JP5277064B2 - Plant control device, thermal power plant control device, and thermal power plant - Google Patents

Abstract

The invention aims to provide a control device for complete equipment and a control device for thermal power generation complete equipment, which can construct a high-precision statistical model and acquire expected control effect under the condition of the learning type complete equipment even if the data used in the control has deviation during constructing the statistical model. The control device for the complete equipment is provided with a statistical model for presuming the value of a measurement signal acquired when a control signal is supplied to the complete equipment, a model construction database for storing data used in construction of the statistical model, an operating method learning part for a generating method input by a learning model in order that the model output reaches a target value, and a model adjusting part for adjusting a radius parameter of the statistical model in the information stored in the model construction database, wherein the statistical model generates the model output by using the adjustment result of the radius parameter obtained by the model adjusting part.

Description

  The present invention relates to a plant control device, and more particularly to a thermal power plant control device that generates power using fossil fuels such as coal.

  The plant control apparatus processes a measurement signal of a state quantity obtained from a plant that is a control target, calculates a control signal (operation signal) to be given to the control target, and transmits the control signal to the control target. An algorithm for calculating an operation signal is mounted on the control device of the plant so that the measurement signal of the state quantity 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 multiplying the deviation between the measurement signal of the plant state quantity and its target value by a proportional gain is added to a value obtained by time-integrating the deviation to derive an operation signal to be given to the controlled object.

  Since the control algorithm using PI control can describe the input / output relationship with a block diagram or the like, the causal relationship between the input and the output is easy to understand, and has a lot of application results. However, when the plant is operated under conditions that are not assumed in advance, such as a change in the operation state of the plant or a change in the environment, an operation such as changing the control logic may be required.

  On the other hand, control methods that can adapt to changes in plant operating conditions and environments include control algorithms and control methods that use adaptive control that automatically corrects parameter values and learning algorithms.

  As a method of deriving the operation signal of the plant controller using the learning algorithm, the plant characteristics are estimated using the measured data of the plant and the data constructed based on the numerical analysis, and the characteristics of the plant are estimated. A general method is to construct a statistical model and autonomously learn optimal control logic for this statistical model.

  The performance of the control method obtained by using this autonomous learning method depends on the estimation accuracy of the statistical model. That is, the closer the characteristic of the statistical model is to the actual plant characteristic, the more control effect equivalent to the control result for the statistical model is obtained. Therefore, construction of a more accurate statistical model becomes a problem in adaptive control technology using a learning algorithm.

  As a technique for improving the accuracy of a statistical model, Patent Document 1 describes a technique for optimizing a radius parameter of a basis function of an RBF network, which is one of statistical methods, using a search algorithm.

  Non-Patent Document 1 discloses a technique in which a statistical model is configured by an RBF network, and a radius parameter of the basis function is determined by a calculation formula derived based on data characteristics (number of data, dimensions, distance) for model construction. Is described.

  In the above known technique, the radius is adjusted based on the characteristics of the model construction data, so that it is possible to construct a statistical model that matches the actual plant characteristics more than when the radius is set at random.

JP 2005-15639 A

Kitayama, Yasuda, Yamazaki: "Integrated optimization using RBF network and Particle Swarm Optimization", IEEJ Transactions C, Vol. 128, no. 4, pp. 636-645 (2008).

  When the techniques disclosed in Patent Document 1 and Non-Patent Document 1 are applied to a plant control device, the estimation accuracy is higher than when a radius parameter is set randomly by setting a radius parameter of a statistical model based on data. Can be improved.

  On the other hand, in plant control, it takes several minutes to several tens of minutes for the characteristics to stabilize after changing the operating conditions by control.By changing the operating conditions using this time as the control cycle, the maximum effect can be obtained. Expected to get.

  Therefore, it is preferable that the parameter adjustment of the statistical model and the learning of the control logic by the learning algorithm are finished within this control cycle.

  However, when the technique of Patent Document 1 is applied to a plant control device, the radius parameter adjustment method using an optimization technique like the technique of Patent Document 1 minimizes the error between the data and the statistical model estimation result. In order to search for the radius parameter, iterative error evaluation according to the number of searches is required. Since the number of error evaluations increases in proportion to the number of data, the calculation cost increases when the data size is large, and parameter adjustment of the statistical model may not be completed within the control cycle.

  On the other hand, in the technique of Non-Patent Document 1, since the radius parameter of the statistical model is determined using a calculation formula derived in advance, iterative error evaluation is unnecessary, and the calculation cost is reduced and the statistical model adjustment is completed within the control cycle. it can. In addition, this technique is based on the premise that data for constructing a statistical model is distributed at a certain interval, and the radius parameters of all bases are set uniformly.

  However, when the technique of Non-Patent Document 1 is applied to a plant control device, plant operation control is premised on stable operation under optimum operating conditions in order to satisfy the performance guarantee value. It is assumed that the distribution of data for model construction acquired will be concentrated in a region that is almost close to the optimum condition, and from the viewpoint of compliance with the performance guarantee values described above, there is also an artificial operation for distributing the data at even intervals. Often not allowed.

  Furthermore, if the model building data is concentrated and distributed under certain conditions, the statistical model basis functions cannot sufficiently cover the input space with a uniform radius parameter, and the estimation accuracy may be significantly reduced. There is sex.

  The object of the present invention is to control the parameters of the statistical model appropriately within the control period according to the distribution of the bias of the data even when controlling the learning type plant in which the bias used in the data used when constructing the statistical model is controlled. An object of the present invention is to provide a plant control device and a thermal power plant control device having a function of adjusting and improving estimation accuracy.

In the plant control device of the present invention, the plant control device includes a control device that takes in a measurement signal that is a state quantity of the plant from the plant and calculates an operation signal that controls the plant using the measurement signal. The control device includes a measurement signal database that captures and stores a measurement signal that is a state quantity of the plant, a model construction database that stores model construction data converted from measurement data of the plant stored in the measurement signal database, and A statistical model for simulating plant control characteristics for estimating a value of a measurement signal that is a state quantity of the plant when a control signal is given to the plant using model building data stored in the model building database; Using the model, the model output corresponding to the measurement signal will achieve the target value. An operation method learning unit that learns a generation method of a model input corresponding to the control signal given to the runt, a learning information database that stores learning information data related to learning constraint conditions and learning results in the operation method learning unit, and the measurement A model that includes a measurement signal in the signal database and a control signal generation unit that calculates a control signal transmitted to the plant using the learning information data in the learning information database, and is stored in the model construction database The construction data is configured to include sparse density information indicating the density of the data, and the control device adjusts the base radius parameter of the statistical model included in the model construction data stored in the model construction database. A model adjustment unit that provides a basis radius parameter by the model adjustment unit. Using the adjustment results of the meter, characterized by being configured to generate a model output.

The control apparatus for a thermal power plant according to the present invention takes in a measurement signal, which is a state quantity of the plant, from a thermal power plant equipped with a boiler, and calculates an operation signal for controlling the thermal power plant using the measurement signal In the control device for a thermal power plant including the device, the measurement signal is at least one of the concentrations of nitrogen oxide, carbon monoxide, and hydrogen sulfide contained in the exhaust gas discharged from the boiler of the thermal power plant. The operation signal is discharged from the boiler, the flow rate of air supplied to the boiler of the thermal power plant, the opening degree of the air damper for adjusting the air flow rate, the flow rate of fuel supplied to the boiler exhaust gases comprises a signal representing at least one of the exhaust gas recirculation flow recirculating in the boiler, the control device, the thermal power Pla A measurement signal database that captures and stores measurement signals that are the state quantities of the engine, an air flow rate that is supplied to the boiler converted from the measurement data of the plant stored in the measurement signal database, and an air damper that adjusts the air flow rate is opened. A model construction database for storing model construction data including at least one of a flow rate of fuel supplied to the boiler and an exhaust gas recirculation flow rate for recirculating exhaust gas discharged from the boiler to the boiler, and the model construction database A statistical model that simulates a control characteristic of a plant that estimates a value of a measurement signal that is a state quantity of the plant when a control signal is given to the plant using stored model construction data, and the statistical model is used to The control signal given to the plant so that the model output corresponding to the measurement signal achieves the target value An operation method learning unit for learning a corresponding model input generation method, a learning information database for storing learning constraint data and learning results in the operation method learning unit, a measurement signal in the measurement signal database, and The model construction data stored in the model construction database includes a control signal generator configured to calculate a control signal transmitted to the plant using the learning information data of the learning information database . The controller is configured to include sparse density information indicating a degree of congestion, and the controller is further provided with a model adjusting unit that adjusts a base radius parameter of a statistical model included in model building data stored in the model building database. The statistical model is modeled using the adjustment result of the base radius parameter by the model adjustment unit. It is configured to generate a digital output.
The thermal power plant of the present invention is a thermal power plant including a control device that takes in a measurement signal that is a state quantity of the plant from the plant and calculates an operation signal for controlling the plant using the measurement signal. The power plant includes a boiler that generates steam, a steam turbine that is driven by the steam generated in the boiler, and a generator that generates power by driving the steam turbine, and the control device is in a state of the plant. A measurement signal database that captures and stores measurement signals that are quantities, a model construction database that stores model construction data converted from plant measurement data stored in the measurement signal database, and a model that is saved in the model construction database When a control signal is given to the plant using construction data, A statistical model that simulates plant control characteristics for estimating the value of a measurement signal that is a state quantity, and the control signal that is applied to the plant using the statistical model so that a model output corresponding to the measurement signal achieves a target value An operation method learning unit that learns a generation method of a model input corresponding to the above, a learning information database that stores learning information data related to learning constraint conditions and learning results in the operation method learning unit, a measurement signal of the measurement signal database, And a control signal generation unit that calculates a control signal transmitted to the plant using the learning information data of the learning information database, and the model construction data stored in the model construction database includes data Sparse density information indicating the degree of congestion of the model construction data is further included in the control device. A model adjustment unit that adjusts the base radius parameter of the statistical model included in the model construction data stored in the database, and the statistical model generates the model output using the adjustment result of the base radius parameter by the model adjustment unit It is characterized by having constituted so .

  According to the present invention, even when controlling a learning type plant in which there is a bias in data used when building a statistical model, the parameters of the statistical model are appropriately set within the control period according to the distribution of the bias in the data. It is possible to realize a plant control apparatus and a thermal power plant control apparatus having a function of adjusting and improving the estimation accuracy.

The block diagram which shows the structure of the control apparatus of the plant which is 1st Example of this invention. The flowchart which shows the operation | movement flow at the time of learning of the operation method in the control apparatus of the plant by 1st Example of this invention described in FIG. The block diagram which shows the structure of the model adjustment part in the control apparatus of the plant by 1st Example of this invention described in FIG. The figure which shows the aspect of the data preserve | saved in the model construction data in the control apparatus of the plant by 1st Example of this invention described in FIG. The flowchart which shows the operation | movement flow of the model adjustment in the control apparatus of the plant by 1st Example of this invention described in FIG. Schematic which shows the concept of the category division | segmentation at the time of model adjustment in the control apparatus of the plant by 1st Example of this invention described in FIG. The schematic diagram explaining the model adjustment mechanism in the control apparatus of the plant by 1st Example of this invention described in FIG. The schematic diagram which shows the mode of the base sparse density distribution change at the time of model adjustment in the control apparatus of the plant by 1st Example of this invention described in FIG. The schematic diagram which shows the mode of the model estimated value distribution change at the time of model adjustment in the control apparatus of the plant by 1st Example of this invention described in FIG. The plant control apparatus by 1st Example of this invention described in FIG. 1 WHEREIN: An example of the screen displayed on an image display apparatus when setting model input / output. The plant control apparatus by 1st Example of this invention described in FIG. 1 WHEREIN: An example of the screen displayed on an image display apparatus when setting model adjustment conditions. The plant control apparatus by 1st Example of this invention described in FIG. 1 WHEREIN: An example of the screen displayed on an image display apparatus, when confirming a model adjustment result. The schematic block diagram which shows the structure of the thermal power plant which is the 2nd Example to which the control apparatus of the plant of this invention is applied. The schematic structure figure which shows the structure of the air heater with which the thermal power plant of 2nd Example described in FIG. 13 was equipped. FIG. 14 is an example of a screen displayed on the image display device when setting model input / output in the thermal power plant control apparatus according to the second embodiment of the present invention described in FIG. 13;

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

In the plant control apparatus having a configuration common to both the plant control apparatus and the thermal power plant control apparatus according to the present invention, the model adjustment unit configuring the control apparatus uses information stored in the model construction database. At least one of a category calculation function for determining a category number of model building data and a radius adjustment function for adjusting a radius parameter of the statistical model using model building data including category information determined by the category calculation function. It is desirable to provide.

  The information stored in the model construction database preferably includes at least one of the coordinates of each data in the model input space, the radius parameter, the data sparse density, and the category number to which the data belongs.

  The category operation function calculates the sparse density distribution range of the model construction data based on the function for calculating the sparse density, which is an index indicating the density of each data, and the category number information input from the external input device. It is desirable to provide at least one of the functions for determining the category number of each data on the basis of the values equally divided by.

  The radius adjustment function calculates the sparse density for the reference model input arbitrarily determined in the model input space when adjusting the radius parameter, and the calculated sparse density is input from an external input device. When the threshold condition is not satisfied, it is desirable to provide at least one of the functions of extracting the category of data located closest to the reference model input and adjusting the radius parameter of the data belonging to the category.

  The control device is connected to the image display device, a function of displaying information stored in the model construction database on the image display device, a function of setting a model adjustment condition used in the model adjustment unit via the image display device, It is desirable to have at least one of the functions for displaying the adjustment result of the statistical model by the model adjustment unit on the image display device.

  By providing a function for inputting model adjustment condition settings via the image display device, the plant operator can set appropriate model adjustment conditions according to the control needs of the plant. Furthermore, by providing a function to display the transition of sparse density due to model adjustment and the transition of error between data and estimation results on the image display device, the plant operator could obtain the desired model estimation accuracy by model adjustment. If it cannot be obtained, model adjustment can be executed again.

  Moreover, when applying the control apparatus of this invention to a thermal power plant, the thermal power plant of the structure provided with the control signal generation part which derives | leads-out the control signal given to a thermal power plant using the measurement signal acquired from a thermal power plant It becomes the control device.

  These measurement signals include signals representing at least one of the concentrations of nitrogen oxide, carbon monoxide, and hydrogen sulfide contained in the gas discharged from the thermal power plant. The control signal includes a signal for determining at least one of the opening degree of the air damper, the air flow rate, the fuel flow rate, and the exhaust gas recirculation flow rate.

  The control device includes a statistical model for estimating a value of a measurement signal when a control signal is given to the thermal power plant, and an opening degree of an air damper, an air flow rate, a fuel used in the construction of the statistical model The control signal so that the model output corresponding to the measurement signal achieves a target value by using a model construction database for storing data including at least one information of the flow rate and the exhaust gas recirculation flow rate, and the statistical model. An operation method learning unit that learns a generation method of a model input corresponding to the above, a learning information database that stores information on learning constraint conditions and learning results in the operation method learning unit, and information stored in the model construction database A model adjustment unit for adjusting a radius parameter of the statistical model included.

  In addition, the control device is connected to the image display device, a function for displaying information stored in the model construction database on the image display device, and a function for setting a model adjustment condition used in the model adjustment unit via the image display device. It is desirable to provide at least one of the functions for displaying the adjustment result of the statistical model by the model adjustment unit on the image display device.

  In an embodiment in which the control device of the present invention is applied to control of a thermal power plant, setting information regarding the burner corresponding to the model input in the thermal power plant and the air amount of the after-air port is input via the image display device.

  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.

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

  FIG. 1 is a system configuration diagram of a plant control apparatus according to a first embodiment of the present invention. As shown in FIG. 1, a plant 100 to be controlled is controlled by a control device 200.

  Since the control device 200 that controls the plant 100 is connected to the maintenance tool 910, an operator of the plant 100 can connect the external input device 900 connected to the maintenance tool 910 and an image display device (for example, a CRT display) 920. Thus, the control device 200 can be controlled.

  The control device 200 includes a measurement signal conversion unit 300, a numerical analysis unit 400, a statistical model 500, a model adjustment unit 600, a control signal generation unit 700, and an operation method learning unit 800 as arithmetic devices. ing.

  The control device 200 is provided with a measurement signal database 210, a model construction database 220, a learning information database 230, a control logic database 240, and a control signal database 250 as databases (DB).

  Further, 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 obtained by measuring various state quantities of the plant from the plant 100 is taken into the measurement signal database 210 of the control device 200 via the external input interface 201. The control signal generator 700 is configured to output the control signal 15 for controlling the plant 100 to be controlled, for example, as the control signal 16 for controlling the supplied air flow rate, via the external output interface 202. ing.

  In the control device 200, the measurement signal 2 obtained by measuring the state quantity of the plant 100 captured from the plant 100 via the external input interface 201 is stored in the measurement signal database 210.

  Further, the control signal 15 generated by the control signal generation unit 700 provided in the control device 200 is stored in the control signal database 250 provided in the control device 200, and the operation signal for the plant 100 from the external output interface 202. 16 is output.

  The measurement signal conversion unit 300 provided in the control device 200 converts the measurement data 3 stored in the measurement signal database 210 into model construction data 4. The model construction data 4 is stored in the model construction database 220. In addition, the operation condition obtained as the control result immediately before included in the measurement data 3 is input to the control signal generation unit 700 provided in the control device 200.

  The numerical analysis unit 400 provided in the control device 200 predicts the characteristics of the plant 100 using a physical model that simulates the plant 100. The numerical analysis data 5 obtained by being executed by the numerical analysis unit 400 is stored in the model construction database 220.

  The model adjustment unit 600 provided in the control device 200 updates the model parameter information included in the model construction data 7 fetched from the model construction database 220 (adjusts the model), and uses the updated model construction data 8 as the model construction database 220. Save to.

  The operation method learning unit 800 provided in the control device 200 generates learning data 12 and stores it in the learning information database 230.

  The statistical model 500 provided in the control device 200 has a function of simulating the control characteristics of the plant 100. That is, the operation signal 16 is given to the plant 100, and a function equivalent to obtaining the measurement signal 1 for the control result is simulated. For this simulation operation, the statistical model 500 uses the model input 9 received from the operation method learning unit 800 and the model construction data 6 stored in the model construction database 220.

  The model input 9 corresponds to the operation signal 16. From the model input 9 and the model construction data 6, the statistical model 500 obtains a model output 10 by simulating a characteristic change due to control of the plant 100 by a statistical method using a basis function.

  The model output 10 obtained by the statistical model 500 is a predicted value of the measurement signal 1 of the plant 100. Note that the number of model inputs 9 and model outputs 10 is not limited to one, and a plurality of types can be prepared.

  Here, a statistical method using a basis function is a method in which a basis function is arranged on a vector space having model inputs as components according to statistical data (corresponding to model construction data 6 in the present invention) information held. This is a method of outputting a simulation calculation result (model characteristic) of the plant characteristic by the linear combination.

Typical techniques, radial basis function networks is one technique of the neural network (R adial B asis F unction N etwork) but are exemplified, the present invention not limited to this respect to the configuration of the statistical model, the basis functions Other techniques used can also be used. As the basis function, a radial function (Gaussian function) is generally used, and its shape is determined by a radius parameter representing the spread of the radiation.

In the control signal generation unit 700 provided in the control device 200, the measurement signal 1 is set to a desired value using the learning information data 13 output from the learning information database 230 and the control logic data 14 stored in the control logic database 240. The control signal 15 is generated as follows.

The control logic database 240 stores a control circuit for calculating the control logic data 14 and control parameters. As a control circuit for calculating the control logic data 14, a publicly known PI (proportional / integral) control can be used.

  The operation method learning unit 800 learns the operation method of the model input 9 using the learning information data 11 including the learning constraint condition and the learning parameter setting condition stored in the learning information database 230. The learning data 12 that is the learning result is stored in the learning information database 230.

  As described above, in the operation of the control device 200, the model adjustment unit 600 has a mechanism for adjusting the model parameter information included in the model construction data 7 stored in the model construction database 220. Therefore, the estimation accuracy of the plant characteristics in the statistical model 500 can be improved.

  In addition, since the model adjustment is executed according to an algorithm formulated based on the density distribution of data, the time required for model adjustment is shortened compared to the trial and error radius parameter determination means described in Patent Document 1. it can.

  Detailed functions of the statistical model 500, the model adjustment unit 600, and the operation method learning unit 800 installed in the control device 200 will be described later.

  The learning data 12 stored in the learning information database 230 from the operation method learning unit 800 includes information about model inputs before and after the operation and model outputs obtained as a result of the operation.

  In the learning information database 230, the learning data 12 corresponding to the current driving condition is selected and input to the control signal generation unit 700 as learning information data 13.

  An operator of the plant 100 is provided in the control device 200 by using an external input device 900 composed of a keyboard 901 and a mouse 902, a maintenance tool 910 that can transmit and receive data to and from the control device 200, and an image display device 920. You can access information stored in various databases.

  Also, by using these devices, parameter setting values, learning constraint conditions, and obtained learning used in the numerical analysis unit 400, the statistical model 500, the model adjustment unit 600, and the operation method learning unit 800 of the control device 200 are used. Setting information necessary for checking the result can be input.

  The maintenance tool 910 includes an external input interface 911, a data transmission / reception processing unit 912, and an external output interface 913, and can transmit / receive data to / from the control device 200 via the data transmission / reception processing unit 912.

  The maintenance tool input signal 91 generated by the external input device 900 is taken into the maintenance tool 910 via the external input interface 911. The data transmission / reception processing unit 912 of the maintenance tool 910 acquires the input / output data information 90 from the control device 200 according to the information of the maintenance tool input signal 92.

  Further, in the data transmission / reception processing unit 912, parameter setting values used in the numerical analysis unit 400, the statistical model 500, the model adjustment unit 600, and the operation method learning unit 800 of the control device 200 according to the information of the maintenance tool input signal 92, the learning method The input / output data information 90 including the constraint conditions and setting information necessary for visual recognition of the obtained learning result is output.

  The data transmission / reception processing unit 912 transmits a maintenance tool output signal 93 obtained as a result of processing the input / output data information 90 to the external output interface 913. The maintenance tool output signal 94 transmitted from the external output interface 913 is displayed on the image display device 920.

In the control device 200, the measurement signal database 210, the model construction database 220, the learning information database 230 , the control logic database 240, and the control signal database 250 are arranged inside the control device 200. Alternatively, a part can be arranged outside the control device 200.

  Further, although the numerical analysis unit 400 is disposed inside the control device 200, it can also be disposed outside the control device 200.

  For example, the numerical analysis unit 400 and the model construction database 220 may be arranged outside the control device 200, and the numerical analysis data 5 may be transmitted to the control device 200 via the Internet.

  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 flowchart showing the operation | movement at the time of the adjustment of the statistical model 500 by the model adjustment part 600 installed in the control apparatus 200 of the plant of 1st Example, and the learning of the operation method by the operation method learning part 800 is shown. .

  The flowchart shown in FIG. 2 executes steps 1000, 1100, 1200, 1300, 1400, 1500, 1600, 1700, 1800, 1900, 2000, 2100, and 2200 in combination. Hereinafter, each step will be described.

  After the operation of the control device 200 is started, first, in step 1000 for setting the model construction condition / learning condition, various parameters such as the execution condition at the time of model construction, the maximum number of times of learning at the time of learning, the maximum number of operations, and the constraint conditions Set the value.

  Next, in step 1100 for adjusting the radius of the statistical model, the model adjustment unit 600 of the control device 200 is operated to update the model parameters included in the model construction data 7. Here, the model parameter includes base radius information, central sparse density information, and category number information of each data. The detailed function and operation of the model adjustment unit 600 will be described later.

  Next, in step 1200 for learning the parameters of the statistical model, the statistical model 500 of the control device 200 is operated, and the parameters used for calculating the estimated value of the statistical model 500 are learned. As specific means for learning, various known and commonly used methods can be used.

  Next, in step 1300 for initializing learning number k (k = 1), learning number k, which is a value indicating the number of repetitions of steps 1400 to 2100, is initialized (set to k = 1).

  Next, in step 1400 for determining the initial value of the model input, the initial value of the model input 9 when starting learning is set. As an initial value of the model input 9, an arbitrary value within a preset operable range can be selected. That is, any state can be selected as an initial condition within the operable range. The model input 9 is usually expressed as a continuous value vector, but a discrete value vector may be used.

  Next, in step 1500 in which the number of operations o is initialized (o = 1), the number of operations o that is the number of repetitions of steps 1600 to 2000 is initialized (set to o = 1).

  Next, in step 1600 for updating the model input, the model input 9 is updated using the determined operation amount of the model input 9.

  Next, in step 1700 for calculating the basis function value of the statistical model, the updated model input 9 is input to the statistical model 500, and each basis function value of the statistical model 500 is calculated.

  Next, in step 1800 for calculating the model output, the model output, which is the calculation result of the statistical model 500, from the parameters of the statistical model 500 obtained in step 1200 and the basis function values of the statistical model 500 obtained in step 1700. 10 is calculated.

  Next, in step 1900 for learning the operation method, the operation method learning unit 800 is operated, and based on the value of the model output 10 calculated by the statistical model 500, the operation method learning unit 800 of the control device 200 is operated, The operation method of the model input 9 is learned using a learning algorithm such as reinforcement learning theory.

  The next step 2000 for determining whether the number of operations o has reached the maximum value is a branch. If the number of operations o is smaller than the maximum number of operations set in step 1000, 1 is added to o and then the process returns to step 1600. If o reaches the maximum number of operations, the process proceeds to step 2100.

  The next step 2100 for determining whether the learning count k has reached the maximum value is also a branch. If the number of learning k is smaller than the maximum number of learning set in step 1000, 1 is added to k and then the process returns to step 1400. If k has reached the maximum number of learning, the process proceeds to step 2200.

  In the last step 2200 of saving the learning result in the learning information database, the learning result of the operation method is saved in the learning information database 230, and the operation proceeds to the step of ending the operation method learning operation in the operation method learning unit 800.

  With the above operation, in model adjustment and learning of the operation method, the model parameter information included in the model construction data 7 is updated based on the model adjustment conditions set by the operator of the plant 100 and the learning conditions, and any model It is possible to autonomously acquire an operation method that reaches an input condition that allows a desired model output to be obtained from the input condition.

  Next, the operation of the model adjustment unit 600 in the control device 200 will be described in detail with reference to FIG. FIG. 3 is a diagram for explaining the operation of the model adjustment unit 600, and shows in detail the part including the model adjustment unit 600 and the model construction database 220 in the control device 200 shown in FIG.

  The model adjustment unit 600 includes a category calculation function unit 601 and a radius adjustment function unit 602. The category calculation function unit 601 uses the model construction data 7 stored in the model construction database 220, and is a parameter indicating the density of each model construction data and the coverage with respect to the input space of the basis arranged in each data. The category number determined as a result of classifying the data based on the density and the sparse density is calculated, and the model construction data 8 updated from them is stored in the model construction database 220.

  The radius adjustment function unit 602 uses the model construction data 7 updated by the category calculation function unit 601 to update the base radius information of each model construction data based on the sparse density and category information included therein, and updates The subsequent model construction data 8 is stored in the model construction database 220. The above operation corresponds to step 1100 for adjusting the radius of the statistical model in the flowchart of FIG.

  FIG. 4 shows an example of data stored in the model construction database 220.

  In the data stored in the model construction database 220 shown in FIG. 4, the data ID 221 is an identification number of each model construction data. The data coordinate 222 is coordinate information in the input space of the data, and at the same time means a central parameter of a statistical basis function arranged for the data.

  The base radius 223 is a parameter indicating the spread of the base arranged for each data. The sparse density 224 is a parameter indicating the coverage of the basis function in the data coordinates by the density distribution of the data and the basis arranged in the input space.

  The category number 225 is a parameter determined based on the sparse density, and data in which the sparse density 224 falls within a specific range is the same category and is given the same category number. The category number 225 is determined when the number of data stored in the model construction database 220 is updated, and is not updated until the model construction database 220 is updated next time.

  Hereinafter, a flowchart (FIG. 5) and a conceptual diagram (FIGS. 6 and 7) of an algorithm for model adjustment by the category calculation function unit 601 and the radius adjustment function unit 602 in the model adjustment unit 600 provided in the control device 200. ) And will be described.

  FIG. 5 is a flowchart showing the algorithm operation of model adjustment by the model adjustment unit 600, and corresponds to step 1100 for adjusting the radius of the statistical model in the flowchart of FIG.

  The flowchart shown in FIG. 5 is executed by combining steps 1101, 1102, 1103, 1104, 1105, 1106, 1107, 1108, 1109, 1110, 1111, and 1112. Hereinafter, each step will be described.

  After the model adjustment algorithm is started, first, in step 1101 for initializing the model adjustment condition and the base radius, the model adjustment unit 600 is operated, and the maximum number of iterations at the time of model adjustment, the number of categories, the sparse density determination threshold, And initialize various parameter values such as the base radius.

  Note that the base radii of all data are initialized to the same value larger than 0 for use in category calculation described later. The radius value is preferably set to a small value, for example, about 5 to 10% of the maximum distance of the model input space.

  Next, in step 1102 in which the number m of model adjustment iterations is initialized (m = 1), the iteration number m, which is the number of iterations in steps 1106 to 1111, is initialized (m = 1 is set).

Next, in step 1103 for calculating the sparse density of each data, the category calculation function unit 601 of the model adjustment unit 600 is operated to calculate the sparse density ρ _i of each model construction data according to the equation (1).

In Equation (1), i and j are subscripts of data, I is the total number of data, c _i and c _j are coordinate vectors of data i and j, respectively, and r _i is a base radius of data i. In addition, the exponential function of Σ in Equation (1) represents the basis function of the statistical model 500. As shown in Equation (1), the sparse density is the average value of all the basis function values with respect to the basis center of the data, that is, the coverage on the data coordinates by the basis function.

  That is, the sparse density is large for data in a region where the distance between data is small (data is dense), and small for data in a region where the distance between data is large (data is sparse). Become. By defining sparse density in this way, it is possible to handle the sparse distribution and coverage of data as a scalar quantity.

Next, in step 1104 for determining the category of each data, the category number to which each data belongs is determined based on the sparse density information of each data calculated in step 1103 for calculating the sparse density of each data. The category number is the number of categories set in step 1101 for initializing the model adjustment condition and the base radius, and the maximum value ρ ^max and the minimum value ρ ^min of the sparse density of each data.
It is determined based on the category boundary conditions [rho ^Ln that is calculated according to equation (2) using ^a. In Equation (2), N means the number of categories, and n means a subscript of the number of categories.

  A category number determination method based on the above definition will be described with reference to FIG.

FIG. 6 is a conceptual diagram showing a distribution of sparse density, where the horizontal axis represents model construction data ID and the vertical axis represents sparse density. As shown in FIG. 6, a range obtained by equally dividing the difference between the sparse density maximum value ρ ^max and the minimum value ρ ^min of each data by the number of categories N is an area of each category, and the area is calculated by Expression (2). It is defined by the category boundary condition ρ ^Ln . That is, the category n is classified data range from [rho ^{Ln + 1} from sparse is [rho ^Ln. (However, the sparse density range of category N is from ρ ^LN to ρ ^max .)
In the radius adjustment method described later, the value of the base radius of data having the same category number is adjusted to be uniform. When the number of categories is large, the number of parameters of the base radius to be adjusted increases, and it becomes possible to finely adjust the base radius, so a greater improvement in estimation accuracy can be expected, but the calculation cost increases accordingly. On the other hand, when the number of categories is small, the improvement in estimation accuracy is small, but the calculation cost is low. The plant operator can arbitrarily set the number of categories according to the needs regarding accuracy and time.

  Next, in step 1105 for storing sparse density / category information in the database, the model construction data 8 updated using the calculated sparse density and category number information is stored in the model construction database 220.

  In subsequent steps 1106 to 1111, the radius adjusting function unit 602 of the model adjusting unit 600 is operated to adjust the base radius of the data based on the determined category information. Here, a specific adjustment algorithm will be described with reference to FIG.

  FIG. 7 is a schematic diagram for explaining the adjustment of the base radius of data in the radius adjustment function unit 602 of the model adjustment unit 600. When the number of model inputs is 2, the model construction database 220 is displayed on the model input space. Is a plot of the model construction data 7 stored in FIG. Each model input has a range of [0, 1], and each data shows a concentric circle corresponding to the value of the base radius and a category number to which each data belongs.

As shown in the upper and lower diagrams of FIG. 7, the base radii of data having the same category number are equal. Adjustment of the base radius determines the model input x _m which first serves as a reference for the base radius adjustment on the model input space. Then, when the sparse density for the model input does not satisfy the threshold condition, data located in the nearest vicinity of the model input is extracted, and the base radius of the data belonging to the same category as the data is adjusted.

In the lower part of FIG. 7, the category number 2 of the data to the nearest neighbor to the model input x _m is selected, shows a case where the base radius of the data having the same category number is adjusted (enlarged). By repeating the above process a predetermined number of times, a base radius parameter capable of covering the model input space sufficiently and sufficiently can be obtained. Below, the detail of each step is demonstrated.

Next, in step 1106 determines the model input conditions, to determine the model input x _m as a reference when the base radius adjustment. As the model input condition, an arbitrary value within the operable range can be selected.

Next, in step 1107 to calculate a sparse for the model input, for the determined model input _{x m,} compute the sparse [rho _m of _{x m} according to equation (3).

Next, step 1108 for determining whether the sparse density satisfies the threshold condition is a branch. Determining whether to satisfy the calculated sparse [rho _m is the threshold conditions set in advance (within upper and lower limits). And for the model input x _m if satisfying the threshold condition, the base radius to be adjusted is regarded as not present, the process proceeds to step 1111, if not satisfied threshold condition is the step for adjusting the base radius Proceed to step 1109.

Next, in step 1109 to determine the category to which the latest neighbor data model input belongs (the nearest) whose distance is small relative to the model input x _m data, and the data to determine the category number belongs.

  Next, in step 1110 for adjusting the base radius belonging to the latest neighbor data, the category number having the same category number to which the nearest neighbor data determined in step 1109 for determining the category to which the latest neighbor data of the model input belongs is assigned. Adjust the base radius for the data.

The base radius is adjusted by increasing the base radius to increase the sparse density by expanding the area covered by the base when the sparse density ρ _m is below the lower limit of the threshold condition, and adjusting the base radius when the base radius is above the upper limit. In order to reduce the sparse density by reducing the area covered by, this is executed according to the policy of reducing the base radius.

  In this way, by adjusting the base radius so that the sparse density for an arbitrary model input condition satisfies the threshold, overestimation of the estimated value due to base competition in the dense region, and the base coverage is sufficient in the sparse region It is possible to reduce the influence of underestimation due to the fact that it cannot be secured, and to improve the model estimation accuracy.

  Next, step 1111 for determining whether m has reached the maximum value is a branch. If the number of iterations m is smaller than the maximum number of iterations set in step 1101 for initializing the model adjustment condition and the base radius, the process returns to step 1106 for determining the model input condition after adding 1 to m, and m is the maximum operation. If the number has been reached, the process proceeds to step 1112.

  Finally, in step 1112 for storing the base radius information in the database, the model building data 8 is updated using the adjusted base radius information and stored in the model building database 220, and the operation of the model adjusting unit 600 is terminated. move on.

  The effects of the model adjustment algorithm performed by the model adjustment unit 600 of the control device 200 described above will be described with reference to FIGS.

  FIG. 8 is a contour diagram of the sparse density distribution in the model input space and data structure similar to that shown in FIG. 7, and the upper diagram of FIG. 8 shows the case where the base radius is set to a uniform value (0.1). FIG. 8 is a contour diagram after radius adjustment by the model adjustment algorithm.

  FIG. 9 is a contour diagram of the model output distribution for the base radius values of the two cases shown in FIG. 8 and 9, the concentric circle of the base radius and the category number of each data are shown as in FIG.

  As shown in the upper figure of FIGS. 8 and 9, when the basis radius is set uniformly, the basis function cannot sufficiently cover the input space, so both the sparse density and the model output are only inside the concentric circle of the basis radius. The value is larger.

  As a result, the estimation accuracy in a region where the data distribution is sparse decreases, and even when the operation method learning unit 800 of the control device 200 performs learning, the desired control performance may not be obtained.

  On the other hand, when the base radius is adjusted by the model adjustment unit 600 of the control device 200, the input space can be sufficiently covered by the basis function, so that the coverage of the base is improved as shown in FIGS. As a result, the sparse density increases almost in the entire input space. The model output value becomes a value corresponding to the data characteristic even in a region where the data distribution is sparse, and the estimation accuracy is improved.

  As is clear from the above description, the model adjustment unit 600 of the control device 200 adjusts the base radius so that the input space can be covered sufficiently and sufficiently based on the sparse density information of the model construction data. As a result, it is possible to eliminate the region where the base is excessively covered and the region where the coverage is reduced, thereby contributing to improvement of the model estimation accuracy.

  Further, since the base radius is adjusted according to a certain policy with the sparseness as a clue, trial and error search processing can be eliminated, and the calculation cost can be reduced as compared with the trial and error radius adjusting means. This is the end of the detailed operation of the model adjustment function 600.

Next, in the plant control apparatus of the first embodiment, the maintenance tool output signal 94 transmitted from the external output interface 913 of the maintenance tool 910 capable of transmitting and receiving data to and from the control apparatus 200 is displayed on the image display device 920. The displayed screen will be described with reference to FIGS. 10 to 12 are specific examples of screens displayed on the image display device 920.

  FIG. 10 is a screen example displayed on the image display device when setting the model input / output in the plant control apparatus according to the first embodiment, and shows a control procedure in the plant control apparatus according to the first embodiment. It is an example of the model construction condition setting screen of step 1000 which sets the model construction condition / learning condition in the flowchart of FIG.

  In the model input / output setting screen shown in FIG. 10, model input / output of the statistical model 500 in the control device 200 can be set by selecting an arbitrary item from the input / output items based on the measurement data information of the plant. .

  With the screen shown in FIG. 10 displayed on the image display device 920, the mouse 902 of the external input device 900 is operated to move the focus to the numerical box on the screen, and a numerical value can be input by using the keyboard 901. Further, by operating the mouse 902 and clicking a button on the screen, the button can be selected (pressed). Similarly, a check can be made by operating the mouse 902 and clicking a check box on the screen.

  In the screen shown in FIG. 10, first, in the model input setting, the selection bar 3001 is selected by moving the focus of the mouse 902 to an arbitrary item with respect to the input item displayed in the input item list 3000 and selecting a button. Can be matched with the input item. Then, by selecting a button 3002, the selected input item can be added to the model input item list 3003.

  Furthermore, the minimum value and the maximum value can be set for the added model input item by shifting the focus to the numerical boxes 3004 and 3005 and inputting the numerical values. When deleting an item from the already added model input item list, the item to be deleted can be deleted from the list by selecting the item to be deleted with the mouse 902 and selecting the button 3006.

  In the screen shown in FIG. 10, next, in the model output setting, the focus of the mouse 902 is moved to an arbitrary item with respect to the output item similarly displayed in the output item list 3007, and a selection bar is selected by selecting a button. 3008 can be matched with the selected output item. Then, by selecting a button 3009, the selected output item can be added to the model output item list 3010.

  When deleting an item from the already added model output item list, the item to be deleted can be deleted from the list by selecting the item to be deleted with the mouse 902 and selecting the button 3011.

  After the above model input / output setting is completed, when a button 3012 is selected, the screen moves to a model adjustment condition setting screen shown in FIG.

  FIG. 11 is a screen example displayed on the image display device when setting the model adjustment conditions in the plant control apparatus according to the first embodiment. In the model adjustment condition setting screen shown in FIG. In step 1104, the model adjustment iteration number which is the maximum value of the model adjustment iteration number m in the flowchart of FIG. 5 is obtained by shifting the focus to each of the boxes 3100, 3101, 3102, 3103, and 3104. The number of categories used, the minimum and maximum values of the sparse density threshold used in step 1108, and the estimated error target value used for evaluating the model adjustment result can be determined.

  After the above model adjustment condition setting is completed, the model adjustment can be started by selecting a button 3105. When the button 3106 is selected, the screen returns to the model input / output setting screen.

  FIG. 12 is a screen displayed on the image display device when confirming the model adjustment result in the plant control apparatus according to the first embodiment, and is a diagram showing a control procedure in the plant control apparatus according to the first embodiment. FIG. 6 is an example of a screen used when evaluating an adjustment result after completion of model adjustment in step 1100 in the flowchart of FIG.

  In the model adjustment result display screen shown in FIG. 12, as a sparse density average transition display screen 3200, a sparse density transition with respect to the number m of model adjustment iterations is displayed as a graph 3201.

Here, the horizontal axis of the graph is the number m of model adjustment iterations, and the vertical axis is the moving average of the sparse density ρ _m at each iteration number. The sparse density average transition display screen 3200 displays the minimum value 3202 and the maximum value 3203 of the sparse density threshold set on the model adjustment condition setting screen of FIG.

  In the model adjustment result display screen shown in FIG. 12, the model estimation error transition display screen 3204 displays a transition of the model estimation error with respect to the number m of model adjustment iterations as a graph 3205. Here, the horizontal axis of the graph is the number m of model adjustment iterations, and the vertical axis is the error between the model evaluation data and the model estimation value at each iteration number. Further, on this model estimated error transition display screen 3204, the estimated error target value 3206 set on the model adjustment condition setting screen of FIG. 11 is displayed.

  The plant operator determines whether or not the model adjustment is properly performed while viewing the model adjustment results displayed on the sparse density average transition display screen 3200 and the model estimation error transition display screen 3204 shown in FIG. can do.

In the sparse average transition display screen 3200, sparse 3201 in the model adjustment late If the has remained between the minimum value 3202 and a maximum value 3203 of sparse threshold, and in the model estimation error transition display screen 3204, If the model estimation error 3205 at the end of the model adjustment is below the estimated error target value 3206, the model adjustment can be ended by selecting the button 3207 on the assumption that a desired model adjustment result has been obtained.

  On the other hand, if the model adjustment result does not satisfy any of the above conditions, selecting the button 3208 returns to the model adjustment condition setting screen of FIG. 11 and the model adjustment can be executed again.

  As described above, the plant control apparatus according to the first embodiment has a function capable of determining whether or not to end the model adjustment according to the information displayed on the model adjustment result display screen shown in FIG. The model adjustment can be repeated until the statistical model performance is obtained. As a result, it is possible to construct a robust statistical model that guarantees a certain level of estimation accuracy regardless of the model construction data configuration.

  Further, by setting the number of iterations of model adjustment to be small in advance, it is possible to reduce redundant calculation costs in model adjustment, and therefore it is possible to shorten the time required for model adjustment compared to the conventional assumption. Therefore, the frequency | count of operating a plant can be increased and the higher control effect can be acquired.

  In the plant control apparatus of the above-described embodiment, the estimation accuracy of the statistical model can be improved by using the radius parameter appropriately adjusted by the model adjustment unit. In addition, the adjustment of the radius parameter does not require an iterative process for evaluating the error between the data and the estimated value, so that the calculation cost can be reduced and the adjustment can be completed within the control cycle.

  Above, description about the screen displayed on the image display apparatus 920 in the control apparatus of the plant which is 1st Example is complete | finished.

  According to this embodiment, even when controlling a learning type plant in which there is a bias in the data used when constructing the statistical model, the parameter of the statistical model is appropriately set within the control cycle according to the distribution of the bias in the data. It is possible to realize a plant control apparatus having a function of improving the estimation accuracy by adjusting in the above.

  Next, a control device for a thermal power plant that is a second embodiment in which the control device 200 according to the present invention is applied to a thermal power plant will be described.

  Needless to say, the control device 200 according to the present invention can also be used when controlling a plant other than the thermal power plant.

  FIG. 13 is a schematic diagram showing the configuration of a thermal power plant 100a to which the control device 200 according to the present invention is applied. First, the mechanism of power generation by the thermal power plant 100a will be briefly described.

  In FIG. 13, pulverized coal, which is fuel obtained by finely pulverizing coal with a mill 110, primary air for transporting pulverized coal, and secondary air for combustion adjustment is supplied to a boiler 101 constituting the thermal power plant 100a. A plurality of burners 102 are provided, and pulverized coal supplied through the burners 102 is burned in 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, respectively.

  Further, the boiler 101 is provided with an after-air port 103 through which air for two-stage combustion is introduced into the boiler 101. The air for two-stage combustion is guided from the pipe 142 to the after air port 103.

  The high-temperature combustion gas generated by burning pulverized coal inside the boiler 101 flows downstream along the path inside the boiler 101 and is supplied to the heat exchanger 106 disposed inside the boiler 101. After generating steam by exchanging heat with the air, it becomes exhaust gas and flows into the air heater 104 installed on the downstream side of the boiler 101. The air supplied to the boiler 101 is heated by this air heater 104 and heated. To do.

  And the exhaust gas which passed this air heater 104 is discharged | emitted from the chimney to air | atmosphere after performing the exhaust gas process which is not shown in figure.

  The feed water circulating through the heat exchanger 106 of the boiler 101 is supplied 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 to become high-temperature and high-pressure steam. In this embodiment, the number of heat exchangers is 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 107, and the steam turbine 108 is driven by the energy of the steam to generate power by the generator 109.

  In the thermal power plant 100a of the second embodiment, various measuring devices that detect state quantities indicating the operating state of the thermal power plant are arranged.

  Since the thermal power plant 100a corresponds to the plant 100 in FIG. 1, the measurement signal of the thermal power plant acquired from these measuring instruments is the control device 200 as the measurement signal 1 from the plant 100 as shown in FIG. To the external input interface 201.

As a measuring instrument, for example, as shown in a thermal power plant 100a in FIG. 13, a temperature measuring instrument 151 that measures the temperature of high-temperature and high-pressure steam supplied from the heat exchanger 106 to the steam turbine 108, and measures the pressure of the steam. A pressure measuring device 152 and a power generation output measuring device 153 that measures the amount of power generated by the power generator 109 are illustrated.

  The feed water generated by cooling the steam by the condenser (not shown) of the steam turbine 108 is supplied to the heat exchanger 106 of the boiler 101 by the feed water pump 105. The flow rate of this feed water is measured by the flow meter 150. It is measured.

In addition, the state quantity relating to the concentration of components (nitrogen oxide (NOx), carbon monoxide (CO), hydrogen sulfide (H ₂ S), etc.) contained in the exhaust gas that is the combustion gas discharged from the boiler 101 The measurement signal is measured by a concentration measuring device 154 provided on the downstream side of the boiler 101.

  That is, in the thermal power plant control apparatus according to the second embodiment in which the control device 200 of the present invention is applied to the thermal power plant 100a, each measurement data item of the thermal power plant 100a measured by the measuring instrument includes each of the above measurements. The fuel flow rate supplied to the boiler 101, which is the state quantity of the thermal power plant 100a measured by the boiler, the air flow rate supplied to the boiler 101, the feed water flow rate supplied to the heat exchanger 106 of the boiler 101, and the heat exchange of the boiler 101 Steam temperature generated in the steam generator 106 and supplied to the steam turbine 108, feed water pressure supplied to the heat exchanger 106 of the boiler 101, gas temperature of the exhaust gas discharged from the boiler 101, gas concentration of the exhaust gas, and An exhaust gas recirculation flow rate for recirculating a part of the exhaust gas discharged from the boiler 101 to the boiler 101 is included.

  These measurement data items are measurement data items determined by the control signal 15 calculated and output by the control signal generation unit 700 in the control device 200 shown in FIG.

  In general, many measuring instruments other than those shown in FIG. 13 are arranged in the thermal power plant 100a, but the illustration is omitted here.

  Next, a path of air that is introduced into the boiler 101, that is, a path of primary air and secondary air that is introduced into the boiler 101 from the burner 102, and an interior of the boiler 101 from the after air port 103. The air path will be described with reference to FIG.

  In the boiler 101 shown in FIG. 13, the primary air is guided from the fan 120 to the pipe 130, and passes through the air heater 104 installed on the downstream side of the boiler 101 and the air heater 104. However, it is branched to the bypass pipe 131 but is joined again as the pipe 133 disposed on the downstream side of the air heater 104 and led to the mill 110 for producing pulverized coal installed on the upstream side of the burner 102. .

  The primary air passing through the air heater 104 is heated by exchanging heat with the combustion gas flowing down the boiler 101. The primary air that bypasses the air heater 104 together with the heated primary air conveys the differential coal crushed in the mill 110 to the burner 102.

  The air introduced from the pipe 140 using the fan 121 is heated in the same manner by the air heater 104 and then branched into a secondary air pipe 141 and an after-air port pipe 142, respectively. To the burner 102 and the after-air port 103.

  In the control apparatus for a thermal power plant according to the second embodiment, as an example of controlling the flow rate of air sent from the fan 121 and introduced into the boiler 101 from the burner 102 and the after air port 103, piping for secondary air 141 and an after-airport piping 142 are provided upstream of an air damper 162 and an air damper 163 as operation end devices, respectively, and the opening degree of these air damper 162 and air damper 163 is adjusted by the control device 200 to be installed inside the boiler 101. The flow rate of the supplied secondary air and after air can be controlled respectively.

  In addition, as an example of controlling the flow rate of air sent from the fan 120 and introduced into the boiler 101 together with the pulverized coal from the burner 102, an air damper serving as an operation end device is connected to the pipe 131 and the pipe 132 immediately before joining the pipe 133. 160 and an air damper 161 are provided, respectively, and the opening degree of the air damper 160 and the air damper 161 is adjusted by the control device 200 so that the flow rate of the air supplied into the boiler 101 can be controlled.

  Since the control device 200 can also control other measurement data items, the installation location of the operation terminal device may be changed according to the control target.

  FIG. 14 is an enlarged view of a piping section related to the air heater 104 installed on the downstream side of the boiler 101 of the thermal power plant 100a shown in FIG.

  As shown in FIG. 14, the air heater 104 is provided with a pipe 130 for supplying air and a pipe 140, and among these, the pipe 140 is disposed through the air heater 104, and the pipe 130 is The pipe 131 and the pipe 132 are branched from the middle, and the pipe 131 is arranged to bypass the air heater 104, and the pipe 132 is arranged to penetrate the air heater 104.

  The pipe 132 passes through the air heater 104 and then becomes a pipe 133 joined with the pipe 131 and is led to the mill 110, and is arranged so as to guide air along with the pulverized coal from the mill 110 to the burner 102 of the boiler 101. It is installed.

  In addition, the pipe 140 passes through the air heater 104 and then branches into a pipe 141 and a pipe 142. Of these, the pipe 141 guides air to the burner 102 of the boiler 101, and the pipe 142 leads to the after-air port 103 of the boiler 101, respectively. It is arranged like this.

  Also, an air damper 160 and an air damper 161 for adjusting the amount of air flowing are installed in the pipe 131 and the pipe 132 immediately before joining the pipe 133, respectively, and the upstream part of the pipe 141 and the pipe 142 flows. An air damper 162 and an air damper 163 for adjusting the amount of air are respectively installed.

  Then, by operating these air dampers 160 to 163, the area through which air passes through the pipes 131, 132, 141, 142 can be changed, so that the boiler 101 passes through the pipes 131, 132, 141, 142. The flow rate of air supplied to the interior of each can be individually adjusted.

  The control signal 15 calculated by the control signal generator 700 of the control device 200 is output as the operation signal 16 for the thermal power plant 100a via the external output interface 202, and installed in the pipes 131, 132, 141, 142 of the boiler 101, respectively. The control end devices such as the air dampers 160, 161, 162, and 163 are operated.

  In this embodiment, the devices such as the air dampers 160, 161, 162, 163 are called operation ends, and the control signal 15 calculated by the control device 200 necessary for operating the devices is sent from the control device 200 to the above-mentioned control device 200. An output signal commanded to the operation end is referred to as an operation signal 16.

  The operation signal 16 calculated by the control signal generator 700 and output to the operation end includes an air flow rate supplied to the boiler 101 through the pipes 131, 132, 141, and 142, and a pipe for supplying air to the boiler 101. 131, 132, 141, 142, the air dampers 160 to 163 that adjust the flow rate of air, the fuel flow rate of pulverized coal supplied to the burner 102 of the boiler 101, and the exhaust gas discharged from the boiler 101 For example, an exhaust gas recirculation flow rate for recirculating a part of the exhaust gas to the boiler 101 is included.

Hereinafter, the air dampers 160, 161, respectively installed in the pipes 131, 132 for adjusting the amount of air supplied to the burner 102 whose operation end is installed in the boiler 101 by applying the control device of the present invention to the thermal power plant 100a. As the air dampers 162 and 163 installed in the pipes 141 and 142 for adjusting the amount of air supplied to the after air port 103 installed in the boiler 101, CO, NOx in the exhaust gas discharged from the boiler 101, and A case where the concentration of H ₂ S is used will be described.

In this embodiment, the operation amount of the operation end of the boiler 101 (the opening degree of the air dampers 160, 161, 162, 163) is the exhaust gas discharged from the boiler 101 as the model input of the statistical model 500 constituting the control device 200. The NOx, CO, and H ₂ S concentrations contained in the model become the model output of the statistical model 500, and the minimization of each model input / output is the learning objective.

  FIG. 15 shows an example of a screen displayed on the image display device 920 when used in the control device 200 of the thermal power plant 100a in the thermal power plant control apparatus according to the second embodiment. Model input / output setting of the statistical model 500 constituting the control device 200 displayed on the image display device, corresponding to FIG. 10 showing an example of the screen displayed when setting the model input / output in the plant control device. It is an example of a screen.

  In the model input / output setting screen example shown in FIG. 15, the model of the statistical model 500 in the control device 200 while grasping the positional relationship with respect to each of the burner 102 and the after-air port 103 which are the operation ends of the boiler 101. Input / output can be set.

  Specifically, in the model input setting screen, the installation position of the item selected by the selection bar 3503 with respect to the input item displayed in the model input item list 3502 displayed on the boiler operation end display screen 3500 is set. A symbol on the boiler diagram before the can displayed on the boiler operation end display screen 3500 shown is indicated by a pointer 3501.

Contrary to this operation, the selection bar is displayed by clicking the mouse 902 of the external input device 900 on the symbol of the specific operation terminal displayed on the boiler operation terminal display screen 3500 and focusing the pointer 3501. It is also possible to move the display position 3503 (select an input item). Then, by selecting a button 3504, the selected input item can be added to the model input item list 3502 .

In FIG. 15, the functions of the numerical value box 3506, the output item selection bar 3509, the screens 3508 and 3511, the buttons 3510, 3512 and 3513, and the button 3507 on the model output setting screen are the same as those in the screen of FIG. It is the same.

  In the air amount control for controlling the air supplied to the boiler 101 by the control device of the thermal power plant of the second embodiment, a priori knowledge exists regarding the adjustment method of the air damper of the specific burner and after-airport, and in many cases Control based on that is executed.

  Therefore, by setting the model input setting in the control device 200 of the present embodiment to the screen configuration as shown in FIG. 15, the plant operator can confirm the position of the operation end of the boiler 101 and perform the prior experiment. The model input / output of the statistical model 500 in the control device 200 can be appropriately selected in consideration of the control method based on knowledge.

  In addition, since it is possible to understand the operation end and minimum / maximum values set using the screen shown in FIG. 15 in association with the plant design information, the model input setting can be made more efficient and setting errors can be reduced. Can contribute.

As described above, if the plant control apparatus 200 of the present invention is applied to a thermal power plant, NOx and CO discharged from the thermal power plant are learned by learning an operation method that satisfies the requirements for environmental regulations and operation costs. , And target values for H ₂ S concentration can be achieved.

  According to this embodiment, even when controlling a learning type plant in which there is a bias in the data used when constructing the statistical model, the parameter of the statistical model is appropriately set within the control cycle according to the distribution of the bias in the data. Thus, it is possible to realize a thermal power plant control apparatus having a function of improving estimation accuracy.

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

  1: Measurement signal, 16: Control signal, 90: Input / output data information, 100: Plant, 100a: Thermal power plant, 101: Boiler, 102: Burner, 103: After airport, 130-133: Piping, 140-142: Piping, 160 to 163: Air damper, 200: Control device, 201: External input interface, 202: External output interface, 210: Measurement signal database, 220: Model construction database, 230: Learning information database, 240: Control logic database, 250 : Control signal database, 300: measurement signal conversion unit, 400: numerical analysis unit, 500: statistical model, 600: model adjustment unit, 700: control signal generation unit, 800: operation method learning unit, 900: external input device, 901 : Keyboard, 902: Mouse, 9 0: maintenance tools, 911: external input interface, 912: data transmission and reception processing unit, 913: external output interface, 920: image display device.

Claims (11)

In a plant control device including a control device that takes in a measurement signal that is a state quantity of the plant from the plant and calculates an operation signal that controls the plant using the measurement signal,
The control device includes a measurement signal database that captures and stores a measurement signal that is a state quantity of the plant, a model construction database that stores model construction data converted from measurement data of the plant stored in the measurement signal database, and A statistical model for simulating plant control characteristics for estimating a value of a measurement signal that is a state quantity of the plant when a control signal is given to the plant using model building data stored in the model building database; An operation method learning unit that learns how to generate a model input corresponding to the control signal given to the plant so that a model output corresponding to the measurement signal achieves a target value using a model, and learning in the operation method learning unit Learning information database for storing learning information data on constraints and learning results , The measurement signal of the measurement signal database, and configured by a said learning information control signal generation unit for calculating a control signal to be transmitted to the plant by using the learning information data in the database,
The model construction data stored in the model construction database is configured to include sparse density information indicating the density of the data,
Further, the control device is provided with a model adjustment unit that adjusts a base radius parameter of a statistical model included in model construction data stored in the model construction database, and the statistical model adjusts a base radius parameter by the model adjustment unit. A plant control apparatus configured to generate a model output using a result. The plant control apparatus according to claim 1,
The model adjustment unit includes a category calculation function unit that determines a category number of the model building data using model building data stored in the model building database, and a model including category information determined by the category calculation function unit A plant control apparatus comprising at least one of a radius adjustment function unit that adjusts a radius parameter of the statistical model using construction data. The plant control apparatus according to claim 2,
The category calculation function unit is configured to calculate a sparse density distribution range of the model construction data by the number of categories based on a function of calculating a sparse density indicating a density of each data and a category number information input from an external input device. A plant control apparatus comprising at least one of functions for determining a category number of each data on the basis of a divided value . The plant control apparatus according to claim 2,
The radius adjustment function unit calculates a sparse density with respect to a reference model input arbitrarily determined in the model input space and adjusts the radius parameter, and the calculated sparse density is input from an external input device. When the threshold condition is not satisfied, it is provided with at least one of functions for extracting a category of data located nearest to the reference model input and adjusting a radius parameter of data belonging to the category. Plant control equipment . The plant control apparatus according to claim 1,
The control device is connected to an image display device, the image display device has a function of displaying model construction data stored in the model construction database, and a function of displaying a result of adjustment of a statistical model by the model adjustment unit, A plant control apparatus comprising at least one of functions for setting model adjustment conditions used in the model adjustment unit . In a control device for a thermal power plant comprising a control device that takes in a measurement signal that is a state quantity of the plant from a thermal power plant including a boiler and calculates an operation signal for controlling the thermal power plant using the measurement signal ,
The measurement signal includes a state quantity signal representing at least one of the concentrations of nitrogen oxide, carbon monoxide, and hydrogen sulfide contained in the exhaust gas discharged from the boiler of the thermal power plant,
The operation signal includes an air flow rate supplied to the boiler of the thermal power plant, an opening degree of an air damper that adjusts the air flow rate, a fuel flow rate supplied to the boiler, and an exhaust gas discharged from the boiler is recirculated to the boiler. A signal representing at least one of the exhaust gas recirculation flow rates,
The control device includes a measurement signal database that captures and stores a measurement signal that is a state quantity of the thermal power plant, an air flow rate that is supplied to the boiler converted from the measurement data of the plant stored in the measurement signal database, and the air Model construction for storing model construction data including at least one of the opening degree of the air damper for adjusting the flow rate, the fuel flow rate supplied to the boiler, and the exhaust gas recirculation flow rate for recirculating exhaust gas discharged from the boiler to the boiler A statistical model for simulating plant control characteristics for estimating a value of a measurement signal that is a state quantity of the plant when a control signal is given to the plant using model construction data stored in the model construction database; Using the statistical model, the model output corresponding to the measurement signal achieves the target value. An operation method learning unit that learns a generation method of a model input corresponding to the control signal given to the plant, a learning information database that stores learning information data related to learning constraint conditions and learning results in the operation method learning unit, and Comprising a measurement signal in the measurement signal database, and a control signal generation unit that calculates a control signal transmitted to the plant using the learning information data in the learning information database;
The model construction data stored in the model construction database is configured to include sparse density information indicating the density of the data,
Further, the control device is provided with a model adjustment unit that adjusts a base radius parameter of a statistical model included in model construction data stored in the model construction database, and the statistical model adjusts a base radius parameter by the model adjustment unit. A control apparatus for a thermal power plant, characterized in that a model output is generated using a result . In the thermal power plant control apparatus according to claim 6,
The model building data stored in the model building database includes the air flow rate supplied to the boiler of the thermal power plant, the opening degree of the air damper that adjusts the air flow rate, the fuel flow rate supplied to the boiler, and the exhaust gas discharged from the boiler Including information related to at least one of the exhaust gas recirculation flow rates for recirculating the exhaust gas to the boiler, the model adjusting unit using the model construction data stored in the model construction database, the category number of the model construction data At least one of a category calculation function unit that determines a radius parameter of the statistical model using model construction data including category information determined by the category calculation function unit. A control device for a thermal power plant . The control device for a thermal power plant according to claim 7,
The category calculation function unit is configured to calculate a sparse density distribution range of the model construction data by the number of categories based on a function of calculating a sparse density indicating a density of each data and a category number information input from an external input device. A control apparatus for a thermal power plant, comprising at least one of functions for determining a category number of each data on the basis of the divided values . The control device for a thermal power plant according to claim 7,
The radius adjustment function unit calculates a sparse density with respect to a reference model input arbitrarily determined in the model input space and adjusts the radius parameter, and the calculated sparse density is input from an external input device. When the threshold condition is not satisfied, it is provided with at least one of functions for extracting a category of data located nearest to the reference model input and adjusting a radius parameter of data belonging to the category. Control device for thermal power plant . In the thermal power plant control apparatus according to claim 6,
The control device is connected to an image display device, the image display device has a function of displaying model construction data stored in the model construction database, and a function of displaying a result of adjustment of a statistical model by the model adjustment unit, A control apparatus for a thermal power plant, comprising at least one of functions for setting model adjustment conditions used in the model adjustment unit . In a thermal power plant equipped with a control device that takes in a measurement signal that is a state quantity of the plant from the plant and calculates an operation signal for controlling the plant using the measurement signal,
The thermal power plant comprises a boiler that generates steam, a steam turbine that is driven by guiding steam generated in the boiler, and a generator that generates power by driving the steam turbine,
The control device includes a measurement signal database that captures and stores a measurement signal that is a state quantity of the plant, a model construction database that stores model construction data converted from measurement data of the plant stored in the measurement signal database, and A statistical model for simulating plant control characteristics for estimating a value of a measurement signal that is a state quantity of the plant when a control signal is given to the plant using model building data stored in the model building database; An operation method learning unit that learns how to generate a model input corresponding to the control signal given to the plant so that a model output corresponding to the measurement signal achieves a target value using a model, and learning in the operation method learning unit Learning information database for storing learning information data on constraints and learning results , The measurement signal of the measurement signal database, and configured by a said learning information control signal generation unit for calculating a control signal to be transmitted to the plant by using the learning information data in the database,
The model construction data stored in the model construction database is configured to include sparse density information indicating the density of the data,
Further, the control device is provided with a model adjustment unit that adjusts a base radius parameter of a statistical model included in model construction data stored in the model construction database, and the statistical model adjusts a base radius parameter by the model adjustment unit. A thermal power plant configured to generate a model output using the result .

Images