JP2009282750A - Apparatus for processing plant data, and method for processing plant data - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an apparatus for processing plant data, which is improved in grasping the knowledge of an accurate plant state of a plant to be measured, and also improved in plant control performance. <P>SOLUTION: The apparatus for processing plant data comprises: an external input interface for capturing plant measurement signals measuring state quantities of a measured plant into the data processing apparatus; a measurement signal database for storing the measured plant measurement signals; a model for simulating plant characteristics of the measured plant with the input of model inputs or the measured plant measurement signals; a learning part for learning parameters given to the model such that model outputs simulated by the model attain model output target values; a processing part for computing processed signals indicating the plant measurement signals with noise and measuring device drift removed according to the results of learning by the learning part; and an external output interface for outputting the processed signals computed by the processing part from the data processing apparatus. <P>COPYRIGHT: (C)2010,JPO&INPIT

Description

  The present invention relates to a plant data processing apparatus and a plant data processing method.

As represented by the power generation plant, various plant data are measured from the plant in order to grasp the operation of the plant and the state of the plant. The data of these plants is in a time series format, and the amount of data increases according to the scale of the plant.

The measured plant data includes not only noise superimposed between the measurement position and the data storage device, but also drift components of the measuring instrument due to plant operation.
Therefore, in order to grasp the true state quantities of these plants, it is necessary to exclude components such as noise and instrument drift from the collected plant data.

The conventional technology for excluding components such as noise and instrument drift from the measured plant data is a method of separating the plant data into frequency components by Fourier transform and excluding unnecessary signal frequency components with a filter. There is a method of grasping the influence of noise by creating a time series model of target data and adding a noise term assumed by Gaussian distribution.

For example, the former, in Japanese Patent Laid-Open No. 6-43238, obtains an amplitude spectrum after fast Fourier transforming a received signal even if the target radiated sound and noise are received at the same level for a passive sonar system. A method for discriminating whether a desired radiated sound or noise is input by inputting into a neural network is described.

The latter is described in, for example, the technical literature “System Identification”, Seto Sagara, Kageo Akiyama, Takayoshi Nakamizo, and Toru Katayama (Corona Inc., published on August 20, 1997), autoregressive model, moving average model, autoregressive A method for adding a noise term to each moving average model, a Gopinath-Budin method for realizing a probability model from input / output data with a white noise sequence added, and the like are described.

JP-A-6-43238 System Identification, Seto Sagara, Akio Akiyama, Takayoshi Nakamizo, Toru Katayama, Corona, August 20, 1997.

The technique described in Japanese Patent Application Laid-Open No. 6-43238 is a process after the received signal is converted into the frequency domain, and the reception target depends greatly on the frequency of sound and clearly gives a boundary frequency with noise. Only if you can.

However, since plant data such as power generation plants measure the operation and state of the plant, the plant data has various frequency bands, and it is difficult to uniquely determine the boundary frequency with noise. .

In addition, the technology described in the technical document “System Identification”, Seto Sagara, Kageo Akiyama, Takayoshi Nakamizo, and Toru Katayama (Corona Inc., issued on August 20, 1997), the plant data is converted into a time series model or Must be defined by a probabilistic model. This probability model is a model that is linear or linearly approximated.

However, since the number of data increases according to the scale and the nonlinearity of the plant data is strong, it is difficult to apply the method based on the linearity data shown in Non-Patent Document 1. It is.

In addition, both of the known techniques described above only remove noise in any case,
It is not considered to remove the drift component of the measuring instrument at the same time.

The object of the present invention is to simultaneously remove the noise contained in the measured plant data and the drift component of the measuring instrument to enable accurate grasp of the plant state and to improve the plant control performance. An object of the present invention is to provide a data processing method for an apparatus and a plant.

A plant data processing apparatus for obtaining a measurement signal obtained by removing noise and drift of a measuring instrument based on a measurement signal obtained by measuring a state quantity of a measurement target plant according to the present invention is a plant that measures a state quantity of a measurement target plant. An external input interface for importing the measurement signal of the plant into the data processing device, a measurement signal database for storing the measurement signal of the measured plant, and a plant characteristic of the plant to be measured by inputting the model input or the measured measurement signal of the plant A learning unit that learns parameters to be given to the model so that the model output simulated by the model achieves a target value of the model output, and a measurement signal of the plant based on a learning result by the learning unit A processing unit that calculates a processing signal that removes noise and instrument drift from the The processed signal calculated in the processing unit, characterized in that it includes an external output interface for outputting from the data processing apparatus to the outside.

Moreover, the data processing apparatus of the plant which obtains the measurement signal from which the noise and the drift of the measuring instrument are removed based on the measurement signal obtained by measuring the state quantity of the measurement target plant according to the present invention measures the state quantity of the measurement target plant. An external input interface that imports the measured signal of the plant into the data processing device, a measurement signal database that stores the measured signal of the measured plant, a model creation unit that creates a model using the measured signal of the plant and model parameters, A model that simulates the plant characteristics of the plant to be measured using the model input or the measured plant measurement signal as input, and parameters that are given to the model so that the model output simulated by the model achieves the target value of the model output And a plan based on a learning result by the learning unit A processing unit that calculates a processing signal that removes noise and instrument drift from the measurement signal, and an external output interface that outputs the processing signal calculated by the processing unit to the outside from the data processing device. Features.

Moreover, the data processing apparatus of the plant which obtains the measurement signal from which the noise and the drift of the measuring instrument are removed based on the measurement signal obtained by measuring the state quantity of the measurement target plant according to the present invention measures the state quantity of the measurement target plant. An external input interface for capturing the measured signal of the plant into the data processing device, a measurement signal database for storing the measured signal of the plant,
A model creation unit that creates a model using plant measurement signals and model parameters, a model that simulates plant characteristics of the plant to be measured using model inputs or measured plant measurement signals as inputs, and the model A learning unit that learns parameters to be given to the model so that the model output achieved achieves a target value of the model output, a model parameter database that stores model parameters learned by the learning unit, and a learning result by the learning unit A processing unit that calculates a processing signal that removes noise and instrument drift from the measurement signal of the plant, and an external output interface that outputs the processing signal calculated by the processing unit to the outside from the data processing device. It is characterized by.

A plant data processing method for obtaining a measurement signal obtained by removing noise and drift of a measuring instrument based on a measurement signal obtained by measuring a state quantity of a measurement target plant according to the present invention is a plant that measures a state quantity of a measurement target plant. The measurement signal is stored in the measurement signal database, the model input using the model or the plant measurement signal stored in the measurement signal database is used as an input to simulate the plant characteristics of the plant to be measured, and the model is simulated Learn the parameters given to the model so that the model output achieves the target value of the model output, and calculate and output a processing signal that removes noise and instrument drift from the measured signal of the plant based on the learning result It was made to do.

The plant data processing method for obtaining a measurement signal obtained by removing noise and drift of the measuring instrument based on the measurement signal obtained by measuring the state quantity of the plant to be measured according to the present invention measures the state quantity of the measurement target plant. The measured signal of the plant is stored in the measured signal database, a model is created using the measured signal of the plant and the model parameter stored in the measured signal database, and the model is input using this model or stored in the measured signal database. The plant measurement signal of the plant to be measured is simulated as an input, and the parameters given to the model are learned so that the model output simulated by the model achieves the target value of the model output. Noise and instrument drift from the plant measurement signal based on learning results And it calculates the processed signal to remove characterized by being adapted to output.

The plant data processing method for obtaining a measurement signal obtained by removing noise and drift of the measuring instrument based on the measurement signal obtained by measuring the state quantity of the plant to be measured according to the present invention measures the state quantity of the measurement target plant. The measured signal of the plant is stored in the measured signal database, a model is created using the measured signal and the model parameter of the plant stored in the measured signal database, and the model input or the measured signal database is created using this model. The plant measurement signal of the plant to be measured is input as an input to the measurement signal stored in the plant, and the parameters given to the model are learned so that the model output simulated by the model achieves the target value of the model output. The learned model parameters are stored in the model parameter database. And so, it is characterized in that so as to calculates and outputs the processed signal to eliminate drift noise and instrument from the measurement signal of the plant based on the learning result of the learning by using the model.

ADVANTAGE OF THE INVENTION According to this invention, while removing the noise contained in the measured plant data and the drift component of a measuring device simultaneously, it is possible to grasp an accurate plant state, and to improve the control performance of the plant. And a plant data processing method can be realized.

A plant data processing apparatus and a plant data processing method according to embodiments of the present invention will be described below with reference to the drawings.

FIG. 1 is a control block diagram showing the configuration of a plant data processing apparatus, which describes a plant data processing apparatus and a plant data processing method according to an embodiment in which the present invention is applied to a plant 100 to be measured.

In FIG. 1, the plant data processing apparatus of the present embodiment is a plant 100 to be measured by a coal-fired boiler, which will be described later, and a plant measured from this plant 100 to grasp the operation of the plant and the state of the plant. The data processing system includes a data processing device 200 that performs processing for removing noise and drift components from plant data obtained by measuring the operation and state of a coal-fired boiler plant, and a maintenance tool 910 that maintains the data processing device 200. An input device 900 for an operator of the plant 100 to input to the maintenance tool 910 and an image display device 950 for displaying various data are provided.

Various state quantities of the plant 100, which is a coal-fired boiler to be measured, measured by various measuring instruments described below are transmitted to the data processing apparatus 200 as measurement information 1.

Then, as shown in FIG. 2, a plant data processing signal from which noise and drift components have been removed by the calculation of the data processing device 200 is input to the boiler control device 1000, and the boiler control device 1000 performs arithmetic processing to perform coal burning. Air dampers 160, 161, 162, which will be described later, serve as operation ends of the coal-fired boiler as command signals for the plant 100, which is a boiler.
, 163 to operate each, and operate the coal fired boiler efficiently.

As shown in FIG. 1, the data processing device 200 is provided with an external input interface 210 that takes in a measurement signal 1 obtained by measuring a plant state quantity from the measurement target plant 100 into the data processing device 200. The measurement signal 1 from the plant 100 is taken in via the interface 210.

Further, the data processing apparatus 200 is provided with an external output interface 270, and a data processing signal of a plant that has performed processing for removing noise and drift components from the plant data measured by the processing unit 600 of the data processing apparatus 200. 9 is transmitted as an output signal 15 to the maintenance tool 910 provided with the data transmission / reception processing unit 930 via the external output interface 270.

The measurement signal 2 which is various state quantities of the plant captured from the plant 100 via the external input interface 210 is transmitted to and stored in the measurement signal database 220 installed in the data processing device 200, and the data processing device. The measurement signal 3 is transmitted to the switching unit 230 installed at 200.

In the switching unit 230, the measurement signal 3 is sent to the model creation unit 300 installed in the data processing apparatus 200 as the measurement signal 4 or the measurement signal to the processing unit 600 according to the switching signal 20 output from the switching unit 230. 8 is transmitted respectively.

That is, if the switching signal 20 output from the switching unit 230 is “model creation”, the switching unit 23.
The measurement signal 4 is transmitted from 0 to the model creation unit 300. If the switching signal 20 output from the switching unit 230 is “data processing”, the measurement signal 8 is transmitted from the switching unit 230 to the processing unit 600.

A learning unit 500 installed in the data processing apparatus 200 is connected to a model 400 installed in the data processing apparatus 200.

The model 400 simulates the plant characteristics of the plant 100 that is the coal-fired boiler to be controlled. When the switching signal 20 is “model creation” by using a model formula based on a physical law or a statistical method. Has a function of simulating and calculating the model output 13 with respect to the model input 12, and has a function of simulating and calculating the model output 14 with respect to the measurement signal 8 when the switching signal 20 is “data processing”.

That is, the model constructed in the model 400 includes a model expression for simulating the plant characteristics of the plant 100, or a statistical method, so that the model output 13 with respect to the model input 12,
Alternatively, the model output 14 for the measurement signal 8 is simulated.

The model parameter 7 necessary for the model 400 is input to the model 400 from the model creation unit 300 installed in the data processing apparatus 200.

The model creation unit 300 uses the previous model parameter 5 stored in the model parameter database 240 installed in the data processing device 200 and the measurement signal 4 transmitted from the switching unit 230 to use the model 400. It has a function to generate a prototype model of.

In addition, the model creation unit 300, when there is no previous model parameter 5 in the model parameter database 240, and the model parameter newly generated by a random number or the like and the switching unit 2
30 has a function of generating a prototype model of the model 400 by using the measurement signal 4 transmitted from 30.

The model creation unit 300 stores the newly created model parameter as the current model parameter 6 in the model parameter database 240.

The model creation unit 300 is connected to the learning unit 500, extracts model parameters 7 necessary for learning from the model parameter database 240, and transmits them to the learning unit 500.

The learning unit 500 obtains the model parameters 7 stored in the model parameter database 240 via the model creation unit 300.

The learning unit 500 learns and generates the model input 12 to be input to the model 400 using the model parameter 7 and the model output 13.

In the learning unit 500, the model parameter 7 in the model 400 is updated via the model input 12 and the model output 13, and the updated learning information is transmitted to the model parameter database 240 as the updated model parameter 10.

In the processing unit 600 installed in the data processing device 200, the switching unit 2 receives the switching signal 20.
A measurement signal 8 transmitted from the switching unit 230 when "data processing" is input to
Similarly, a model output 14 obtained by simulating the model parameter 7 to which the model parameter 7 for estimation is transmitted is obtained from the model creation unit 300 that has received the switching signal 20, and noise and drift components are removed from the measured plant data. A processing signal 9 that has been processed is generated.

The processing signal 9, which is plant data from which noise and drift components generated by the processing unit 600 are removed, is transmitted to the processing signal database 250 and the external output interface 270 installed in the data processing device 200.

A user involved in the plant 100 uses a measurement signal database provided in the data processing device 200 by using an input device 900 including a keyboard 901 and a mouse 902 and a maintenance tool 910 connected to the image display device 950. The information stored in the model parameter database 240 and the signal processing database 250 can be accessed.

The maintenance tool 910 includes an external input interface 920 and a data transmission / reception processing unit 930.
And an external output interface 940, and the data processing device 20
The data processing signal 9 of the plant which has been processed by the processing unit 600 of 0 to remove noise and drift components is output as the output signal 15 via the external output interface 270 to the maintenance tool 9.
10 is output.

Further, the input signal 31 generated by the input device 900 is taken into the data transmission / reception processing unit 930 of the maintenance tool 910 via the external input interface 920, and the data transmission / reception processing unit 930 inputs input via the external input interface 920. According to the information of the signal 32, the database information 30 provided in the data processing device 200 is acquired.

The data transmission / reception processing unit 930 transmits an output signal 33 obtained as a result of processing the database information 30 of the data processing device 200 to the external output interface 940, and outputs the output signal 34 from the external output interface 940 to the image display device 950. To display.

In the plant data processing apparatus of this embodiment, the database is entirely the data processing apparatus 2.
Although these are arranged inside 00, they can also be arranged outside the data processing apparatus 200.

In this embodiment, all the signal processing functions for generating the output signal 15 are arranged inside the data processing apparatus 200, but these may be arranged outside the data processing apparatus 200.

Next, the data processing apparatus of the plant which is an embodiment in which the data processing apparatus according to the present invention is applied to a thermal power plant equipped with a coal-fired boiler is stored in a database installed in the data processing apparatus of the present embodiment. Information and signal processing functions will be described.

First, the power generation mechanism of a thermal power plant including a coal-fired boiler, which is the plant 100 to be measured, will be described with reference to FIG.

In FIG. 2, fuel coal is supplied from a coal bunker 111 storing coal to a coal feeder 1.
12 to the mill 110. In the mill 110, the coal is finely pulverized by an internal roller into pulverized coal, and the furnace 101 is passed through a burner 102 installed in the furnace 101 constituting the coal-fired boiler together with primary air for transporting coal and secondary air for combustion adjustment. The furnace 1
The pulverized coal of fuel is burned inside 01.

The fuel 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, and after-air for two-stage combustion is introduced into the furnace 101 through the after-air port 103 installed in the furnace 101. The after air is supplied from the pipe 142 to the after air port 103.
Led to.

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

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

The water supply circulating through the furnace 101 is a condenser 11 that condenses steam that has passed through the turbine 108 into condensate.
3 to the furnace 101 through the feed water pump 105 and installed in the furnace 101
In 06, it is heated by the combustion gas flowing down inside the furnace 101 to become high-temperature and high-pressure steam. In this embodiment, the number of heat exchangers 106 is shown as one, but a plurality of heat exchangers may be arranged.

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

As various measuring instruments that detect plant data, which is the operating state quantity of a thermal power plant,
For example, in FIG. 2, steam generated in the heat exchanger 106 disposed in the furnace 101 is converted into steam turbine 1.
A temperature measuring device 151 and a pressure measuring device 152 are installed in the path supplied to 08 to measure the temperature T and pressure P of the steam.

In addition, a flow rate measuring device 150 is installed in a path for supplying this condensate to the heat exchanger 106 from a condenser 113 that cools the steam flowing down the steam turbine 108 and condenses it into condensate. The flow rate of water is measured, and the generator 109 is provided with a generation output measuring instrument 153 to measure the amount of power generated by the generator 109.

Further, the furnace 101 is provided with a concentration measuring device 154 for measuring NOx concentration and / or CO concentration in combustion gas generated by burning pulverized coal of fuel inside the furnace 101.

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

Information on various state quantities of the plant 100 to be measured, which is a coal-fired boiler measured by these various measuring instruments described above, is transmitted to the data processing device 200 as the measurement signal 1.

Then, a processing signal 51 obtained by removing noise and drift components from the plant data measured by the calculation of the data processing device 200 is input to the boiler control device 1000, and is processed by the boiler control device 1000 to be a coal fired boiler. The air dampers 160, 161, 162, which will be the operation ends of the coal-fired boiler based on the command signal for 100,
By operating each of 163, the coal-fired boiler can be operated efficiently according to the type of coal or the coal mixture ratio.

Next, primary air and secondary air that are input into the furnace 101 from the burner 102 that is installed in the furnace 101 of the coal-fired boiler that is the plant 100, and that that is input into the furnace 101 from the after air port 103 that is installed in the furnace 101 The after-air route will be described.

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

The air passing through the air heater 104 is heated by the combustion exhaust gas discharged from the furnace 101, and pulverized coal generated by the mill 110 is conveyed to the burner 102 through the pipe 133 using this primary air.

The secondary air and the after air are led from the fan 121 to the pipe 140, and the air heater 1
After flowing down the pipe 140 passing through the inside of the pipe 04 and being heated, it was branched into a pipe 141 for secondary air and a pipe 142 for after-air on the downstream side of the pipe 140 and installed in the furnace 101 respectively. It is configured to be guided to the burner 102 and the after-air port 103.

In the thermal power plant including the coal-fired boiler that is the control target 100 shown in FIG. 2, the above-described various measuring devices that detect the operating state of the thermal power plant are arranged and obtained from these measuring devices. The measurement signal 1 indicating the state quantity of the thermal power plant that is the plant 100 is configured to be input to the external input interface 210 installed in the data processing separate apparatus 200.

Further, as shown in FIG. 3, the air damper 16 includes a secondary air pipe 141 and an after-air pipe 142 branched on the downstream side of the pipe 140 provided in the air heater 104.
2 and the air damper 163 are arranged. Similarly, the pipe 132 provided inside the air heater 104 and the pipe 131 bypassing the air heater 104 are provided with the air damper 16.
1 and an air damper 160 are respectively arranged.

By operating these air dampers 160, 161, 162, and 163, the areas through which air passes in the pipes 131, 132, 141, and 142 are changed, and the air dampers 160, 161, 162, and 163 pass through these pipes 131, 132, 141, and 142. The air flow rate supplied to the furnace 101 is individually adjusted.

And based on the measurement signal 1 which shows the state quantity of the thermal power plant provided with the coal-fired boiler input from the plant 100 to be measured, the data processing device 200 performs a calculation process to remove noise and drift of the measuring instrument. The processing signal 9 is output as an output signal 15, and the water supply pump 105, the mill 110, and the air dampers 160, 161 are based on the output signal 15.
, 162, 163 and the like are operated to control the operation of the plant 100.

Next, information stored in the measurement signal database 220 installed in the plant data processing apparatus 200 according to the present embodiment will be described.

FIG. 4 is a diagram for explaining an aspect of information stored in the measurement signal database 220 installed in the plant data processing apparatus 200 according to the present embodiment, and FIG. 5 is a diagram illustrating the plant data processing apparatus 200 according to the present embodiment. It is a figure explaining the outline of the model structure of the model 400 installed in FIG.

As shown in FIG. 4, the measurement signal database 22 installed in the data processing device 200.
At 0, plant state information, which is plant data measured by various measuring instruments with respect to the plant 100, is input via the external input interface 210, and stored together with each measurement time for each measuring instrument.

For example, the flow rate measuring device 150 and the temperature measuring device 1 installed in the thermal power plant shown in FIG.
51, the flow rate value F, the temperature value T, the pressure value P, the power generation output value E, and the NOx concentration D contained in the combustion gas measured by the pressure measuring device 152, the power generation output measuring device 153, and the concentration measuring device 154, respectively, Are stored in the measurement signal database 220 of the data processing device 200.

In the measurement signal data shown in FIG. 4, data is stored at a cycle of 1 second, but the sampling cycle of data collection can be arbitrarily set.

In order to make it easy to use the measurement signal data stored in the measurement signal database 220, each measurement signal data is assigned a unique number called a PID number.

The model creation unit 300 installed in the data processing device 200 sets the model parameter 7 in the model 400 according to the state of the switching signal 20. If the switching signal 20 is “model creation”, the model parameter 7 necessary for newly constructing a model is output from the model creation unit 300 to the model 400 to set a new model.

When utilizing the past model parameters, the past model parameters stored in the model parameter database 240 are loaded.

If the switching signal 20 is “data processing”, the model parameter 7 of the model 400 used for data processing is loaded from the model parameter database 240 into the model 400 via the model creation unit 300.

Next, the operation of the model 400 and the learning unit 500 installed in the data processing device 200 constituting the plant data processing device according to the present embodiment will be described.

FIG. 5 is a diagram for explaining a model structure constructed in the model 400. The model structure built in the model 400 is composed of an input layer, a mapping layer, a bottleneck layer, a demapping layer, and an output layer, and the nodes of each layer are connected to each other.

The node portion uses a linear or nonlinear function, but generally uses a sigmoid function.

Each node connection has a weighting coefficient, and represents the strength of the mutual relationship between the nodes. Normal,
Model parameter 7 refers to this weighting factor.

As is clear from the model structure described above, the model structure built in this model 400 is 5
A neural network consisting of layers and symmetric about the bottleneck layer is called a self-associative neural network.

In the model 400, a node called a removal node is newly added to the bottleneck layer of the self-associative neural network.

Such a self-associative neural network is called a robust self-associative neural network. The output side of this removal node is not connected to any layer. This removal node has a mechanism for removing noise and drift components of the measuring instrument.

As an input signal for the model 400, an associated measurement signal fetched from the plant 100 is input. For example, measurement signals such as pressure, temperature, and flow rate related to the water supply system of the thermal power plant may be selected. However, the present invention is not limited to this, and all measurement signals of the thermal power plant may be input.

The learning unit 500 determines a model parameter 7 for the model 400. As a technique, a teacher signal is given to the output signal of the model 400, and the parameter is corrected so that the absolute value or the square error of the difference becomes small. In general, such a parameter correction method is called learning.

In the model 400, as shown in FIG. 5, the model structure constituting the model 400 has five layers, and a removal node is added based on a self-associative neural network having a symmetrical number of nodes around the bottleneck layer. A new configuration called a robust self-associative neural network is used.

Since the structure of the self-associative neural network that is the base and the learning method thereof are well-known techniques, the following “MA Kramer: Nonlinear P” is used.
rincipal Component Analysis Usage Autoas
social Neural Networks, AlChE Journal
, Vol. 37, no. 2,1991 "is shown as a reference.

The feature of the self-associative neural network is that the same signal as the input signal is given as a teacher signal. When an input signal is given to the model 400, the same signal as the input signal is obtained as an output signal by learning.

Thereby, a non-linear relationship between input signals can be established in the model 400. In general, the number of bottleneck layers is generally smaller than the number of input signals, and this also has the function of extracting only the main components of the input signals.

In the robust self-associative neural network constituting the model 400 shown in FIG. 5, a removal node is newly added to the bottleneck layer of this self-associative neural network. A learning method in this case will be described.

First, since the output layer of the neural network is the same as that of a conventional self-associative neural network, the learning method is also the same.

Therefore, the learning method described in the reference is applied to the weighting coefficient of each node not connected to the removal node.

On the other hand, a different learning method is applied to the weighting coefficient of each node connected to the removal node. In the removal node, to extract components related to noise and instrument drift,
The weighting factor needs to be corrected. Noise is generally called white noise and follows a normal distribution.

Since it is a well-known technique that instrument drift follows a normal distribution, the following “N. Kus”
umi et al: Development of Basic System for
r Sensor Calibration Support in Nuclear
Power Plant, Proceedings of International
l Conference on Nuclear Thermal Hydrauli
cs, Operations and Safety, 2004 "is shown as a reference.

Therefore, if the output signal of the removal node in the robust self-associative neural network can be learned so as to follow a normal distribution, the plant noise and the drift component of the measuring instrument can be collected in the removal node.

As a general property of the normal distribution, it is known that the skewness and kurtosis become zero. “
“The University of Tokyo College of Liberal Arts, Statistics, Volume, Introduction to Statistics, University of Tokyo Press, published in 1991” is shown as a reference.

From this, it is only necessary to calculate the skewness and kurtosis of the output signal of the removal node and to modify the associated weighting coefficient so that their values become zero.

If the measurement signal 1 measured from the plant 100 is a variable according to a certain probability distribution, the model input 12 input to the model 400 can also be regarded as a random variable.

Therefore, if x (t) (t = 1,..., T) is a random variable, the skewness is shown in Equation (2) based on the expected value obtained in Equation (1), and the sharpness is shown in Equation (3). Degrees.

In the learning unit 500, if the output signal of the removal node in the robust self-associative neural network constituting the model 400 can be learned so as to follow a normal distribution, the noise of the measurement signal of the plant and the drift component of the measuring instrument are removed. Therefore, the related weighting factors are corrected so that the skewness and kurtosis values of the signal components become zero.

As described above, in order to correct the weighting factor so that the output of the removal node follows the normal distribution, the weighting factor may be corrected so that the evaluation function J shown in Expression (4) becomes zero.

A function for performing the calculations of the above formulas (1) to (4) is incorporated in the learning unit 500.

A specific algorithm for correcting the weighting factor will be described with reference to the flowchart in the arithmetic processing of the learning unit 500 shown in FIG.

FIG. 6 is a diagram illustrating a flowchart of arithmetic processing in the learning unit 500 for creating the model 400 installed in the data processing device 200 constituting the plant data processing device of the present embodiment. It should be noted that parameters necessary for execution of arithmetic processing in each step of this flowchart are stored in the model parameter database 240. The form of information stored in the model parameter database 240 will be described later.

As shown in the flowchart of FIG. 6 as each step, first, in step 401, whether to use the model parameter 7 set in the past or to newly create the model parameter 7 is selected.

When a new model parameter is created, the initial value of the model parameter is set using a random number in the next step 402.

Next, in step 403, the measurement signal 4 that is the input signal of the model 400 and the teacher signal is extracted from the measurement signal database 220.

In the next step 404, parameters necessary for learning, such as the number of learning, the learning coefficient, and the number of nodes, are set. When a new model parameter is created, a default value stored in the model parameter database 240 is used.

  In the next step 405, the model parameter 7 is sequentially updated from the initial value by learning.

The robust self-associative neural network constituting the model 400 shown in FIG. 5 includes a model parameter 7 updated by learning regarding the output signal of each node in the output layer and a model updated by learning regarding the output signal of the removal node. There is parameter 7.

The update method of the model parameter 7 regarding the output signal of each node of the output layer is as follows.
The ropagation method is used.

The above learning method is described in detail in “Neural Nets and Measurement Control, Nishikawa? Ichi, Kitamura Shinzo, Asakura Shoten, published January 25, 1995”.

Basically, the model parameters are updated so that the difference between the output signal and the teacher signal when the input signal is given to the model 400 is eliminated.

The difference between the output signal from the model 400 and the teacher signal is generally expressed by a square error and is called an evaluation function.

Therefore, the learning unit 500 performs partial differential calculation of the variation of the evaluation function when each model parameter is varied, and the obtained value multiplied by the learning coefficient is used as the update of the model parameter 7. When this is repeated, the model parameter 7 is updated so that the output signal of the model 400 becomes the same signal as the teacher signal.

The method for updating the model parameter 7 related to the output signal of the removal node is to calculate the variation of the evaluation function shown by the equation (4) by partial differentiation, and update the model parameter 7 by multiplying the calculated value by the learning coefficient. .

  A specific calculation formula for the above evaluation function will be described below.

In order to partially differentiate equation (4), partial differentiation of expected values included in each term must be obtained.

About the 1st term of a formula (4), it becomes a formula (6)-(8). The function f (x) of each node is
The sigmoid function shown in Formula (5) is used. The output value of the removal node is x, and the output value of the jth node in the mapping layer is hj.

Let wjk be the weighting factor between the jth node in the mapping layer and the kth removal node in the bottleneck layer.

Further, since the output value of each node is a time-series signal, it is related to time. However, as described above, since the average value of all times is calculated by the expected value calculation of Expression (1), Is no longer included.

In addition, functions for performing calculations of the following formulas (5) to (15) are incorporated in the learning unit 500.

  For the second term of equation (4), equations (9) to (13) are obtained by calculating in the same manner.

  From the above, the partial differential calculation expression representing the variation of Expression (5) is as follows.

  Therefore, the model parameter update formula is the following formula (15).

Using model (15), the model parameter 7 related to the output signal of the removal node is updated.

In the process of repeatedly updating the model parameter 7 related to the output of each node of the output layer and the model parameter 7 related to the output signal of the removal node, when the evaluation function approaches zero, the value of the partial differential also approaches zero, The model parameter update amount approaches zero.

In the numerical calculation, it is never completely zero. Therefore, when the evaluation function is equal to or less than the value set in step 406 shown in the flowchart of FIG. 6, it is considered that learning is finished, and the model creation is finished.

If the learning end condition is not satisfied, the iterative calculation is stopped when the number of learning repetitions reaches the set number, and the process returns to step 404 to set the learning parameters again.

In the flowchart shown in FIG. 6, if the use of the past model parameter is selected in the step 401, the process proceeds to the step 407, where the past model parameter 7 is set as an initial value and corrected by learning. Choose whether or not.

If it is selected in step 407 to correct the past model parameter 7, the process proceeds from step 407 to step 403.

If it is selected not to modify the past model parameter 7 in step 407, the past model parameter 7 is used as it is, so that it is not necessary to reconstruct the model 400, and the model creation ends.

In this embodiment, the sigmoid function is used for the nodes of the robust self-associative neural network constituting the model 400, but other basis functions are used depending on the situation of the plant that is the model and the characteristics of the measurement signal. A network model may be used.

FIG. 7 is an explanatory diagram for explaining the form of information stored in the model parameter database 240 installed in the data processing device 200 constituting the data processing device of the plant of this embodiment.

As shown in FIG. 7, the model parameter database 240 includes model names, IDs, creation dates and times, learning coefficients, learning times, termination conditions, and the number of nodes as input / output layers, mapping / demapping layers, bottleneck layers, parameters. The values are specific examples of parameters for determining the weighting coefficient and the basis function.

Here, the weighting coefficient is the mutual coupling of the nodes, and is stored as W11, W12,. The parameters that determine the basis function are, for example, a slope and an input side translation amount in the case of a sigmoid function, and a center value, a variance value, and an input signal translation amount in the case of a Gaussian function.

An ID value of 000 indicates a default value of a learning parameter when a model parameter of the model 400 is newly created. The number of nodes and parameter values are usually blank for new creation.

Next, operations of the model 400 and the processing unit 600 installed in the data processing device 200 constituting the data processing device of the plant of this embodiment will be described with reference to FIG.

In FIG. 1, the processing unit 600 is measured in the plant 100 and the external interface 2
The measurement signal 2 obtained through 10 is input as a measurement signal 8 and output as a processed signal 9 from which noise and drift of the measuring instrument are removed.

First, when the switching signal 20 is “data processing”, the model parameter 7 used for data processing is set in the model 400 from the model creation unit 300.

Further, the measurement signal 8 stored in the measurement signal database 220 by the switching unit 230 is the model 4.
00 and the processing unit 600 are input.

In the model 400, the measurement signal 8 is input as an input signal to the model constructed with the model parameter 7 set by the model creation unit 300. As a result, the output signal obtained by the calculation of the model 400 is transmitted to the processing unit 600 as the model output 14.

The processing unit 600 calculates the difference between the model output 14 and the measurement signal 8 for each time, and outputs it to the external output interface 270 and the processing signal database 250 as the processing signal 9 together with the model output 14.

Next, as shown in FIG. 8, the processing signal database 250 installed in the data processing device 200 constituting the data processing device of the plant of this embodiment includes, for example, a model name, PID.
, The input signal, the model output 14, and the processed signal 9 are stored for each time.

Further, as shown in FIG. 5 as the signal ex in the output signal of the removal node, the PID representing the type of the input signal of each model is ex, and the corresponding signal is stored in the model output column of FIG. Other input signals and differential signal portions are blank.

The external output interface 270 transmits the measurement signal 8 and the processed signal difference signal 9 calculated by the processing unit 600 to the maintenance tool 910 as the output signal 15.

Next, a user related to the plant 100 to be measured is provided by the maintenance tool 91 of the input device 900.
A method of displaying the output signal 15 and information of each database on the image display device 950 using 0 will be described.

9 to 13 show an image display device 950 installed in the data processing device of the plant of this embodiment.
It is an Example of the screen displayed on. The user performs an operation such as inputting a parameter value in a blank area of these screens using the keyboard 901 or the mouse 902 of the input device 900.

FIG. 9 is an initial screen displayed on the image display device 950. The user selects a necessary button from the learning condition setting button 951, the data processing condition setting button 952, and the information display button 953 shown in FIG. 9, and moves the cursor 954 in FIG. By clicking the mouse 902, a necessary button is selected and pressed from these.

FIG. 10 is a diagram for explaining a learning condition setting screen in the data processing device 200 installed in the data processing device of the plant of the present embodiment. When the user clicks the learning condition setting button 951 in FIG. 9, the screen in FIG. 10 is displayed.

The user inputs settings necessary for the model 400 for each model. A model is selected from the model switching menu 963 displayed at the upper right of the screen of FIG.

In the model generation field 961 of the screen of FIG. 10, the learning coefficient, the number of learning times, and the end condition necessary for executing the learning process flowchart in the learning unit 500 shown in FIG. It is possible to modify the values entered in or already entered.

In addition, the user can change the default value whose ID is 000 by using this function. In the model structure field 962 of the screen of FIG. 10, the structure of the model 400, the types of basis functions to be set for the nodes, and values necessary for setting the parameters are input.

When the save button 964 displayed at the bottom of FIG. 10 is clicked, the information input to the model generation field 961 and the model structure field 962 is stored in the model switching database 963 in the model parameter database 240 in the data processing apparatus 200 of FIG. Saved for each model entered in.

In addition, the information input to the model generation field 961 and the model structure field 962 is set to “model learning” in the switching signal 20 and transferred to the model creation unit 300 in the data processing device 200 of FIG. .

When a cancel button 965 displayed at the bottom of FIG. 10 is clicked, information input in the model creation field 961, the model structure field 962, and the model switching menu 963 is cancelled. Further, when the return button 966 is clicked, the screen of FIG. 9 is restored.

FIG. 11 is a diagram illustrating a data processing condition setting screen in the data processing apparatus 200 of the plant data processing apparatus according to the present embodiment. When the user clicks the data processing condition setting button 952 with the mouse in FIG. 9, the screen in FIG. 11 is displayed.

The user selects the model 400 to be used for data processing for each target data from the data processing model switching menu 973 displayed on the upper right of the screen of FIG.

The model generation information column 971 on the screen of FIG. 11 displays the unique ID of the selected model, the learning coefficient used for learning, the number of learnings when learning is completed, and the learning error. The model 400 whose ID is 000 is not displayed.

In the model structure information column 972 of the screen of FIG. 11, the structure of the selected model 400, basis functions, and information on parameters set to the basis functions are displayed.

When the transfer button 974 displayed at the bottom of the screen of FIG. 11 is clicked, the information input to the model generation information column 971, the model structure information column 972, and the discrimination setting column 977 is stored in the data processing device 200 of FIG. The “switching signal 20” is set to “data processing” and transferred to the model creation unit 300. The model creation unit 300 sets the model parameter 7 in the processing unit 600 based on the transferred information.

When a cancel button 975 displayed at the bottom of the screen of FIG. 11 is clicked, information input to the model creation information column 971, the model structure information column 972, and the data processing model switching menu 973 is cancelled. Further, when the return button 976 is clicked, the screen of FIG. 9 is restored.

FIG. 12 shows the conditions for displaying the information stored in the measurement signal database 220 and the processing signal database 250 on the image display device 950 in the data processing apparatus 200 of the plant data processing apparatus according to the present embodiment. It is a screen to set.

In FIG. 9, when the user clicks the information display button 953, the screen of FIG. 12 is displayed.

The user can input a measurement signal, an operation signal, a processing signal difference signal 9 which is a calculation result in the processing unit 600, an output signal at the removal node, and the like in the input field 981 on the screen of FIG. Input with the range (upper limit / lower limit). Also, the time to be displayed is input in the time input field 982 on the screen of FIG.

The signal name can be set by designating a PID number in advance by creating a list corresponding to the PID number. Alternatively, if a list of measurement signals or the like is registered in the pull-down menu, it can be selected from the menu.

When the user clicks the display button 983 on the screen in FIG. 12, a trend graph as shown in FIG. 13 is displayed on the image display device 950. Also, by clicking the return button 991 displayed at the bottom of FIG. 13, the screen of FIG. 12 is returned.

Also, by clicking the back button 984 displayed at the bottom of FIG.
You can return to the screen.

In addition to the images described above, arbitrary information stored in the database in the data processing device 200 can be displayed on the image display device 950 in an arbitrary manner.

Next, operational effects of thermal power plant operation control, operation monitoring, and plant state diagnosis when the plant data processing apparatus and the plant data processing method according to the present invention are applied to a thermal power plant will be described.

When the plant data processing apparatus and the plant data processing method according to the present invention are applied to a coal-fired boiler of a thermal power plant, efficient operation control of the plant becomes possible.

Specifically, by applying the plant data processing device and the plant data processing method of the present invention to the operation control of a coal-fired boiler of a thermal power plant, noise and drift of a measuring instrument are detected from a measurement signal obtained from the thermal power plant. By operating an air damper that adjusts the amount of air supplied to the coal-fired boiler based on the state quantity of the excluded plant, the CO and NOx in the combustion gas discharged from the coal-fired boiler are made efficient to desired values. Operation control can be reduced.

Regulation values are set for CO and NOx discharged into the atmosphere from the chimney of a thermal power plant. Especially for NOx, the gas at the boiler outlet is led to a denitration device, where the regulation value of NOx is passed through denitration treatment. Protecting.

Since the consumption of ammonia used in the denitration apparatus increases as the NOx concentration at the denitration apparatus inlet increases, reducing the NOx amount at the denitration apparatus inlet as much as possible improves the cost merit.

Also, in plant operation monitoring and condition diagnosis, because the plant state can be grasped with high accuracy based on the accurate measurement signal excluding noise and instrument drift from the measurement signal obtained from the thermal power plant, Safer and more accurate thermal power plant operation monitoring and condition diagnosis are possible.

As described above, according to the embodiment of the present invention, the noise contained in the measured plant data and the drift component of the measuring instrument are simultaneously removed to enable an accurate grasp of the plant state and the control performance of the plant. An improved plant data processing apparatus and plant data processing method can be realized.

The present invention is applicable to a plant data processing apparatus and a plant data processing method such as a power generation plant from which various data are collected in order to grasp the operation and state of the plant.

The control block diagram which shows schematic structure of the data processing apparatus of the plant which is an Example of this invention. The schematic block diagram of the thermal power plant provided with the coal burning boiler to which the processing apparatus of the plant data which is the Example of this invention shown in FIG. 1 is applied. The fragmentary figure which shows the air heater of the thermal power plant provided with the coal burning boiler shown in FIG. Explanatory drawing which shows the aspect of the data memorize | stored in the measurement signal database installed in the data processing apparatus in the data processing apparatus of the plant of the present Example shown in FIG. Explanatory drawing which shows the structure of the model installed in the data processing apparatus in the data processing apparatus of the plant of the present Example shown in FIG. The flowchart figure which shows the operation | movement in the model preparation part installed in the data processing apparatus in the data processing apparatus of the plant of the present Example shown in FIG. Explanatory drawing which shows the aspect of the data memorize | stored in the model parameter database installed in the data processing apparatus in the data processing apparatus of the plant of the present Example shown in FIG. Explanatory drawing which shows the aspect of the data memorize | stored in the processing signal database installed in the data processing apparatus in the data processing apparatus of the plant of the present Example shown in FIG. The initial screen displayed on the image display apparatus installed in the data processing apparatus in the data processing apparatus of the plant of a present Example shown in FIG. The learning condition setting screen displayed on the image display apparatus installed in the data processing apparatus in the data processing apparatus of the plant of a present Example shown in FIG. The data processing condition setting screen displayed on the image display apparatus installed in the data processing apparatus in the data processing apparatus of the plant of the present Example shown in FIG. The display information setting screen displayed on the image display apparatus installed in the data processing apparatus in the data processing apparatus of the plant of the present Example shown in FIG. The trend graph of the selected measurement signal displayed on the image display apparatus installed in the data processing apparatus in the data processing apparatus of the plant of the present Example shown in FIG.

Explanation of symbols

100: Plant, 200: Data processing device, 210: External input interface, 2
20: measurement signal database, 230: switching unit, 240: model parameter database, 250: processing signal database, 260: learning parameter database, 300: model creation unit, 400: model, 500: learning unit, 600: processing unit, 900 : Input device, 90
1: Keyboard, 902: Mouse, 910: Maintenance tool, 920: External input interface, 930: Data transmission / reception processing unit, 940: External output interface, 950: Image display device.

Claims (14)

In the data processing device of a plant that obtains a measurement signal obtained by removing noise and drift of a measuring instrument based on a measurement signal obtained by measuring a state quantity of a plant to be measured,
The data processing device includes an external input interface for importing a measurement signal of the plant that measures the state quantity of the plant to be measured into the data processing device, a measurement signal database that stores the measured measurement signal of the plant, and model input or measurement. A learning unit that learns a model that simulates plant characteristics of the plant to be measured by using a measurement signal of the plant as an input, and a parameter that is given to the model so that the model output simulated by the model achieves the target value of the model output A processing unit that calculates a processing signal for removing noise and drift of the measuring instrument from the measurement signal of the plant based on a learning result by the learning unit, and a processing signal calculated by the processing unit from the data processing device. With an external output interface that outputs to Data processing device. In the data processing device of a plant that obtains a measurement signal from which noise and drift of the measuring instrument are removed based on the measurement signal obtained by measuring the state quantity of the plant to be measured,
The data processing apparatus includes an external input interface for importing a measurement signal of the plant that measures the state quantity of the plant to be measured into the data processing apparatus, a measurement signal database that stores the measured measurement signal of the plant, a measurement signal of the plant, A model creation unit that creates a model using model parameters, a model that simulates the plant characteristics of the plant to be measured by using a model input or a measured plant measurement signal as an input, and a model output that is simulated by the model is a model. A learning unit that learns parameters to be given to the model so as to achieve an output target value, and a processing signal that removes noise and instrument drift from the measurement signal of the plant based on a learning result by the learning unit The processing unit and the processing signal calculated by the processing unit The data processing apparatus of the plant, characterized in that it comprises an external output interface for outputting to the outside. In a data processing apparatus of a plant that obtains a measurement signal obtained by removing noise and drift of a measuring instrument based on a measurement signal obtained by measuring a state quantity of a measurement target plant, the data processing apparatus measures a state quantity of the measurement target plant. An external input interface for importing plant measurement signals into the data processing device, a measurement signal database for storing the measured plant measurement signals, a model creation unit for creating a model using the plant measurement signals and model parameters, and a model A model that simulates the plant characteristics of the plant to be measured using the measurement signal of the plant that is input or measured, and parameters that are given to the model so that the model output simulated by the model achieves the target value of the model output A learning unit that learns and model parameters learned by the learning unit A model parameter database for storing data, a processing unit for calculating a processing signal for removing noise and drift of the measuring instrument from the measurement signal of the plant based on a learning result by the learning unit, and the processing signal calculated by the processing unit And an external output interface for outputting the data from the data processing device to the outside. 4. The plant data processing apparatus according to claim 1, wherein the learning unit extracts a signal component according to a normal distribution from a measurement signal stored in a measurement signal database, and the learning unit extracts the signal component. A plant data processing apparatus, wherein the processing unit calculates a processing signal for removing noise and drift of the measuring instrument from the measurement signal of the plant based on the signal component. 5. The plant data processing apparatus according to claim 4, wherein the learning unit calculates a skewness and a kurtosis of a signal component extracted from a measurement signal stored in a measurement signal database, corrects a weighting coefficient, and performs the learning. A plant data processing apparatus, wherein the processing unit calculates a processing signal for removing noise and drift of the measuring instrument from the measurement signal of the plant based on the weighting coefficient corrected by the unit. 6. The plant data processing apparatus according to claim 1, wherein the data processing apparatus determines a model structure, a basis function type, and a basis function shape as an external input function. A plant data processing apparatus comprising a user interface for inputting at least one of the parameters for use. The plant data processing apparatus according to any one of claims 1 to 6, wherein the data processing apparatus is configured to detect noise and drift of a measuring instrument from a measurement signal of the plant to be measured, which is acquired from an external input interface. A plant data processing apparatus comprising a display device for displaying a processing signal from which components have been removed. In a plant data processing method for obtaining a measurement signal from which noise and drift of a measuring instrument are removed based on a measurement signal obtained by measuring a state quantity of a plant to be measured,
The measurement signal of the plant that has measured the state quantity of the plant to be measured is stored in the measurement signal database, and the model input using the model or the measurement signal of the plant stored in the measurement signal database as an input is input to the measurement target plant. The plant characteristics are simulated, and the parameters given to the model are learned so that the model output simulated by the model achieves the target value of the model output. Based on the learning result, noise and measuring instrument A plant data processing method characterized in that a processing signal for removing drift is calculated and output. In the plant data processing method for obtaining a measurement signal obtained by removing noise and drift of the measuring instrument based on the measurement signal obtained by measuring the state quantity of the plant to be measured,
The measurement signal of the plant that measured the state quantity of the plant to be measured is stored in the measurement signal database, a model is created using the measurement signal of the plant and the model parameter stored in the measurement signal database, and this model is used. Model input or measurement signal of the plant stored in the measurement signal database is used as an input to simulate plant characteristics of the plant to be measured, and the model output simulated by the model achieves a target value of model output A plant data processing method characterized by learning a parameter to be given to a model, and calculating and outputting a processing signal for removing noise and instrument drift from the plant measurement signal based on the learning result . In the plant data processing method for obtaining a measurement signal obtained by removing noise and drift of the measuring instrument based on the measurement signal obtained by measuring the state quantity of the plant to be measured,
The measurement signal of the plant that measured the state quantity of the plant to be measured is stored in the measurement signal database, a model is created using the measurement signal of the plant and the model parameter stored in the measurement signal database, and this model is Using the model input or the measurement signal of the plant stored in the measurement signal database as an input to simulate the plant characteristics of the plant to be measured so that the model output simulated by the model achieves the target value of the model output The parameters to be given to the model are learned, and the learned model parameters are stored in the model parameter database. Based on the learning results learned using the model, noise and instrument drift are measured from the measurement signal of the plant. Calculate and output the processed signal to be removed Data processing method of the plant, characterized in that the. The plant data processing method according to any one of claims 8 to 10, wherein a signal component according to a normal distribution is extracted by learning from a measurement signal stored in a measurement signal database, and the extracted signal component A plant data processing method characterized in that a processing signal for removing noise and instrument drift is calculated from the plant measurement signal. 12. The plant data processing method according to claim 11, wherein a skewness and a kurtosis of the signal component extracted from a measurement signal stored in a measurement signal database are calculated to correct a weighting factor, and based on the weighting factor. A plant data processing method characterized in that a processing signal for removing noise and instrument drift is calculated from the plant measurement signal. The plant data processing method according to any one of claims 8 to 10, wherein parameters for determining a model structure, a basis function type, and a basis function shape are externally determined for the model. A plant data processing method characterized in that at least one of them can be input. 12. The data processing method for a plant according to claim 8, wherein noise and a drift component of the measuring instrument are removed from the measurement signal of the plant that measures the state quantity of the measurement target plant. A data processing method for a plant, characterized in that the processed signal that has been processed is displayed on a display device.

Images