to control at least one of nitrogen oxide, carbon monoxide concentration, carbon dioxide concentration, sulfide oxide concentration and mercury concentration contained in the exhaust gas discharged from the thermal power plant 100a to a desired value. It becomes possible.

  In the model 620 constituting the control device 200 in this embodiment, the opening degree of the air damper, the air flow rate, the air temperature, the fuel flow rate, and the exhaust gas recirculation flow rate, which are the model construction data 6 stored in the model construction database 240. Is input, and the values of nitrogen oxide, carbon monoxide concentration, carbon dioxide concentration, sulfide oxide concentration and mercury concentration at that time are predicted and output.

  The plant control apparatus of the first embodiment shown in FIG. 1 and the thermal power plant control apparatus of the second embodiment shown in FIG. 7 are the same as those in the second embodiment shown in FIG. The control device 200 is different in that a plurality of learning means 610 and models 620 are provided.

  In the control apparatus 200 of the present embodiment, by providing a plurality of learning means 610 and models 620 as described above, it is possible to cope with switching of operating conditions of the thermal power plant 100a.

  When the thermal power plant 100a performs an operation of changing the burner pattern, load level, and coal type of the boiler 101, the plant characteristics greatly change.

  In the present embodiment, a model switching unit 630 is installed in the control device 200 as a method for maintaining a good operating state of the thermal power plant 100a even when the plant characteristics change greatly, and various operations are performed in the model switching unit 630. A plurality of models 620 corresponding to the conditions are prepared, and a plurality of learning means 610 for learning the operation method are arranged for each of the models 620 prepared corresponding to these various driving conditions. ing.

  In addition, there are as many types of learning information data 9 learned by the learning means 610 prepared for the model 620 provided for each driving condition in the present embodiment as many as the types of models 620.

  The learning results learned by these learning means 610 are stored in the learning information database 250.

  Further, when the learning signal 16 is generated by the learning signal generation unit 600, the determination of which learning result is used, that is, the determination of the current plant operating condition is performed by the operation result evaluation unit 400.

  FIG. 8 shows a schematic configuration of a thermal power plant including a boiler that uses coal as fuel to be controlled by the thermal power plant control apparatus according to the second embodiment of the present invention shown in FIG.

  First, the power generation mechanism of the thermal power plant 100a including the boiler 101 will be described with reference to FIG.

  In FIG. 8 (a), coal as fuel is pulverized by a mill 110 and is converted into pulverized coal to a boiler 101 through a burner 102 installed in the boiler 101 together with primary air for transporting coal and secondary air for combustion adjustment. The fuel coal is burned in the furnace of the boiler 101.

  The fuel coal and the primary air are led from the pipe 134 and the secondary air is led from the pipe 141 to the burner 102.

  Further, after-air for two-stage combustion is introduced into the boiler 101 through an after air port 103 installed in the boiler 101. This after air is guided from the pipe 142 to the after air port 103.

  High-temperature combustion gas generated by burning fuel coal inside the furnace of the boiler 101 flows downstream along the path indicated by the arrow in the furnace of the boiler 101, and the heat exchanger 106 disposed in the boiler 101. After passing through and exchanging heat, it becomes combustion exhaust gas and is discharged from the boiler 101 and flows down to the air heater 104 installed outside the boiler 101.

  The combustion exhaust gas that has passed through the air heater 104 is then released from the chimney to the atmosphere after removing harmful substances contained in the combustion exhaust gas with an exhaust gas treatment device (not shown).

  The feed water circulating in the boiler 101 is led to the boiler 101 through a feed water pump 105 from a condenser (not shown) installed in the turbine 108, and the furnace 101 of the boiler 101 is installed in the heat exchanger 106 installed in the furnace of the boiler 101. It is heated by the combustion gas flowing down the inside and becomes high-temperature and high-pressure steam.

  In this embodiment, the number of heat exchangers 106 is shown as one, but a plurality of heat exchangers may be arranged.

  The high-temperature and high-pressure steam generated in the heat exchanger 106 is guided to the steam turbine 108 via the turbine governor valve 107, and the steam turbine 108 is driven by the energy of the steam, and the generator 109 connected to the steam turbine 108 is connected to the generator 109. Rotate to generate electricity.

  Next, the primary air and the secondary air introduced into the furnace 101 of the boiler 101 from the burner 102 installed in the furnace of the boiler 101, and the after air port 103 installed in the furnace of the boiler 101 are introduced into the furnace of the boiler 101. The after-air route will be described.

  The primary air is guided from the fan 120 to the pipe 130, and is branched into a pipe 132 that passes through the inside of the air heater 104 and a pipe 131 that bypasses the air heater 104, and flows down the pipe 132 and the pipe 131. The primary air joins again in the pipe 133 and is guided to the mill 110.

  Air passing through the air heater 104 is heated by the combustion exhaust gas discharged from the furnace of the boiler 101.

  The primary air is used to convey coal (pulverized coal) generated in the mill 110 to the burner 102 through the pipe 133.

  The secondary air and the after air are led from the fan 121 to the pipe 140 and flow down the pipe 140 passing through the inside of the air heater 104 and heated, and then the secondary air pipe 141 on the downstream side of the pipe 140, The pipe is branched to an after-air pipe 142 and led to a burner 102 and an after-air port 103 installed in the furnace of the boiler 101, respectively.

  The control device 200 of the thermal power plant 100a provided with the boiler according to the present embodiment reduces the NOx and CO concentrations in the exhaust gas of the boiler, and the amount of air that is input from the burner 102 to the boiler 101 and the boiler from the after-air port 103. 101 has a function of adjusting the amount of air to be supplied to the apparatus 101.

  Various measuring devices for detecting the operating state of the thermal power plant 100a are arranged in the thermal power plant 100a, and plant measurement signals obtained from these measuring devices are transmitted to the control device 200 as measurement signals 1. Sent.

As various measuring instruments for detecting the operating state of the thermal power plant 100a, for example, in FIG. 8, a flow measuring instrument 150, a temperature measuring instrument 151, a pressure measuring instrument 152, a power generation output measuring instrument 153, and an O ₂ concentration and / or CO ₂ are measured. A concentration measuring device 154 for measuring the concentration is shown respectively.

  The flow rate measuring device 150 measures the flow rate of the feed water supplied from the feed water pump 105 to the boiler 101. Further, the temperature measuring device 151 and the pressure measuring device 152 are used to supply steam generated by heat exchange with the combustion gas flowing down the boiler 101 in the heat exchanger 106 disposed in the boiler 101 to the steam turbine 108. Measure temperature and pressure respectively.

  The amount of power generated by the generator 109 rotated by the steam turbine 108 driven by the steam generated in the heat exchanger 106 is measured by a power generation output measuring device 153.

In addition, information on the concentration of components (CO, NOx, etc.) contained in the combustion exhaust gas flowing down the boiler 101 is the O ₂ concentration and / or CO concentration provided in the boiler outlet channel on the downstream side of the boiler 101. It is measured by a concentration measuring device 154 that measures the above.

  In general, many measuring instruments other than those shown in FIG. 8 are arranged in the thermal power plant 100, but the illustration is omitted here.

  FIG. 8B is a partially enlarged view showing the air heater 104 installed on the downstream side of the boiler 101 constituting the thermal power plant 100 a and the pipes arranged in the air heater 104.

  As shown in FIG. 8B, the secondary air pipe 141 and the after-air pipe 142 branched on the downstream side of the pipe 140 arranged inside the air heater 104 and the air heater 104 are arranged. Air dampers 162, 163, 161, and 160 are disposed in the pipe 132 and the pipe 131 that bypasses the air heater 104, respectively.

  Then, by operating these air dampers 160 to 163, the area through which air passes in the pipes 131, 132, 141, 142 is changed, and the flow rate of air passing through these pipes 131, 132, 141, 142 is individually set. adjust.

  Then, using the operation signal 18 generated by the control device 200 for controlling the thermal power plant 100a and output to the thermal power plant 100a, devices such as the feed water pump 105, the mill 110, the air dampers 160, 161, 162, and 163 are used. To operate.

  In the thermal power plant control apparatus according to this embodiment, devices that adjust the state quantities of the thermal power plant, such as the feed pump 105, the mill 110, and the air dampers 160, 161, 162, and 163, are called operation terminals. A command signal necessary to operate this is called an operation signal.

  In addition, when air such as combustion or fuel such as pulverized coal is introduced into the boiler 101, a function capable of moving the discharge angle up, down, left and right is added to the burner 102 and the after air port 103 installed in the boiler 101. Then, a command signal for adjusting the mounting angle of the burner 102 and the after air port 103 can be included in the operation signal 18.

  Next, the determination method of the plant operation condition by the operation result evaluation means 400 installed in the control apparatus 200 of the present embodiment will be described with reference to FIG.

  9 (a) to 9 (c), the operation result evaluation means 400 has a function of grasping the burner pattern of the boiler 101 based on the value of the measurement signal of the coal flow rate supplied to the mill 110, the output demand or the power generation. It has a function of grasping the load level based on the value of the measurement signal output from the machine 109 and a function of grasping the coal type based on the value of the measurement signal of the mill classifier rotation speed.

  FIG. 9A is a diagram for explaining a function of grasping the burner pattern of the boiler 101 by the operation result evaluation unit 400.

  When the time on the horizontal axis is t1, the coal is supplied to the boiler 101 from the mills A, B, and D constituting the mill 110 as indicated by the flow rate of coal on the vertical axis, and the time t2 is the boundary. Then, supply of coal from the mill C starts (for example, when the time is t3, coal is supplied from all of the mills A, B, C, and D).

  As shown in the schematic diagram of the boiler on the right side of FIG. 9 (a), from the mills A, B, C and D, coal is supplied to each of five burners installed along the horizontal direction before and after the boiler 101. Supply.

  The operation result evaluation unit 400 detects the measurement signal 3 of the thermal power plant 100a through the measurement database 130, and supplies coal supplied from the mills A, B, C, and D shown in FIG. The burner pattern can be grasped from the correspondence relationship between B, C, D and the five burners installed in the boiler 101.

  By grasping the above burner pattern, it is possible to grasp the combustion state of the boiler 101 in the furnace.

  FIG. 9B is a diagram for explaining the function of grasping the load level of the thermal power plant 100a by the operation result evaluation unit 400.

  The generator output is obtained by measuring the power generation amount of the generator 109 by the power generation output measuring device 153 shown in FIG. 8, and the thermal power generation shown on the vertical axis when the time on the horizontal axis is t4 in FIG. 9B. The generator output amount of the plant 100a is M4, and the generator output amount when the time is t5 is M5.

  In this way, the load level of the thermal power plant 100a can be grasped based on the value of the measurement signal of the generator output measured by the power generation output measuring device 153.

  FIG. 9C is a diagram for explaining a first coal type discrimination method used for the fuel of the boiler 101 of the thermal power plant 100 a by the operation result evaluation unit 400.

  With respect to the mill classifier that supplies the boiler 101 with pulverized coal obtained by pulverizing fuel coal installed in the mill 110 of the thermal power plant shown in FIG. 8, FIG. The rotation speed of the mill classifier shown on the vertical axis in FIG. 4 is R6, and the rotation speed when the time is t7 is R7.

  The rotational speed of the mill classifier is adjusted so that the particle size of the pulverized coal supplied from the mill 110 to the boiler 101 becomes a target value.

  Since the hardness of the coal varies depending on the coal type, the classifier rotation speed when the particle size of the pulverized coal matches the target value varies depending on the coal type.

  Therefore, the coal type of coal supplied to the boiler 101 can be determined from the classifier rotation speed.

  The learning signal generating means 600 of the control device 200 generates the learning signal 16 according to the following procedure.

  First, the operation result evaluation means 400 has a function of grasping the burner pattern grasped by the operation result evaluation means 400 based on the measurement signal 3 of the thermal power plant 100a obtained through the measurement signal database 230, a function of grasping the load level, And the flag 13 corresponding to the plant operating condition determined by the function of grasping the coal type is generated, and this flag 13 is output to the learning signal generating means 600.

  In the model 620, learning information data 9 learned by the learning means 610 corresponding to the model 620 is generated using the model 620 corresponding to the plant operating condition determined by the operation result evaluation means 400, and the learning information data 250 is transmitted via the learning information database 250. Then, the learning signal generation unit 600 outputs the information as learning information data 11.

  The learning signal generation unit 600 is configured to calculate and generate the learning signal 16 based on the flag 13 and the learning information data 11 and output the learning signal 16 to the operation signal generation unit 800.

  FIG. 10 is a diagram illustrating a second coal type determination method in the operation result evaluation unit 400 installed in the control device 200 of the present embodiment.

  In FIG. 10 (a), the operation result evaluation unit 400 calculates the weight between the models 620 prepared by the model switching unit 630 according to various operating conditions based on the coal composition information described above. The signal generation means 700 generates the correction signal 18 by using the weight value of the model 620 calculated and the learning result of the learning means 610 and using the equation (9) provided in the correction signal generation means 700 as an arithmetic function. To do.

  Where 1 ≦ i ≦ n, n is the number of coal types, di is the distance between the i th coal and the actual sample value, Si is the value of the learning signal value generated according to the learning result using the i th coal model is there.

  Formula (9) means that the value obtained by adding all (Si / di) in the range of 1 ≦ i ≦ n is divided by the value obtained by adding all (1 / di) in the range of 1 ≦ i ≦ n. To do.

So = Σ (Si / di) / Σ (1 / di) (9)
The distance di is calculated using a calculation algorithm for the Euclidean distance and the Mahalanobis distance.

  In the drawings of component 1 and component 2 shown in FIG. 10 (a), the actual sample value subjected to the coal type determination is the distance d1 between the coal type A sample value and the coal type B sample value. If the reciprocals of the distances d1 and d2 are taken from the relationship with the distance d2, the coal type A sample value has a large weight and the coal type B sample value has a small weight.

  Therefore, the operation result evaluation unit 400 can determine the coal type of the actual sample by calculating each model 620 with the weight value described above.

  Further, as shown in the schematic diagram of the left boiler in FIG. 10 (b), the correction signal generation means 700 provided in the control device 200 of the present embodiment contains more sulfur than the threshold as fuel for the boiler 101. When using coal, the correction signal 18 is generated and output to the operation signal generating means 800 so as to increase the value of the air flow rate supplied from the burner on the furnace wall surface of the boiler 101 and the air port.

  The operation signal generation means 800 generates an operation signal 22 based on the correction signal 18 and outputs and controls it as an operation signal 23 for the thermal power plant 100a via the external interface 220.

  In addition, the graph of the air flow rate on the right side of FIG. 10B schematically shows the state of the air flow rate supplied from the burner on the furnace wall surface portion and the air port of the schematic diagram of the left side boiler of FIG. 10B. .

  In particular, when coal with a high sulfur content is used as fuel, it is important to adjust the air flow rate so as to suppress corrosion of the furnace wall surface of the boiler 101.

  Further, as shown in FIG. 10B, in the control device 200 of the present embodiment, when coal having a high sulfur content is used as the fuel for the boiler 101, the air flow rate at the furnace wall surface of the boiler 101 is increased. By generating the correction signal 18 by the correction signal generation means 700, the corrosion of the furnace wall along the furnace wall from the burner 102 and the air port 103 installed on the furnace wall of the boiler 101 can be suppressed.

  FIG. 11 is an explanatory diagram for explaining the operation of the control device 200 in the thermal power plant control device according to the second embodiment of the present invention shown in FIGS. 7 and 8.

  FIG. 11A is a diagram illustrating an example of the characteristics of the model 620 provided in the control device 200 of the present embodiment. In the learning unit 610 installed corresponding to the model 620, the model 620 is targeted. Learn operating conditions that minimize CO concentration.

  For example, by using reinforcement learning when the learning means 610 is mounted on the control device 200 and setting so that the reward increases as the CO concentration is lower (in FIG. 11A). It is possible to learn an operation method that reaches the operation condition y).

  As shown in FIG. 11A, the operating condition is taken on the blade axis and the CO concentration is taken on the vertical axis, and the relationship between the two is shown.

  When the air flow rate is the operating condition, the air flow rate supplied to the boiler 101 varies with time. Therefore, even if the operation signal matches the operating condition y, the air flow rate actually supplied to the boiler 101 changes. The CO concentration may increase.

  In order to avoid such a phenomenon, the control device 200 of the present embodiment can employ the method described below when operating the learning unit 610.

  That is, the reinforcement learning is applied when the learning unit 610 is mounted on the control device 200, and the first reward value that increases as the CO concentration decreases and the equation (10) included in the learning unit 610 as an arithmetic function are satisfied. In the case of using a remuneration obtained by adding a second remuneration having a negative value.

  In equation (10), CO (I) is an estimated CO concentration value (model output) when the operating condition is I, Δ is a minute value, and Ω is a preset threshold value.

ABS (CO (I) −CO (I + Δ)) / Δ> Ω (10)
By adopting the above reward for the learning means 610, the operation method that reaches the operation condition (operation condition x in FIG. 11A) that minimizes the CO concentration under the condition that the change rate of the CO concentration is small is learned. it can.

  This contributes to stable operation when controlling the thermal power plant 100a.

  FIG. 11B is an embodiment (control circuit) of the operation signal generating means 800 provided in the control device 200 in the present embodiment whose control object is the thermal power plant 100a.

  As shown in the control circuit of FIG. 11B, in order to set the oxygen concentration at the boiler outlet to a desired value in the control circuit constituting the operation signal generating means 800, the deviation between the measured oxygen concentration value and the target value is calculated. By adding the input output signal of the PI controller and the air flow rate set value, an operation signal related to the air flow rate is generated.

  Based on the generated air flow rate operation signal, the operation signal generation means 800 determines the total air flow rate to be introduced into the boiler 101 by calculation and outputs it as an operation signal.

  The correction signal 18 generated by the correction signal generation unit 700 is input to the operation signal generation unit 800. The correction signal 18 is input to the control signal constituting the operation signal generation unit 800 shown in FIG. 11B, for example. , And are input to the positions indicated as the correction signal a and the correction signal b.

  That is, the correction signal a and the correction signal b are input as a correction signal a for correcting the oxygen concentration target value and a correction signal b for correcting the air flow rate operation signal.

  FIG. 11C is an example of a method for generating the correction signal 18 in the correction signal generation means 700 provided in the control device 200 in this embodiment.

  As shown in FIG. 11C, the generation of the correction signal 18 by the correction signal generation means 700 is corrected based on the characteristic of the error between the model characteristics of the model 620 and the actual machine data of the thermal power plant 100a that is the control target. A signal 18 is generated.

  That is, the correction signal generation means 700 estimates an error factor from the error characteristics between the model characteristics of the model 620 and the actual machine data of the thermal power plant 100a, and generates the correction signal 18 based on the error factor.

  As an error factor, there may be mentioned an error caused by adjusting the momentum of the air introduced into the furnace 101 of the boiler 101 from the after air port 103 installed on the furnace wall of the boiler 101.

  Thus, adjustment of the momentum of the air that is introduced into the furnace from the after-air port 103 will be described.

  FIG. 11D is a schematic structural diagram of the after-air port 103 installed on the furnace wall of the boiler 101 in this embodiment.

  In FIG. 11D, the air supplied to the after air port 103 is supplied into the furnace through the nozzles 181 and 182 of the after air port 103.

  The distribution of the air supplied from the nozzles 181 and 182 into the furnace can be changed by operating the air dampers 163a and 163b constituting a part of the air damper 163 shown in FIG.

  That is, in FIG. 11D, when the position of the air damper 163b is moved to the right, the flow path of the air damper 163b is narrowed, so that the flow rate of air supplied from the nozzle 182 into the furnace decreases.

  In the thermal power plant 100a, the air dampers 163a and 163b of the after-air port 103 are operated in this way to adjust the flow rate, flow velocity, and momentum of the air that is introduced into the furnace from the after-air port 103.

  The opening degree of the air dampers 163a and 163b of the after-air port 103 is manually operated to a set value.

  Further, there is a possibility that an error occurs between the target air amount and the air amount actually supplied to the boiler 101 between the set value and the air amount resulting from the operation result.

  Therefore, the model 620 of the control device 200 is configured to predict the carbon monoxide concentration in the combustion exhaust gas discharged from the boiler 101 in the case of the set values based on the opening set values of the air dampers 163a and 163b. Has been.

  Therefore, as described above, when an error occurs between the set value of the air amount input to the boiler 101 and the air amount according to the operation result, the predicted value of the model 620 and the state amount of the thermal power plant 100a are used. There is a possibility that a certain measured value does not match.

  Therefore, in the control device 200 of the present embodiment, when the predicted value of the model 620 and the measured value from the thermal power plant 100a do not match, the value between the set value of the amount of air input to the boiler 101 and the amount of air based on the operation result. The operation signal 18 for correcting the opening degree of the air dampers 163a and 163b of the after-air port 103 is generated by the correction signal generation means 700 so as to eliminate the error of the after air port 103, and is input to the operation signal generation means 800 to be input by the operation signal generation means 800. The operation signal 22 is generated.

  Further, the value of the operation signal 18 can be displayed as an operation guidance value on the image display device 950 shown in FIG. 7 so that the operator of the thermal power plant 100a can know.

  As described above, the concentration of nitrogen oxides and carbon monoxide contained in the combustion exhaust gas discharged from the boiler 101 is adjusted by adjusting the opening degree of the air 163a, 163b of the after-air port 103 using the control device 200 of the present embodiment. , Carbon dioxide concentration, sulfide oxide, mercury, or the like can be controlled to a desired value.

  In the present embodiment, the case where the control device 200 adjusts the flow rate, flow velocity, momentum, etc. of the air introduced into the furnace from the after air port 103 installed on the furnace wall of the boiler 101 has been described. May be applied to control for adjusting the flow rate, flow velocity, momentum, and the like of the air introduced into the furnace from the burner 102 installed on the furnace wall of the boiler 101.

  According to the embodiment of the present invention described above, control of a thermal power plant that can maintain good plant control characteristics even when the characteristics of the model for predicting the characteristics of the plant to be controlled are different from the characteristics of the actual plant. A device can be realized.

  The present invention is applicable to a plant control device and a control device for a thermal power plant including a boiler.

The control block diagram which shows the whole structure of the control apparatus of the plant which is the 1st Example of this invention. The flowchart which shows the procedure of control in the control apparatus of the plant which is the 1st Example shown in FIG. The setting screen of the information preserve | saved in the correction signal calculation database in the control apparatus of the plant which is the 1st Example shown in FIG. Explanatory drawing of each operation | movement of the correction signal production | generation means in the control apparatus of the plant which is the 1st Example shown in FIG. Explanatory drawing of the operation | movement which learns the optimal operation method by the correction signal production | generation means in the control apparatus of the plant which is the 1st Example shown in FIG. Explanatory drawing of the interface which changes the range of trial and error by the correction signal production | generation means in the control apparatus of the plant which is the 1st Example shown in FIG. The control block diagram which shows the whole structure of the control apparatus of the thermal power plant which is the 2nd Example of this invention. The schematic block diagram which shows the whole structure of the thermal power plant which is a 2nd Example shown in FIG. Explanatory drawing of the determination method of the plant operating condition by the operation result evaluation means in the control apparatus of the thermal power plant which is a 2nd Example shown in FIG. Explanatory drawing which shows the coal type determination method by the operation result evaluation means 400 in the control apparatus of the thermal power plant which is the 2nd Example shown in FIG. Explanatory drawing of the operation | movement by the model and learning means in the control apparatus of the thermal power plant which is the 2nd Example shown in FIG.

Explanation of symbols

  1, 2, 4: Measurement signal, 7: Model output, 8: Model input, 9, 19: Learning information data, 16: Learning signal, 18: Correction signal, 22, 23: Operation signal, 100: Plant, 100a: Thermal power plant, 101: boiler, 200: control device, 210: external input interface, 220: external output interface, 230: measurement signal database, 240: database for model construction, 250: learning result information database, 260: correction signal calculation Database, 270: operation signal database, 300: measurement signal conversion means, 400: operation result evaluation means, 500: reference signal generation means, 600: learning signal generation means, 610: learning means, 620: model, 700: correction signal Generation means, 800: operation signal generation means, 900: external input device, 910: Mamoru tool, 950: image display device.

Claims (10)

In the plant control device that calculates the operation signal for controlling the plant using the measurement signal obtained by measuring the operation state of the plant,
In the control device, an operation signal generating means for generating an operation signal to be transmitted to the plant, a model for simulating the characteristics of the plant, and an input signal of the model so that the output signal simulated by the model satisfies a preset target And a learning signal generating means for calculating a learning signal according to a learning result in the learning means, and further, a first operation signal of the plant obtained as a result of giving a certain operation signal and an updated operation signal to the plant. An operation result evaluation unit that calculates a first error and a second error, which are errors between the first measurement signal and the second measurement signal, and a target value thereof, and a second error calculated by the operation result evaluation unit Correction signal generation means for generating a correction signal for the operation signal generated by the operation signal generation means when the error is larger than the first error, wherein the correction signal generation means is a model. The operation signal generation unit is configured to calculate a correction signal based on the extracted feature amount of the model characteristic, and the operation signal generation unit includes at least the learning signal calculated by the learning signal generation unit and the correction signal calculated by the correction signal generation unit. A plant control apparatus characterized in that the operation signal for controlling the plant is calculated.   2. The plant control device according to claim 1, wherein in the model, a model characteristic curve when the current operation signal for the plant is changed to an operation signal calculated based on a learning result by the learning unit is calculated, and the correction is performed. In the signal generation means, the number of polarization points of the model characteristic curve is calculated, connected to the control device of the plant, an image display device is installed, and the model characteristic curve in which the position of the polarization point is emphasized in the image display device, or A plant control apparatus characterized in that at least one of the number of polarization points is displayed on a screen. In the plant control device that calculates the operation signal for controlling the plant using the measurement signal obtained by measuring the operation state of the plant,
In the control device, an operation signal generating means for generating an operation signal to be transmitted to the plant, a model for simulating the characteristics of the plant, and an input signal of the model so that the output signal simulated by the model satisfies a preset target And a learning signal generating means for calculating a learning signal according to a learning result in the learning means, and further, a certain operation signal and an updated operation signal obtained as a result of giving the plant to the plant. The operation result evaluation means for calculating the first error and the second error, which are errors between the first measurement signal and the second measurement signal and the target value of the measurement signal, respectively, and the operation result evaluation means Correction signal generation means for generating a correction signal for the operation signal generated by the operation signal generation means when the second error is larger than the first error, the learning means, An operation condition that minimizes an error between the Dell output and the target value is searched and learned within a predetermined threshold range, and the correction signal generating means is within the threshold range of the learning means. Search and change the operation signal to calculate a correction signal that minimizes the error between the measurement signal of the plant and the target value of the measurement signal, and the operation signal generation means is at least the learning signal generation means A plant control apparatus configured to calculate the operation signal for controlling the plant using the calculated learning signal and the correction signal calculated by the correction signal generation means.   The plant control apparatus according to claim 3, further comprising a user interface for arbitrarily setting a search range within a threshold range by the learning means in a learning information database storing parameters used for learning by the learning means. A plant control device. In the plant control device that calculates the operation signal for controlling the plant using the measurement signal obtained by measuring the operation state of the plant,
In the control device, an operation signal generating means for generating an operation signal to be transmitted to the plant, a model for simulating the characteristics of the plant, and an input signal of the model so that the output signal simulated by the model satisfies a preset target And a learning signal generating means for calculating a learning signal according to a learning result in the learning means, and further, a first operation signal of the plant obtained as a result of giving a certain operation signal and an updated operation signal to the plant. An operation result evaluation unit that calculates a first error and a second error, which are errors between the first measurement signal and the second measurement signal, and a target value thereof, and a second error calculated by the operation result evaluation unit Correction signal generation means for generating a correction signal for the operation signal generated by the operation signal generation means when the error is larger than the first error, and the operation result evaluation means includes the error A weight parameter value of the learning signal and the correction signal is calculated on the basis of the learning signal, and the operation signal generation unit is configured to calculate at least the weight parameter value calculated by the operation result evaluation unit and the learning signal calculated by the learning signal generation unit. A plant control apparatus configured to calculate the operation signal for controlling the plant using a signal and a correction signal calculated by the correction signal generation means. In a control device for a thermal power plant that calculates an operation signal for controlling the thermal power plant using a measurement signal obtained by measuring the operating state of the thermal power plant,
Measurement signals include nitrogen oxide concentration, carbon monoxide concentration, carbon dioxide concentration, sulfide oxide concentration, mercury concentration, coal flow rate, mill classifier rotation speed, power generation in exhaust gas discharged from boilers of thermal power plants At least one measurement signal of the machine output is included, and the operation signal includes at least one operation signal of the opening degree of the air damper of the boiler, air flow rate, air temperature, fuel flow rate, exhaust gas recirculation flow rate In the control device, an operation signal generating means for generating an operation signal to be transmitted to the thermal power plant, a model for simulating the characteristics of the thermal power plant, and an output signal simulated by the model satisfy a preset target. A learning means for generating an input signal of the model, and a learning signal generating means for calculating the learning signal according to the learning result of the learning means. A first error and a second error, which are errors between the first measurement signal and the second measurement signal of the thermal power plant acquired as a result of giving the operation signal to the plant and the target value, are calculated. An operation result evaluation unit; and a correction signal generation unit that generates a correction signal of the operation signal generated by the operation signal generation unit when the second error calculated by the operation result evaluation unit is larger than the first error. The model includes a plurality of models corresponding to the burner pattern, load level, and coal type of the boiler of the thermal power plant, and the operation result evaluation means is a value of a measurement signal of the coal flow rate supplied to the boiler mill. The function to grasp the burner pattern based on the output, the function to grasp the load level based on the output demand or the value of the measurement signal of the generator output, and the number of revolutions of the mill classifier A function of grasping the coal type based on the value of the signal is provided, and the learning signal generating means generates a learning signal according to a result learned using a model corresponding to each condition grasped by the operation result evaluating means. The operation signal generation means calculates the operation signal for controlling the thermal power plant using at least the learning signal calculated by the learning signal generation means and the correction signal calculated by the correction signal generation means. The control apparatus of the thermal power plant characterized by comprising as follows.   The thermal power plant control apparatus according to claim 6, wherein the model includes a plurality of models so as to simulate model characteristics when coal of a plurality of coal types is used in a boiler of the thermal power plant. The operation result evaluation means is configured to calculate model weights among a plurality of models based on coal composition information, and the correction signal generation means controls the thermal power plant based on the model weights. A control apparatus for a thermal power plant, wherein the correction signal is calculated.   In the control apparatus for a thermal power plant according to claim 6, the correction signal generating means, when using coal containing more sulfur than the threshold, a burner installed on the furnace wall surface of the boiler, And a control apparatus for a thermal power plant, wherein a correction signal is generated so that an air flow rate supplied from an air port increases. In the control device for a thermal power plant that calculates an operation signal for controlling the thermal power plant using a measurement signal obtained by measuring the operating state of the thermal power plant,
The measurement signal includes at least one of nitrogen oxide concentration, carbon monoxide concentration, carbon dioxide concentration, sulfide oxide concentration, and mercury concentration contained in the exhaust gas discharged from the thermal power plant, and the operation signal includes the boiler. Operation signal generation including at least one operation signal of an air damper opening, air flow rate, air temperature, fuel flow rate, exhaust gas recirculation flow rate, and generating an operation signal to be transmitted to the thermal power plant in the control device Means, a model for simulating the characteristics of a thermal power plant, a learning means for generating an input signal of the model so that an output signal simulated by the model satisfies a preset target, and learning according to a learning result by the learning means Learning signal generation means for calculating a signal, and further, the thermal power generation plan acquired as a result of giving a certain operation signal and an updated operation signal to the plant The operation result evaluation means for calculating the first error and the second error, which are errors between the first measurement signal and the second measurement signal and the target value, respectively, and the second result calculated by the operation result evaluation means Correction signal generation means for generating a correction signal for the operation signal generated by the operation signal generation means when the error is larger than the first error, and the model includes at least the carbon monoxide concentration of the boiler of the thermal power plant A model that reaches the operating condition in which the rate of change in the predicted value of the carbon monoxide concentration is small and the predicted value of the carbon monoxide concentration is minimized. The operation signal generation unit is configured to learn an input signal generation method, and the operation signal generation unit uses at least the learning signal calculated by the learning signal generation unit and the correction signal calculated by the correction signal generation unit. Control device for a thermal power plant which is characterized by being configured to calculate the operation signal for controlling the electric plant. In a control device for a thermal power plant that calculates an operation signal for controlling the thermal power plant using a measurement signal obtained by measuring the operating state of the thermal power plant,
Measurement signals include nitrogen oxide concentration, carbon monoxide concentration, carbon dioxide concentration, sulfide oxide concentration, mercury concentration, coal flow rate, mill classifier rotation speed, generator included in the gas discharged from the thermal power plant At least one of the outputs, and the operation signal includes at least one operation signal of the air damper opening, air flow rate, air temperature, fuel flow rate, exhaust gas recirculation flow rate, and , An operation signal generating means for generating an operation signal to be transmitted to the thermal power plant, a model for simulating the characteristics of the thermal power plant, and an input signal of the model so that the output signal simulated by the model satisfies a preset target And a learning signal generating means for calculating a learning signal according to a learning result in the learning means, and further supplying an operation signal and an updated operation signal to the plant. Operation result evaluation means for calculating a first error and a second error, which are errors between the first measurement signal and the second measurement signal of the thermal power plant acquired as a result of the measurement, and a target value of the measurement signal, respectively. And a correction signal generation unit that generates a correction signal of the operation signal generated by the operation signal generation unit when the second error calculated by the operation result evaluation unit is larger than the first error, and the correction The signal generation means is configured to estimate an error factor from the characteristics of the first error or the second error and generate a correction signal based on the error factor, and the operation signal generation means is at least the learning The fire is characterized in that the operation signal for controlling the thermal power plant is calculated using the learning signal calculated by the signal generation means and the correction signal calculated by the correction signal generation means. Control unit of the power plant.

Images