JP4627509B2 - Plant control apparatus and plant control method - Google Patents

Description

  The present invention relates to a control device for a plant such as a thermal power plant and a method for controlling the plant.

  The plant control device processes a measurement signal obtained from a plant that is a control target, and calculates an operation signal to be given to the control target. The control device is implemented with an algorithm for calculating an operation signal so that the measurement signal of the plant achieves the operation target.

  As a control algorithm used for plant control, there is a PI (proportional / integral) control algorithm. In PI control, a value obtained by multiplying the deviation between the operation target value and the plant measurement signal by a proportional gain is added to a value obtained by time-integrating the deviation to derive an operation signal of a control device that controls the plant. Moreover, the operation signal of the control apparatus which controls a plant may be derived | led-out using a learning algorithm.

  As a method of deriving an operation signal of a control device that controls a plant using a learning algorithm, Japanese Patent Application Laid-Open No. 2000-35956 describes a technique related to an agent learning device.

  Techniques relating to a method using the Dyna-architecture are described in pages 247 to 253 of Reinforcement Learning in the technical literature.

  In the methods based on these technologies, a learning unit that learns in advance a model that predicts the characteristics of a control target in a control device and a model input generation method that achieves a model output target as a model output that is a prediction result of the model. The operation signal to be given to the controlled object is generated according to the learning result of the learning unit.

  If there is an error between the model and the control characteristics of the controlled object, the model is corrected using the measurement signal that is the result of operating the controlled object, and an operation signal is generated for the corrected model. Learn to learn again.

JP 2000-35956 A Reinforcement Learning, Sadayoshi Mikami and Masaaki Minagawa, Morikita Publishing Co., Ltd., published on December 20, 2000

  When learning the method of generating the operation signal for the control device using the methods described in Patent Literature 1 and Non-Patent Literature 1, it is necessary to determine the constraint conditions for learning. For example, when the operating speed of the operation end of the plant to be controlled changes, the range of the operation amount that can be moved by one operation changes, so the learning result also changes. Therefore, in order to obtain a learning result, it is necessary to appropriately set a learning constraint condition using information on the operation speed of the operation end.

  However, it is difficult to set such learning constraint conditions in advance. In plant control, a plant is operated using a plurality of operation ends of a control device, and even if the operation ends have the same design specifications, the actual operation speed often varies. In addition, there is a possibility that the operating speed is lowered due to deterioration of these operation ends over time.

  When a variation in operation speed or a decrease in operation speed occurs at the operation end, a desired control result cannot be obtained even if an operation signal generated according to the learned model input generation method is given to the plant to be controlled.

  An object of the present invention is to control a plant satisfactorily even when there are variations in the operation speeds of a plurality of operation ends used for plant control, or even when the operation ends deteriorate over time. An object of the present invention is to provide a plant control apparatus and a plant control method having a function of appropriately determining learning constraint conditions.

The plant control apparatus of the present invention is a plant control apparatus including an operation signal generation unit that calculates an operation signal that is a control command to be given to the plant using a measurement signal that is an operation state quantity of the plant. A model that simulates the control characteristics of the plant to be controlled, a control logic database that stores control logic data including control parameters used to calculate operation signals in the operation signal generator, and a state quantity of the plant is controlled An operation end specification database storing operation end specification data of an operation end, an operation signal database storing past operation signals, a measurement signal database storing past measurement signals, and a control logic database; Using control logic data and operation end specification data stored in the operation end specification database, A function that determines the initial value of the learning parameter including the operation limit speed, upper limit value, and lower limit value of the operation end, which is the limit value of the operation signal change width per time, and is stored in the operation signal database and the measurement signal database. Included in the learning parameter that is control logic data when the amount of change in measurement signal data per unit time, which is the sample control period, is smaller than the amount of change in operation signal data Learning condition determination unit having a function of updating the operation limit speed of the operation end to be changed to the value of the change amount of the measurement signal data, and learning the limit value of the operation signal change width per unit time included in the learning parameter How to generate a model input in which the model output simulated by the model achieves the target value of the model output And a learning information database in which learning information data, which is a result of learning a model input generation method by the learning unit, is stored, and the operation signal generation unit is an operation state quantity of the plant. A learning signal generation unit that calculates an operation signal for the plant using the measurement signal and learning information data stored in a learning information database is provided.
The plant control device of the present invention is a plant control device that controls the thermal power plant by calculating an operation signal as a control command to be given to the thermal power plant using a measurement signal that is an operation state quantity of the thermal power plant. The control device simulates the control characteristics of a thermal power plant to be controlled, and an operation signal generator that calculates an operation signal that is a control command given to the plant using a measurement signal that is an operation state quantity of the thermal power plant. A control logic database in which control logic data including a control parameter used to calculate an operation signal in the operation signal generation unit is stored, and operation end specification data of an operation end that controls the state quantity of the thermal power plant A stored operation end specification database, an operation signal database storing past operation signals, and a past Operation that is the limit value of operation signal change width per unit time using measurement signal database in which measurement signals are stored, control logic data stored in control logic database and operation end specification database, and operation end use data Sample control using the function to determine the initial value of the learning parameter including the upper limit speed, upper limit value, and lower limit value, and the operation signal data and measurement signal data stored in the operation signal database and measurement signal database When the change amount of the measurement signal data per unit time, which is a cycle, is smaller than the change amount of the operation signal data, the change in the measurement signal data is determined as the operation limit speed of the operation end included in the learning parameter that is the control logic data. A learning condition determination unit that has the function of updating to a quantity value, and the unit time included in the learning parameter Learning how to operate a thermal power plant using the model to set the limit value of the change width of the operation signal per hit as a learning constraint, so that the model output simulated by the model achieves the target value of the model output And a learning information database in which learning information data, which is a result of learning a model input generation method by the learning unit, is stored, and the operation signal generation unit measures a plant operating state quantity. A learning signal generation unit that calculates an operation signal for the thermal power plant using the signal and learning information data stored in the learning information database is provided.

Further, the plant control method of the present invention is a plant control method for controlling a plant by calculating an operation signal serving as a control command to be given to the plant using a measurement signal that is an operation state quantity of the plant. The control model of the plant to be controlled is simulated by the equipped model, and the control logic data including the control parameters used to calculate the operation signal is stored in the control logic database of the control device by the operation signal generator provided in the control device. The operation end specification data of the operation end for controlling the state quantity of the plant is stored in the operation end specification database provided in the control device, the past operation signal is stored in the operation signal database provided in the control device, and the past measurement signal is stored. Are stored in the measurement signal database provided in the control device and controlled by the learning condition determination unit provided in the control device. Operation limit speed, upper limit value, and lower limit value that are the limit value of the operation signal change width per unit time using control logic data and operation end usage data stored in the control database and operation end specification database The amount of change in measurement signal data per unit time, which is the sample control period, is determined using the operation signal data and measurement signal data stored in the operation signal database and measurement signal database . Is smaller than the change amount of the operation signal data, the operation limit speed value of the operation end included in the learning parameter which is the control logic data is updated to the change amount value of the measurement signal data, and the control device Learn the limit value of the operation signal change width per unit time included in the learning parameter by the learning unit The learning unit simulates the characteristics of the plant and learns how to generate the model input so that the model output that is set as a constraint condition and simulated by the model using the model achieves the target value of the model output. The learning information data, which is the result of learning the generation method of the model input in the unit, is stored in the learning information database, and the measurement signal and the learning information database that are the operation state amount of the plant by the learning signal generation unit provided in the operation signal generation unit The operation information serving as a control command to be given to the plant is calculated using the learning information data stored in the plant to control the plant.
The plant control method of the present invention is a plant control method for controlling a thermal power plant by calculating an operation signal as a control command to be given to the thermal power plant using a measurement signal which is an operation state quantity of the thermal power plant. The control device simulates the control characteristics of the plant to be controlled by the model provided in the control device of the plant, and the control logic data including the control parameters used for calculating the operation signal by the operation signal generation unit provided in the control device is supplied to the control device. Save to the control logic database provided, save the operation end specification data of the operation end that controls the state quantity of the plant to the operation end specification database, save the past operation signal to the operation signal database provided in the control device, Is stored in a measurement signal database provided in the control device, and a learning condition determination unit provided in the control device Therefore, the learning parameters including the operation limit speed, the upper limit value, and the lower limit value of the operation end which are the limit values of the operation signal change width per unit time using the data stored in the control logic database and the operation end specification database. And the change amount of the measurement signal data per unit time, which is a sample control cycle, is determined by using the operation signal data and the operation signal data stored in the measurement signal database and the measurement signal data. When the change amount of the signal data is smaller , the value of the operation limit speed of the operation end included in the learning parameter that is the control logic data is updated to the value of the change amount of the measurement signal data. Learning the limit value of the operation signal change width per unit time included in the learning parameter by the learning unit The learning unit simulates the characteristics of the plant and learns how to generate the model input so that the model output that is set as a constraint condition and simulated by the model using the model achieves the target value of the model output. The learning information data, which is the result of learning the generation method of the model input in the unit, is stored in the learning information database, and the measurement signal and the learning information database that are the operation state amount of the plant by the learning signal generation unit provided in the operation signal generation unit The operation information serving as a control command to be given to the plant is calculated using the learning information data stored in the plant to control the plant.

  According to the present invention, it is possible to control a plant satisfactorily even when the operation speeds of a plurality of operation ends used for plant control vary or even when the operation ends deteriorate due to aging. In addition, it is possible to realize a plant control apparatus and a plant control method having a function of appropriately determining learning constraint conditions.

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

FIG. 1 is a control system diagram showing a plant control apparatus according to an embodiment of the present invention.

  In FIG. 1, the plant 100 is configured to be controlled by a control device 200.

  The control device 200 that controls the plant 100 to be controlled includes an operation signal generation unit 300, a learning unit 400, a model 500, an evaluation value calculation unit 600, a learning condition determination unit 700, and a learning information addition unit 800 as arithmetic devices. Are provided.

  Further, the control device 200 includes, as a database, a measurement signal database 210, an operation end specification database 220, an operation signal database 230, a control logic database 240, a learning parameter database 250, an evaluation value calculation parameter database 260, a model parameter database 270, And a learning information database 280 are provided.

  In addition, the control device 200 is provided with an external input interface 201 and an external output interface 202 as interfaces with the outside.

  In the control device 200, the measurement signal 1 that is a control output of the plant 100 is taken into the control device 200 from the plant 100 via the external input interface 201. Further, an operation signal 24 serving as a control command is transmitted from the control device 200 to the controlled object 100 via the external output interface 202.

  Next, the details of the control in the control apparatus 200 will be described. The measurement signal 2 taken into the external input interface 201 as the measurement signal 1 of the plant 100 is transmitted to the operation signal generation unit 300 and also stored in the measurement signal database 210. The Further, the operation signal 23 generated by the operation signal generation unit 300 is transmitted to the external output interface 202 and is stored in the operation signal database 230.

  In the operation signal generation unit 300, the measurement signal 1 of the plant 100 uses the control logic data 11 stored in the control logic database 240 and the learning information data 22 stored in the learning information database 280 to determine the operation target value. An operation signal 23 is generated to achieve this.

  The control logic database 240 stores control circuits and control parameters for calculating the control logic data 11 in order to output the control logic data 11 to the operation signal generation unit 300.

  The learning information data stored in the learning information database 280 is generated by the learning unit 400 or the learning information adding unit 800. The learning unit 400 is connected to the model 500, the evaluation value calculation unit 600, and the learning condition determination unit 700, respectively.

The model 500 has a function of simulating the control characteristics of the plant 100. That is, the operation signal 24 serving as a control command is given to the plant 100 and the same operation as that for obtaining the measurement signal 1 as a result of the control is simulated.
For this simulation calculation, the model input 17 for operating the model 500 is received from the learning unit 400, the control operation of the plant 100 is simulated by the model 500, and the model output 18 of the simulation calculation result is obtained. Has been. Here, the model output 18 is a predicted value of the measurement signal 1 of the plant 100.

  The model 500 has a model for simulating the control characteristics of the plant 100. A physical model using a model formula based on a physical law, a statistical model using a statistical method such as a neural network, or a physical model and It has a function of calculating the model output 18 for the model input 17 by using the statistical model together.

  In the model 500, other data necessary for calculating the model output 18 by simulating the control of the plant 100 based on the model input 17 is input to the model 500 as data stored in the model parameter database 270. To use.

  The evaluation value calculation unit 600 calculates the evaluation value 19 using the evaluation value calculation parameter 15 stored in the evaluation value calculation parameter database 260 and the model output 18 input from the model 500.

  The learning unit 400 generates a model input 17 to be input to the model 500 using the learning information data 21 stored in the learning information database 280 and the learning parameter 14 stored in the learning parameter database 250.

  In the model 500, the model input 17 is inputted, and the model output 18 simulated by using the internal simulation model is outputted.

  The evaluation value calculation unit 600 calculates an evaluation value 19 from the model output 18 simulated by the model 500 and inputs the evaluation value 19 to the learning unit 400.

  The learning unit 400 simulates the model 500 in order to learn the plant operation method using the model by setting the limit value of the operation signal change width per unit time included in the learning parameter as a learning constraint condition. A model input generation method in which the calculated model output 18 achieves the model output target value is learned using the model output 18 or the evaluation value 19. The learning information data 20 that is a learning result is stored in the learning information database 280.

  In the learning condition determination unit 700, the operation end specification data 4 of the operable range and the operation speed of the operation end of the plant stored in the operation end specification database 220 and the control logic data 6 stored in the control logic database 240 are stored. The initial value of the learning parameter 8 including the limit value of the operation signal change width per unit time is generated.

  In the learning condition determination unit 700, measurement signal data 3 that is a past measurement signal stored in the measurement signal database 210, operation signal data 5 that is a past operation signal stored in the operation signal database 230, and The learning parameter 8 is updated using the learning parameter 9 stored in the learning parameter database 250.

  When the values of the learning parameter 9 and the learning parameter 8 are different, the learning trigger 7 is set to “1”, and this value is transmitted to the learning unit 400 and the learning information adding unit 800. In other cases, the learning trigger 7 has a value of “0”.

  The learning information adding unit 800 uses the learning parameter 10 stored in the learning parameter database 250 and the learning information data 12 stored in the learning information database 280 when the learning trigger 7 becomes “1”. Additional learning information data 13 is generated. The additional learning information data 13 is stored in the learning information database 280.

  An operator of the plant 100 uses an external input device 900 including a keyboard 901 and a mouse 902, a maintenance tool 910 including a data transmission / reception processing unit 930 capable of transmitting / receiving data to / from the control device 200, and an image display device 950. Information stored in various databases included in the control device 200 can be accessed.

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

  The maintenance tool input signal 31 generated by the input device 900 is taken into the maintenance tool 910 via the external input interface 920. The data transmission / reception processing unit 930 of the maintenance tool 910 acquires the database information 30 provided in the control device 200 according to the information of the maintenance tool input signal 32.

  The data transmission / reception processing unit 930 transmits a maintenance tool output signal 33 obtained as a result of processing the database information 30 to the external output interface 940. The maintenance tool output signal 34 is displayed on the image display device 950.

  In the control device 200 of the embodiment of the present invention described above, the measurement signal database 210, the operation end specification database 220, the operation signal database 230, the control logic database 240, the learning parameter database 250, the evaluation value calculation parameter database 260, the model parameters. Although the database 270 and the learning information database 280 are arranged inside the control device 200, all or part of them can be arranged outside the control device 200.

  Similarly, a learning unit 400, a model 500, an evaluation value calculation unit 600, a learning condition determination unit 700, and a learning information addition unit 800 are arranged inside the control device 200, but all or a part of them are controlled. It can also be arranged outside the device 200.

  For example, the learning unit 400, the model 500, the evaluation value calculation unit 600, the learning parameter database 250, the evaluation value calculation parameter database 260, and the model parameter database 270 are configured as external systems, and the external system and the control device 200 are configured. The learning information data 20 generated by the learning unit 400 of the external system may be transmitted to the control device 200 via the Internet by connecting via the Internet.

  Further, if the control device 200 is constructed without using one or both of the evaluation value calculation unit 600 and the learning information addition unit 800, the plant can be controlled although the advanced control function is reduced.

  In addition, a function of correcting the model parameter 16 stored in the model parameter database 270 may be added so that the characteristics of the plant 100 and the model 500 match.

  Below, the case where the control apparatus 200 with respect to the plant which is an Example of this invention is applied to the thermal power plant 100a is demonstrated. Needless to say, the control device 200 of the embodiment of the present invention can also be used when controlling a plant other than the thermal power plant.

  FIG. 2 is a diagram showing a schematic system of a plant when the thermal power plant 100a is a control target plant. First, a power generation mechanism in the thermal power plant 100a will be described.

  A boiler 101 constituting the thermal power plant 100a has a burner 102 for supplying pulverized coal, which is fuel obtained by finely pulverizing coal in a mill 110, primary air for conveying pulverized coal, and secondary air for combustion adjustment. The pulverized coal supplied via the burner 102 is combusted inside the boiler 101. The pulverized 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, the boiler 101 is provided with an after air port 103 for introducing after-air for two-stage combustion into the boiler 101, and the after air is led from the pipe 142 to the after-air port 103.

  The high-temperature combustion gas generated by the combustion of the pulverized coal flows downstream along the path inside the boiler 101 and then passes through the heat exchanger 106 disposed in the boiler 101 to exchange heat. High-temperature and high-pressure steam is generated by the heater 104. After that, after exhaust gas treatment, it is emitted from the chimney to the atmosphere.

  The feed water circulating through the heat exchanger 106 of the boiler 101 is supplied with the feed water to the heat exchanger 106 via the feed water pump 105, and is superheated by the combustion gas flowing down the boiler 101 in the heat exchanger 106. Become. In this embodiment, the number of heat exchangers 106 is one, but a plurality of heat exchangers 106 may be arranged.

  The high-temperature and high-pressure steam that has passed through the heat exchanger 106 is guided to the steam turbine 108 through the turbine governor 107, and the steam turbine 108 is driven by the energy of the steam to generate power by the generator 109.

  Various measuring devices for detecting the operating state of the thermal power plant are arranged in the thermal power plant 100a, and information regarding the control output of the plant acquired from these measuring devices is transmitted to the control device 200 as measurement information 1. Is done. For example, FIG. 2 illustrates a flow rate measuring device 150, a temperature measuring device 151, a pressure measuring device 152, a power generation output measuring device 153, and a concentration measuring device 154 for detecting information related to plant control output. .

  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. The temperature measuring device 151 and the pressure measuring device 152 measure the temperature and pressure of the steam supplied from the heat exchanger 106 to the steam turbine 108.

  The amount of power generated by the power generator 109 is measured by a power generation output measuring device 153. Information on the concentration of components (CO, NOx, etc.) contained in the combustion gas passing through the boiler 101 can be measured by a concentration measuring device 154 provided on the downstream side of the boiler 101.

  In general, a number of measuring instruments other than those shown in FIG. 2 are arranged in the thermal power plant, but the illustration is omitted here.

  Next, the paths of primary air and secondary air that are input from the burner 102 into the boiler 101 and the path of after-air that is input from the after-air port 103 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 air heater 104 installed on the downstream side of the boiler 101 and a pipe 131 that bypasses the air heater 104 without passing through, and is again pipe 133. And is guided to a mill 110 installed on the upstream side of the burner 102.

  The air passing through the air heater 104 is heated by the combustion gas flowing down the boiler 101. Using this primary air, the pulverized coal pulverized in the mill 110 is conveyed to the burner 102 together with the primary air.

  The secondary air and the after air are led from the fan 121 to the pipe 140 and heated in the same manner by the air heater 104, and then branched into the secondary air pipe 141 and the after air pipe 142, respectively. And the after-air port 103.

  FIG. 3 is an enlarged view showing the piping part of the piping 130, 131, 132, 133, 140, 141, 142 through which the primary air, the secondary air, and the after air shown in FIG. is there.

  As shown in FIG. 3, among these pipes, air dampers 160, 161, 162, and 163 are arranged on the pipes 131, 132, 141, and 142, respectively. By operating these air dampers 160, 161, 162, and 163, the area through which the air passes through each of the pipes 131, 132, 141, and 142 can be changed, so that the pipes 131, 132, 141, and 142 are passed through. The air flow to be adjusted can be adjusted individually.

  A water supply pump 105, a mill 110, air dampers 160, 161, 162, 163, and the like that constitute an operation end for controlling the state quantity of the thermal power plant 100a to be controlled using various operation signals 24 generated by the control device 200, etc. Each of the devices. In the present embodiment, devices such as the water supply pump 105, the mill 110, and the air dampers 160, 161, 162, and 163 are referred to as operation ends, and a command signal necessary for operating them is referred to as an operation signal 24.

  In addition, when fuel such as air for combustion or fuel such as pulverized coal is introduced into the boiler 101, a function capable of moving the discharge angle up and down is added to the burner 102 and the after-air port 103, and these angles are set. It can also be included in the operation signal 24.

  FIG. 4 is a detailed diagram illustrating signal processing in the operation signal generation unit 300 of the control device 200. In FIG. 4, in the operation signal generation unit 300, the measurement signal 2 collected from the measurement signal 1 of the plant 100 via the external input interface 201, the learning information data 22 stored in the learning information database 280, and the control logic database 240. The control logic data 11 stored in the operation signal 24 is respectively input, and an operation signal 24 that is a control command for the plant 100 calculated by the operation signal generation unit 300 with reference to these signals and data is input via the external input interface 202. An operation signal 23 to be output is generated.

  The operation signal generation unit 300 includes a learning signal generation unit 310, an operation target value 320, adders / subtractors 330, 331 and 332, a proportional integration controller 340, change rate limiters 350 and 351, high value selectors 360 and 361, and a low value. Selectors 370 and 371 are arranged, and each of these devices is connected in the manner shown in FIG.

  The control parameters necessary for operating each device of the operation signal generation unit 300 are input and used from those stored in the control logic database 240 and the learning information database 280. Note that the configuration of the operation signal generation unit 300 may be other than the device configuration shown in FIG.

  In the adders / subtracters 330, 331, and 332, the signal value is added to the zero value or the subtraction operation is performed using the two input signals. In FIG. 4, a signal to be added is represented by “+”, and a signal to be subtracted is represented by “−”.

  The adder / subtractor 330 calculates the signal 381 using the measurement signal 2 and the operation target value signal 380 taken into the operation signal generation unit 300 based on the function of the expression (1) incorporated in the adder / subtractor 330.

Here, χ ₁ is the value of the signal 381, χ ₂ is the operation target value signal 380, and χ ₃ is the value of the measurement signal 2.

  Next, the proportional integration controller 340 uses the previous value of the signal 381 and the signal 381 and the previous value of the reference signal 382 and the reference signal 382 based on the function of the equation (2) incorporated in the proportional integration controller 340. Calculate The previous value means a value before one sample control period.

Here, P ₁ and P ₂ are control parameters, χ ₄ is the value of the reference signal 382, χ ₅ is the signal 381, χ ₆ is the previous value of the signal 381, and χ ₇ is the previous value of the reference signal 382.

  In addition, the learning signal generation unit 310 derives the recommendation signal 383 using the measurement signal 2 while referring to the learning information data 22 stored in the learning information database 280. This recommendation signal 383 is a recommended value of the operation signal 23.

  The learning information data 22 stored in the learning information database 280 is data necessary for constructing a function for generating the model input 17 from the evaluation value 19 in the learning unit 400. The learning signal generation unit 310 generates the recommendation signal 383 from the measurement signal 2 in the same manner as the learning unit 400 generates the model input 17 from the evaluation value 19.

  The adder / subtractor 331 calculates the signal 384 using the reference signal 382 and the recommendation signal 383 based on the function of the expression (3) incorporated in the adder / subtracter 331.

Here, χ ₈ is the value of the signal 384, χ ₉ is the recommendation signal 383, and χ ₁₀ is the value of the reference signal 382.

  The change rate limiter 350 limits the value of the signal 384 that changes per sample control period. The change rate limiter 350 calculates the signal 385 based on the function of the equation (4) incorporated in the change rate limiter 350.

Here, P ₃ and P ₄ are control parameters, χ ₁₁ is the signal 385, χ ₁₂ is the previous value of the signal 384, and χ ₁₃ is the value of the signal 384. P ₃ and P ₄ are called an increase rate parameter and a decrease rate parameter, respectively.

  By using the change rate limiter 350, the value of the signal 385 can be limited so that the value of the operation signal 384 that changes per sample control period is within the range of the value of the increase rate parameter and the value of the decrease rate parameter.

  The high value selector 360 has a function of preventing the signal 386 from becoming a value below a certain threshold value. The high value selector 360 calculates the signal 386 based on the function of the equation (5) incorporated in the high value selector 360.

Here, P ₅ is a control parameter, χ ₁₄ is the value of signal 386, and χ ₁₅ is the value of signal 385. P ₅ is called a decrease parameter. By using the high value selector 360, the value of the signal 386 can be prevented from becoming less than or equal to the value of P _5.

  The low value selector 370 has a function of preventing the correction signal 387 from becoming a value greater than a certain threshold value. The low value selector 370 calculates the correction signal 387 based on the function of the equation (6) incorporated in the low value selector 370.

Here, P ₆ is a control parameter, χ ₁₆ is the value of the correction signal 387, and χ ₁₇ is the value of the signal 386. P ₆ is called an upper limit parameter. By using the low value selector 370, the value of the signal 387 can be prevented from becoming greater than or equal to the value of P _6.

  In FIG. 4, a plurality of change rate limiters (RL), high value selectors (HL), and low value selectors (LL) are used, but the operation content is the same as the functions of equations (4) to (6). It is. The control parameters of the change rate limiters 350 and 351, the high value selectors 360 and 361, and the low value selectors 370 and 371 can be set individually.

  These control parameters are set by an operator of the plant 100 using the external input device 900, the maintenance tool 910, and the image display device 950.

  Using the reference signal 382 and the correction signal 387 calculated by the above devices, the adder / subtractor 332 adds the two signals to calculate the signal 388. The signal 389 is calculated from the signal 388 using the change rate limiter 351, the signal 390 is calculated from the signal 389 using the high value selector 361, and finally the operation signal 23 is calculated from the signal 390 using the low value selector 371. The operation signal 23 is calculated and output from the control device 200 as the command signal 24 to the plant 100 from the external interface 202.

  By configuring the operation signal generation unit 300 of the control device 200 as shown in FIG. 4, the following effects can be obtained.

  First, the operation signal generator 300 is provided with a change rate limiter 351, a high value selector 361, and a low value selector 362, so that the operation signal 23 is limited within a preset allowable range, and further a preset value. It can suppress that it changes rapidly more than the above.

  Accordingly, it is possible to prevent the operation signal 23 deviating from the operation speed and operation range of the operation end from being calculated and output as the command signal 24.

  In addition, depending on the operation status of the plant 100, if the operation signal 23 serving as the command signal 24 is greatly changed, the safe operation of the plant 100 may be hindered. Even in such a case, the plant 100 can be safely operated by appropriately setting the control parameter of the change rate limiter 351.

  4 does not directly calculate the operation signal 23 by using the recommended signal 383 calculated by the learning signal generation unit 310, but the adder / subtracter 331 generates the reference signal from the recommended signal 383. 382 is subtracted, and after applying the change rate limiter 350, the high value selector 360, and the low value selector 370, the reference signal 382 is added again.

  In the learning signal generation unit 310, the recommended signal 383 is generated with reference to the learning information database 280 in which the result of learning using the model 500 is stored. Therefore, if the characteristics of the model 500 and the plant 100 are different. Even if the recommendation signal 383 is given to the plant 100 as the command signal 24, there is a possibility that the desired performance cannot be obtained.

  Further, by giving the recommendation signal 383 to the plant 100 as the command signal 24, there is a possibility that the plant 100 cannot be operated safely.

  In order to avoid such a situation, the operation signal generation unit 300 uses the change rate limiter 350, the high value selector 360, and the low value selector 370, and appropriately sets this control parameter, whereby the learning signal generation unit The degree of contribution of the recommended signal 383 generated by 310 to the operation signal 23 can be adjusted.

  For example, at the beginning of the introduction of the learning signal generation unit 310, there is no information regarding the difference in characteristics between the model 500 and the plant 100, so the control parameters are set so that the influence of the recommendation signal 383 on the operation signal 23 is reduced. After confirming that the characteristics match, it is possible to implement a countermeasure such as resetting the control parameter so that the influence of the recommendation signal 383 on the operation signal 23 is increased.

  In the thermal power plant 100a, the power generation output constant operation that keeps the power generation output constant, the power generation output change operation that changes the power generation output, the burner switching operation that switches the ignition of the burner of the boiler 101, and the coal type switching that switches the type of coal used as fuel There are various driving modes such as driving. Moreover, even if it is a power generation output fixed operation, the charcoal used as fuel may be different.

  In the control device 200 of the thermal power plant 100a which is an embodiment of the present invention, the control parameter can be determined for each of such various operation modes, so that a command signal that matches the plant operation mode can be generated.

  FIG. 5 shows an example of a control parameter setting screen by the plant control apparatus 200 according to the embodiment of the present invention. FIG. 5 shows a screen for setting control parameters in the change rate limiter 350 included in the operation signal generation unit 300 provided in the control device 200 of the thermal power plant 100a.

  As shown in FIG. 5, the change rate limiter 350 included in the operation signal generation unit 300 represents a situation in which each parameter for increasing rate and each parameter for decreasing rate are set for each operation mode of the thermal power plant 100 a. .

  Next, a learning condition determination unit 700 that determines learning parameters stored in the learning parameter database 250 provided in the control device 200 shown in FIG. 1 will be described. The learning condition determination unit 700 determines a learning parameter 14 that is referred to when the learning unit 400 performs learning.

  When the learning unit 400 performs learning, a change range of the model input 17 that can be moved per sampling control period, an upper limit value of the model input 17, and a lower limit value of the model input 17 are required.

  In the learning condition determination unit 700 of the control device 200, the control logic data 6 stored in the control logic database 240, the operation end specification data 4 stored in the operation end specification database 220, and the measurement signal database 210 are stored. The learning parameter 8 to be stored in the learning parameter database 250 is determined with reference to the measured signal data 3.

  Since the measurement signal cannot be obtained before the plant 100 is operated, the learning condition determination unit 700 determines the initial value of the learning parameter 8 from the control logic data 6 and the operation end specification data 4, operates the plant 100, After obtaining the measurement signal, the learning parameter 8 is updated using the measurement signal data 3 as well.

  FIG. 6 is a diagram illustrating a method for determining the initial value of the learning parameter 8 in the learning condition determination unit 700 provided in the plant control apparatus 200 according to the embodiment of the present invention.

  In FIG. 6, the data regarding the rate-of-change restriction | limiting, an upper limit, and a lower limit are described for every operation end. The value of the control logic data 6 is reflected and displayed in the RL, LL, and HL fields, and the value of the operation end specification data 4 is reflected and displayed in the specifications field. The value of the control logic data 6 is, for example, a control parameter set by the change rate limiter 350 included in the operation signal generation unit 300 illustrated in FIG. The values of the operating end specification data 4 are, for example, the operation limit speed, the upper limit value, and the lower limit value of the operating end, and these values are set by the operator of the plant 100.

  The learning condition determination unit 700 selects a value having the minimum degree of freedom when generating the model input 17 from the values described in FIG. 6, and uses this value as an initial value of the learning parameter 8 as a learning parameter database. 250. For example, the increase rate and the decrease rate of the change rate limiting parameter can increase the degree of change of the model input that can be moved in one sample control period as the absolute value increases, and the degree of freedom also increases.

  Conversely, when the absolute value of the change rate limiting parameter is small, the degree of freedom is also small. Therefore, the increase rate and the decrease rate of the change rate limiting parameter are transmitted to the learning parameter database 250 with a value having a small absolute value as an initial value of the learning parameter 8.

  Further, by selecting the minimum value for the upper limit value and the maximum value for the lower limit value, the degree of freedom in generating the model input 17 can be minimized.

  In this embodiment, a value that minimizes the degree of freedom in generating the model input 17 is selected and the initial value of the learning parameter 8 is determined. However, the operation end specification data stored in the operation end specification database 220 is selected. Various selection methods can be set, such as determining the value of 4 as the initial value of the learning parameter 8 as it is.

  The learning condition determination unit 700 has a function of estimating the current operation mode of the plant 100 by processing the signal included in the control logic data 6 or the measurement signal data 3. By using this function, it is possible to determine which value is currently used among the control parameters set for each operation mode of the plant.

  Next, a method for updating the learning parameter 8 will be described. First, when the operation mode of the plant 100 is changed and the value of the control logic data 6 is changed, the learning parameter 8 is determined by using the changed value of the control logic data 6 by using the method described in FIG. To do.

  Further, the learning condition determination unit 700 updates the learning parameter 8 using the measurement signal data 3 and the operation signal data 5. A method for updating the learning parameter 8 in the learning condition determination unit 700 will be described with reference to FIG.

FIG. 7 shows an example of a method for updating the learning parameter 8 in the learning condition determination unit 700. FIG. 7 shows the operation signal data 3 and the measurement signal data 5 related to the operation end A at times t ₁ and t ₂ . . Δt is 1 sample control cycle time, C ₁ is the value of the operation signal A at time t _1, C ₂ is the value of the operation signal data 3 at time t _2, C ₃ is the measured signal data 5 at time t ₂ Value.

In FIG. 7, the operation signal data 3 as the operation signal A changes by the difference signal of C ₂ −C ₁ during the time Δt from the time t ₁ to the time t ₂ , whereas the measurement signal The data 5 changes only for the difference signal of C ₃ -C ₁ , and the change width of the measurement signal data is smaller than the change width of the operation signal data.

This is an event that occurs when the change width of the operation signal of the operation end A is larger than the operation limit speed per one sample control period. In such a case, the value of the learning parameter 8 regarding the increase rate of the operation signal A is set to the value of the difference signal of C ₃ -C ₁ .

  The learning parameter 8 is determined by the learning condition determination unit 700 by the above method, and the learning parameter 8 is stored in the learning parameter database 250. The learning parameter 8 is also updated when the operation mode changes and the control parameter changes.

  Next, an example in which the learning unit 400 of the control device 200 determines the model input 17 for the model 500 and reduces nitrogen oxide (NOx) that is one of the model outputs 18 output from the model 500 will be described. To do.

  In addition to the nitrogen oxide, the model output 18 includes the case where carbon monoxide (CO), carbon dioxide concentration, sulfide oxide, mercury, vapor temperature, vapor pressure, and the like are controlled to desired values. It is controllable by using the plant control apparatus of the embodiment.

  FIG. 8 illustrates the relationship between the model input 17 input to the model 500 and the model output 18 output from the model 500. In FIG. 8, two types of model input A and model input B are used as model input 17, and NOx is used as model output 18.

As shown in FIG. 8, if the model input A is A ₁ and the model input B is B ₁ , the NOx of the model output 18 is NOx high, and if the model input A is A ₂ and the model input B is B ₂ , the model output 18 NOx becomes NOx low. As described above, the learning unit 400 can learn a method for reaching the NOx low region from the initial state as shown in FIG.

  FIG. 9 illustrates an example of the result of learning the model input generation method for the model by the learning unit 400. In FIG. 9, the NOx low region is reached with the smallest possible number of operations, and The result of learning under the condition that state transition does not occur in the NOx high region is shown.

  The reason why the NOx low region is not reached directly by one operation is that the values of the model input A and the model input B that can be moved per one sample control period are limited.

  The value of the model input 17 that can be moved per sample is based on the learning parameter 8 (learning parameter 14) such as the increase rate and decrease rate of the operation end described in FIG. It is determined to correspond.

  As shown in FIG. 9, the learning unit 400 reaches the NOx low region by two operations so as to indicate that the NOx low region has been reached after the second operation after the first operation. Learned how to do.

  FIG. 10 illustrates the relationship between the operation signal A and the operation signal B, which is an example of the result of learning the operation signal generation method by the learning unit 400 as in FIG. The model input B and the operation signal B correspond to each other.

  The operation method indicated by the dotted arrow in FIG. 10 represents the result learned by the learning unit 400 of the control device 200. In FIG. 10, when the operation speed of the operation signal A is low, the state transitions to a region where the NOx is high after one operation.

  This means that when the operation limit speeds of the operation signal 24 and the model input 17 are different, the learning unit 400 reaches the NOx low region with as few operations as possible and does not transition to the NOx high region. If the operation signal 24 is generated according to the learning result of the generation method of the model input 17 under the condition and given to the plant, this means that the condition set at the time of learning may not be satisfied. .

  In the embodiment of the present invention, in order to avoid such a situation, the following measures are taken. That is, in the present embodiment, the learning condition determining unit 700 is provided in the control device 200, the learning parameter 8 including the operation limit speed at the operation end of the plant 100 is determined as described above, and the learning parameter 8 is determined as the learning parameter. Save in the database 250. The learning unit 400 refers to the learning parameter 14 stored in the learning parameter database 250 to perform learning on the assumption that the operation signal 24 and the operation limit speed of the model input 17 match.

  Next, the control operation of the control device 200 will be described using the flowchart shown in FIG.

  FIG. 11 is a flowchart showing a calculation process for the simulation and learning contents of the plant model in the plant control apparatus 200 in the embodiment of the present invention shown in FIG.

  The flowchart of the control operation of the control device 200 shown in FIG. 11 can be applied even when the learning information adding unit 800 shown in FIG. 1 is not provided. The operation content of the learning unit information adding unit 800 and a flowchart in the case where the learning unit information adding unit 800 is provided will be described later.

  As illustrated in FIG. 11, the flowchart of the control operation of the control device 200 is executed by combining steps 1010, 1020, 1030, 1040, 1050, and 1060. Each step will be described below.

  First, in step 1010, the learning unit 400 and the model 500 are operated to learn a method for generating the model input 17 so that the model output 18 achieves the model output target value.

  The evaluation value calculation unit 600 uses the evaluation value calculation parameter data 15 to determine whether the model output 18 has achieved the model output target value, or the model output 18 and the model output target value are close to each other. The learning may be performed using an evaluation value 19 that is a value evaluated quantitatively.

  The evaluation value calculation parameter database 260 stores parameter values necessary for calculating the evaluation value 19 such as a model output target value. Optimization methods such as genetic algorithms, dynamic programming, and reinforcement learning can be applied to learning.

  Next, in step 1020, the learning unit 400 is operated, and the result learned in step 1010 is transmitted from the learning unit 400 to the learning information database 280 as learning information data 20. The learning information data 20 is information relating to a function necessary for generating the model input 17 from the model output 18, for example.

  Next, in step 1030, the operation signal generator 300 is operated to generate the operation signal 23. The operation signal 23 is transmitted to the operation signal database 230 and the external output interface 202, and the operation signal 24 serving as a control command is given from the external output interface 202 to the plant 100.

  Next, in step 1040, the external input interface 201 is operated, the measurement signal 1 that is the control output of the plant 100 is taken into the control device 200, and the measurement signal 2 is transmitted to the operation signal generator 300 and the measurement signal database 210. To do.

  Next, in step 1050, the learning parameter determining unit 700 determines the learning parameter 8 that becomes the learning condition, and transmits the learning parameter 8 to the learning parameter database 260.

  In step 1060, the learning condition determination unit 700 compares the learning parameter 9, which is the previous value of the learning parameter stored in the learning parameter database 250, with the learning parameter 8. Is set to “0”, and if different, the learning trigger 7 is set to “1” and transmitted to the learning unit 400.

  When the learning trigger 7 is “1”, it means that the value of the learning parameter has been changed, and the processing returns to step 1010 and learning is performed using the new learning parameter 14. This is called relearning.

  Note that the learning unit 400 can perform relearning using the learning information data 21 that is the previous learning result. When the learning trigger 7 is “0” and the relearning is not performed, the process returns to step 1030.

  FIG. 12 is a diagram for explaining the learning effect learned using the calculation method shown in the flowchart of the control operation by the control device 200 according to the embodiment of the present invention shown in FIG.

  In FIG. 12, the learning condition determination unit 700 of the control device 200 sets the operation limit speed of the model input 17 as the learning parameter 8 in consideration of the operation limit speed of the operation signal 24. For this reason, by giving the plant 100 an operation signal 24 as a control command in accordance with a method of generating the model input 17 (upper diagram in FIG. 12) learned by the learning unit 400 using the model 500 of the control device 200, FIG. As shown in the figure below, it is shown that the NOx low region can be reached in the state after four operations from the initial state without making a state transition to the NOx high region.

  In addition, although a plurality of operation ends having the same design specification data are used, even when the actual operation speed varies, learning can be performed in consideration of the operation limit speed of each operation end. In addition, even when the operating end deteriorates over time and the operation speed decreases, it can be set as a condition for learning the decreased operation speed.

  Furthermore, even when the operating parameters of the plant change, such as the power generation output change operation, burner switching operation, and coal type switching operation, and the control parameters such as the change rate limiter are changed, learning should be performed under the changed conditions. Can do. Further, even when the operator of the plant 100 changes the control parameter, it is possible to learn with the changed condition.

  As a result, by giving the operation signal 24 generated according to the learned generation method of the model input 17 to the plant 100 as a control command, a desired control result can be obtained as plant control.

  In addition, since the learning condition determination unit 700 of the control device 200 automatically determines the learning constraint condition, the operation of the plant operator to determine the learning constraint condition becomes unnecessary, and the usability of the control device is improved. There is also an effect that the condition setting period for learning can be shortened.

  By the way, in the flowchart of the control operation of the control device 200 shown in FIG. 11, when the learning parameter becomes a value different from the previous value in the learning condition determination unit 700, it is necessary to perform re-learning in step 1010. is there. Since this learning requires computational resources, it is necessary to use a control device capable of high-speed computation or to spend time on learning.

  Using a control device capable of high-speed computation is expensive. Further, when taking time for learning, it is necessary to stop the operation of the learning signal generation unit 310 during the learning period, and the result learned by the learning unit 400 and the model 500 cannot be reflected in the generation of the operation signal 24.

  Therefore, as a countermeasure, in the embodiment of the present invention, the learning information adding unit 800 is provided in the control device 200 shown in FIG. When the learning trigger 7 becomes “1”, the learning information adding unit 800 generates learning information data 13 using the learning parameter data 14 and the learning information data 12 and transmits the learning information data 280 to the learning information database 280. By using the learning information adding unit 800, it is possible to generate learning information data 13 that is a learning result when the learning parameter 14 is used as a learning condition without performing relearning.

  Therefore, in consideration of the case where the learning parameter determination unit 700 changes the learning parameter, a control device capable of high-speed calculation is used, or the learning signal is changed when the learning parameter determination unit 700 changes the learning parameter. The function of the generation unit 310 does not stop.

  Next, a control operation when the learning information adding unit 800 is provided in the control device 200 will be described with reference to a flowchart shown in FIG.

  FIG. 13 is a flowchart showing the contents of calculation processing for the simulation and learning contents of the plant model in the control device 200 when the learning information adding unit 800 is installed in the plant control device according to the embodiment of the present invention. is there.

  As shown in FIG. 13, the flowchart of the control operation of the control device 200 is executed by combining steps 1110, 1120, 1130, 1140, 1150, 1160, 1170. Each step will be described below.

  First, in step 1110, the learning unit 400 learns a generation method of the model input 17 such that the model output 18 achieves the model output target value for the model 500. Note that learning may be performed using the evaluation value calculation unit 600 in the same manner as in the step 1010 of the flowchart of FIG. An optimization method similar to that in step 1010 can also be used.

  When learning is performed in step 1110, learning is performed by dividing the input space into regions using the minimum setting value of the change width of the model input 17. The minimum set value of the change width of the model input 17 is a value set by the operator of the plant 100.

  FIG. 14 is an explanatory diagram when the learning unit 400 learns the generation method of the model input 17 in step 1110 and divides the input space into regions.

  As illustrated in FIG. 14, the learning unit 400 divides the operable range of the model input A and the model input B into the minimum set value of the model input change width. Next, learning is performed by limiting the change width of the model input that can be changed by one operation to the minimum setting value of the model input change width.

  That is, in each area, an operation method for moving to an adjacent area is learned. For example, when the operation is started from the initial state using the result of learning under the condition that the number of operations is the minimum and the state is not shifted to the region where the NOx is high, the route shown in FIG. Follow the arrival route to reach the low NOx region.

  Next, in step 1120, the learning unit 400 is operated, and the result learned in step 1210 is transmitted from the learning unit 400 to the learning information database 280 as learning information data 20.

  Next, in step 1130, the learning condition determination unit 700 is operated to determine the learning condition, and the learning parameter 8 is transmitted to the learning parameter database 250.

  In step 1140, the learning condition determination unit 700 compares the learning parameter 9, which is the previous value of the learning parameter stored in the learning parameter database 250, with the learning parameter 8. If the learning trigger 7 is “1”, the process proceeds to step 1150. If the learning trigger 7 is “0”, the process proceeds to step 1160.

  Next, in step 1150, the learning information adding unit 800 is operated to use the learning information data 12 stored in the learning information database and the learning parameter 10 stored in the learning parameter database 250, and the additional learning information data. 13 is generated and transmitted to the learning information database 280.

  Note that the learning information data 12 used in step 1150 is the result of learning in step 1110.

  Next, the control operation of the learning information adding unit 800 provided in the control device 200 will be described.

  FIG. 15 is a flowchart for explaining the details of the operation of the learning information adding unit 800 provided in the control device 200 shown in FIG. 1, and explaining the details of step 1150 in the flowchart shown in FIG.

  In FIG. 15, in step 810, the number of operations required to reach the target state for each region is derived using the learning information data 12 obtained as a result of learning in step 1110. This can be derived, for example, by performing the operation of setting a certain area as an initial state and obtaining the number of operations from that point until reaching the target state in all areas.

  Next, in step 820, for each region, the learning parameter 10 is used to determine a state range (operational range) that can be changed by a single operation, and the operation obtained in step 810 for the region within the operationable range. Extract all count values.

  Next, in step 830, it is determined that the operation method for transitioning to the region in which the value of the number of operations extracted in step 820 is the smallest in a certain region is the optimal operation method, and the operation method is added to the additional learning information. The data 13 is transmitted from the learning information adding unit 800 to the learning information database 280.

  FIG. 16 is an explanatory diagram for explaining the learning result in the flowchart of FIG. 15 for explaining the operation content of the learning information adding unit 800. As shown in FIG. 16, the additional learning information data 13 generated by the learning information adding unit 800 includes the operation as indicated by the arrow in the drawing in the initial state.

  When an operation is performed in accordance with the arrow from the initial state in FIG. 16, it is possible to reach an area where the number of operations required to reach the NOx low area is minimized from the operable range in the initial state.

  The above description is the operation description of step 1150 shown in FIG.

  Next, in step 1160, the operation signal generator 300 is operated to generate the operation signal 23 using the learning information data 22 and the control logic data 11 generated in step 1150. This operation signal 23 is transmitted to the plant 100 through the external output interface 202 as an operation signal 24 serving as a control command.

  Next, in step 1170, the external input interface 201 is operated, and the measurement signal 1 which is the plant control output is taken into the control device 200. Then, it progresses to step 1130 and repeats operation | movement of the above-mentioned step 1130-step 1170.

  By the way, in the flowchart of the control operation of the control device 200 shown in FIG. 11, when the learning trigger 7 becomes “1” in the learning condition determination unit 700 of the control device 200, it is necessary to proceed to step 1010 and perform relearning again. It was.

  In contrast, in the flowchart of the control operation of the control device 200 shown in FIG. 13, even when the learning trigger 7 is “1”, the learning information adding unit 800 is operated using the result learned in step 1110. By doing this, it is possible to generate the same learning information data as when learning the generation method of the model input 17 when the learning parameter 14 (learning parameter 10) is used as a learning condition.

  As a result, in addition to the effect obtained by using the flowchart of FIG. 11, the plant can be controlled without stopping the function of the learning signal generation unit 310 even when a control device capable of high-speed calculation is not used. An effect is obtained.

  As an effect of the embodiment in which the plant control device and the control method of the present invention are applied to a thermal power plant, the concentration of NOx in exhaust gas discharged from the thermal power plant can be reduced.

  Further, as the NOx concentration is reduced, the amount of ammonia used in the denitration apparatus necessary for reducing NOx from the exhaust gas can be reduced, and the catalyst activity of the denitration apparatus can be maintained for a long time.

  Further, according to the plant control apparatus of the embodiment of the present invention, the initial value of the learning parameter used for determining the learning constraint condition is determined using the prior information (specification) regarding the operation limit speed of the operation end. Moreover, since the learning parameter is sequentially corrected using the measurement signal, the operation speed of the operation end of the plant can be reflected in the learning parameter.

  For example, when a plurality of operation terminals with design specifications are used and the actual operation speed varies, learning can be performed in consideration of the operation speed of each operation terminal. In addition, even if the operating end deteriorates over time and the operating speed decreases, it is possible to learn the reduced operating speed as a constraint and control the plant well, so that the plant can be operated safely. An effect is obtained.

  In addition, the use of the plant control device of this embodiment eliminates the need for the plant operator to determine learning constraint conditions, thereby improving the usability of the control device and shortening the condition setting period for learning. The effect is also obtained.

The present invention can be applied to a plant control apparatus and a plant control method such as a thermal power plant.

The block diagram which shows the whole structure of the control apparatus of the plant which is one Example of this invention. The block diagram of the thermal power plant to which the control apparatus of the plant which is one Example of this invention is applied. The enlarged view of the piping part and air heater part of the thermal power plant shown in FIG. The block diagram of the operation signal production | generation part in the control apparatus of the plant shown in FIG. Explanatory drawing of the control parameter setting screen in the control apparatus of the plant shown in FIG. Explanatory drawing of the function of the learning condition determination part in the control apparatus of the plant shown in FIG. Explanatory drawing which shows an example of the learning parameter update method of the learning condition determination part in the control apparatus of the plant shown in FIG. Explanatory drawing which shows the relationship between the model input and model output of the model in the control apparatus of the plant shown in FIG. Explanatory drawing which shows the learning result which learned the production | generation method of model input for the model of the learning part in the control apparatus of the plant shown in FIG. Explanatory drawing which shows the learning result of the operation signal learned and produced | generated by the learning part in the control apparatus of the plant shown in FIG. The flowchart which shows the arithmetic processing content of the control apparatus of the plant which is one Example of this invention. Explanatory drawing which shows the learning result of the model input and operation signal which were learned based on the flowchart shown in FIG. The flowchart which shows the arithmetic processing content at the time of installing the learning information addition part in the control apparatus of the plant which is one Example of this invention. Explanatory drawing of the method of dividing | segmenting the input space of the model input learned based on the flowchart shown in FIG. 13 into an area | region. 14 is a flowchart showing details of step 1150 in the flowchart shown in FIG. 13. Explanatory drawing which shows the learning result learned using the flowchart shown in FIG.

Explanation of symbols

1, 2: Measurement signal, 3: Measurement signal data, 8, 9, 10: Learning parameter, 17: Model input, 18: Model output, 19: Evaluation value, 23: Operation signal, 24: Command signal, 100: Plant 100a: Thermal power plant, 101: Boiler for pulverized coal, 200: Control device, 201: External input interface, 202: External output interface, 210: Measurement signal database, 220: Operation end specification database, 230: Operation signal database, 240: control logic database, 250: learning parameter database, 260: evaluation value calculation parameter database, 270: model parameter database, 280: learning information database, 300: operation signal generation unit, 400: learning unit, 500: model, 600: Evaluation value calculation unit, 700: learning Condition determining unit, 800: learning information adding unit, 900: external input device, 901: keyboard, 902: mouse, 910: maintenance tool, 920: external input interface, 930: data transmission / reception processing unit, 940: external output interface, 950 : Image display device.

Claims (13)

In a plant control apparatus including an operation signal generation unit that calculates an operation signal that is a control command to be given to the plant using a measurement signal that is an operation state quantity of the plant,
The control device includes a model that simulates the control characteristics of the plant to be controlled, a control logic database that stores control logic data including control parameters used to calculate operation signals in the operation signal generation unit, An operation end specification database storing operation end specification data of an operation end for controlling the state quantity, an operation signal database storing past operation signals, and a measurement signal database storing past measurement signals ,
The operation limit speed, upper limit value, and lower limit value of the operation end that is the limit value of the operation signal change width per unit time using the control logic data and the operation end specification data stored in the control logic database and the operation end specification database. The measurement signal data per unit time, which is the sample control cycle, is determined using the function for determining the initial value of the learning parameter including the operation signal data and the measurement signal data stored in the operation signal database and the measurement signal database . Learning with a function of updating the operation limit speed of the operation end included in the learning parameter, which is control logic data, to the value of the change amount of the measurement signal data when the change amount is smaller than the change amount of the operation signal data A condition determination unit;
Model input that the model output simulated by the model achieves the target value of the model output using the model with the limit value of the operation signal change width per unit time included in the learning parameter set as the learning constraint condition A learning unit that learns how to generate
Each of the learning units is equipped with a learning information database storing learning information data that is a result of learning how to generate a model input.
The operation signal generation unit includes a learning signal generation unit that calculates an operation signal for the plant using a measurement signal that is an operation state quantity of the plant and learning information data stored in a learning information database. Plant control device. The plant control apparatus according to claim 1,
When the learning parameter is changed to a value different from the learning parameter of the previous value, the control device uses the learning information data stored in the learning information database to per unit time included in the learning parameter. A function for generating learning information data when learning is performed in the learning unit by setting the limit value of the operation signal change width in the learning constraint and transmitting the generated additional learning information data to the learning information database. A plant control apparatus comprising a learning information adding unit. In a plant control device for controlling a thermal power plant by calculating an operation signal as a control command to be given to the thermal power plant using a measurement signal which is an operation state quantity of the thermal power plant,
The control device simulates the control characteristics of a thermal power plant to be controlled, and an operation signal generation unit that calculates an operation signal as a control command to be given to the plant using a measurement signal that is an operation state quantity of the thermal power plant. Control logic database that stores the model and control logic data including control parameters used to calculate the operation signal in the operation signal generator, and operation end specification data for the operation end that controls the state quantity of the thermal power plant Operation terminal specification database, operation signal database in which past operation signals are stored, measurement signal database in which past measurement signals are stored,
Using the control logic data and operation end usage data stored in the control logic database and the operation end specification database, the operation limit speed, upper limit value, and lower limit value of the operation end that is the limit value of the operation signal change width per unit time are Changes in measurement signal data per unit time, which is the sample control cycle, using the function to determine the initial values of the included learning parameters and the operation signal data and measurement signal data stored in the operation signal database and measurement signal database When the amount is smaller than the change amount of the operation signal data, the learning condition determination has a function of updating the operation limit speed of the operation end included in the learning parameter that is the control logic data to the value of the change amount of the measurement signal data. And
Model input that the model output simulated by the model achieves the target value of the model output using the model with the limit value of the operation signal change width per unit time included in the learning parameter set as the learning constraint condition A learning unit that learns how to generate
Each of the learning units is equipped with a learning information database storing learning information data that is a result of learning how to generate a model input.
The operation signal generation unit includes a learning signal generation unit that calculates an operation signal for a thermal power plant using a measurement signal that is an operation state quantity of the plant and learning information data stored in a learning information database. A plant control device. In the plant control apparatus according to claim 3,
The measurement signal includes at least one of nitrogen oxide concentration, carbon monoxide concentration, carbon dioxide concentration, sulfide oxide, and mercury, and the operation signal determines at least one of the air damper opening, air flow rate, and fuel flow rate The thermal power plant uses the data stored in the control logic database, the operation signal database, and the measurement signal database in the learning condition determination unit provided in the control device, and the burner switching operation, the coal type switching operation, and the load A function that estimates whether or not driving including at least one of the changing driving is performed and updates the learning parameter based on the estimation result, and the operation using the data stored in the operation signal database and the measurement signal database. The function to update the learning parameters based on the estimation result by estimating the movement speed of the edge The plant control system according to claim. In the plant control apparatus according to claim 4,
When the learning parameter is changed to a value different from the previous learning parameter by the control device, the operation per unit time included in the learning parameter using the learning information data stored in the learning information database It has a function to generate learning information data when learning is performed in the learning unit by setting the limit value of the signal change width as a learning constraint condition, and to transmit the generated additional learning information data to the learning information database A plant comprising a learning information adding unit, wherein the learning signal generating unit of the operation signal generating unit is configured to calculate an operation signal using additional learning information data stored in a learning information database. Control device. In the plant control apparatus according to claim 1 or 3,
Of the control parameters stored in the control logic database, the parameters set to limit the signal change width per unit time and the operation end specification database are stored in the learning condition determining unit of the control device. A plant control device characterized in that it has a function of comparing values of operation speeds at the operating end and setting a value having a small absolute value as an initial value of a learning parameter. In the plant control apparatus according to claim 3,
A plant characterized in that a user interface for setting control parameters used in the control device is provided for each of normal operation, burner switching operation, coal type switching operation or load change operation which is an operation form of a thermal power plant. Control device. In a plant control method for controlling a plant by calculating an operation signal as a control command to be given to the plant using a measurement signal that is an operation state quantity of the plant,
Simulate the control characteristics of the plant to be controlled by the model provided in the plant control device and control the control logic data including the control parameters used to calculate the operation signal by the operation signal generator provided in the control device. Save in the logic database, save the operating end specification data of the operating end that controls the state quantity of the plant in the operating end specification database provided in the control device, and save the past operation signals in the operation signal database provided in the control device. , Store past measurement signals in the measurement signal database provided in the control device,
By using the control logic data and operation end usage data stored in the control logic database and the operation end specification database by the learning condition determination unit provided in the control device, the operation end that is the limit value of the operation signal change width per unit time is set. The initial value of the learning parameter including the operation limit speed, the upper limit value, and the lower limit value is determined, and the operation signal data and the measurement signal data stored in the operation signal database and the measurement signal database are used as a sample control cycle. When the change amount of the measurement signal data per unit time is smaller than the change amount of the operation signal data, the value of the operation limit speed of the operation end included in the learning parameter, which is control logic data, is changed. Update to the value of
The model output simulated by the model using the model by setting the limit value of the operation signal change width per unit time included in the learning parameter as a learning constraint by the learning unit provided in the control device is the model output. In order to achieve the target value, the learning unit simulates the characteristics of the plant to learn the model input generation method,
The learning information data that is the result of learning how to generate the model input in the learning unit is stored in the learning information database,
A learning signal generation unit provided in the operation signal generation unit calculates an operation signal serving as a control command to be given to the plant using a measurement signal that is a plant operating state quantity and learning information data stored in a learning information database. A plant control method characterized by controlling the plant. The plant control method according to claim 8,
When learning the model input generation method by simulating the characteristics of the plant, the learning information data stored in the learning information database is used when the learning parameter is changed to a value different from the previous learning parameter. The learning information data is generated when learning is performed with the limit value of the operation signal change width per unit time included in the learning parameter set as the learning constraint, and the generated additional learning information data is learned. A plant control method characterized in that an operation signal serving as a control command given to a plant is calculated in addition to learning information data in an information database to control the plant. In a plant control method for controlling a thermal power plant by calculating an operation signal as a control command to be given to the thermal power plant using a measurement signal that is an operation state quantity of the thermal power plant,
The control device is provided with control logic data including control parameters used to calculate the operation signal by the operation signal generation unit provided in the control device, simulating the control characteristics of the plant to be controlled by the model provided in the plant control device. Stored in the control logic database, the operation end specification data of the operation end for controlling the state quantity of the plant is stored in the operation end specification database provided in the control device, and the past operation signals are stored in the operation signal database provided in the control device. Save and save past measurement signals in the measurement signal database provided in the control device,
The operation limit speed limit and the upper limit value of the operation end, which is the limit value of the operation signal change width per unit time, using the data stored in the control logic database and the operation end specification database by the learning condition determination unit provided in the control device. And an initial value of the learning parameter including the lower limit value, and the operation signal data and measurement signal data stored in the operation signal database and the measurement signal database are used per unit time which is a sample control period. When the change amount of the measurement signal data is smaller than the change amount of the operation signal data, the value of the operation limit speed of the operation end included in the learning parameter that is the control logic data is set to the change amount value of the measurement signal data. To update,
The model output simulated by the model using the model by setting the limit value of the operation signal change width per unit time included in the learning parameter as a learning constraint by the learning unit provided in the control device is the model output. In order to achieve the target value, the learning unit simulates the characteristics of the plant and learns how to generate the model input,
The learning information data, which is the result of learning how to generate the model input in the learning unit, is stored in the learning information database,
A learning signal generation unit provided in the operation signal generation unit calculates an operation signal serving as a control command to be given to the plant using a measurement signal that is a plant operating state quantity and learning information data stored in a learning information database. A plant control method characterized by controlling the plant. The plant control method according to claim 10,
Use the data stored in the control logic database and operation end specification database, including the concentration of the components contained in the combustion gas in the measurement signal, the signal necessary for operating the equipment constituting the operation end in the operation signal Determine the initial values of the learning parameters, and determine the control parameters for each of the operation forms of the power generation output constant operation, the power generation output change operation, the burner switching operation, and the coal type switching operation, which are the operation forms of the thermal power plant. A control method for a plant, wherein the learning parameter is updated based on the control parameter. The plant control method according to claim 11,
When the learning parameter is changed to a value different from the learning parameter of the previous value, using the learning information data stored in the learning information database, an operation signal change width per unit time included in the learning parameter To generate learning information data when learning is performed in the learning unit with the limit value of learning set as a learning constraint condition, and an operation signal for the thermal power plant is calculated using the generated additional learning information data A plant control method characterized by that. In the plant control method according to claim 8 or 10,
Of the control parameters stored in the control logic database, compare the parameter set to limit the change width of the signal per unit time with the operation speed value stored in the operation terminal specification database. A plant control method characterized in that a value having a small absolute value is set as an initial value of a learning parameter.

Images