JP7561529B2 - Prediction device, plant, prediction method, program, and configuration program - Google Patents

Description

This disclosure relates to a prediction device, a plant, a prediction method, a program, and a configuration program.

Catalysts are used in plants equipped with boilers, etc. In plants where catalysts are used, the catalysts deteriorate as the plants operate. Therefore, in plants where catalysts are used, it is desirable to grasp the degree of deterioration of the catalysts.
Japanese Patent Application Laid-Open No. 2003-233633 discloses a related technique for predicting the degree of deterioration of a catalyst.

Patent No. 6278296

Incidentally, in order to predict the degree of deterioration of a catalyst in a plant, a user of the plant generally needs to carry out complicated processing such as analyzing the properties of exhaust gas each time the prediction is made.
For this reason, there is a demand for technology that allows users to know the degree of catalyst deterioration without having to perform complex processing in plants.

The present disclosure aims to provide a prediction device, a plant, a prediction method, a program, and a configuration program that can solve the above problems.

In order to solve the above problem, the deterioration prediction device of the present disclosure includes: a model generation unit that generates a first prediction model, which is a neural network that predicts a deterioration level of a catalyst in a second plant different from the first plant, based on learning data including first data related to a catalyst in past operation of a first plant and second data related to the state of the past operation, the learning data having planned values, plant data, and catalyst data after poisoning as inputs and catalyst deterioration data as output, and includes: a model generation unit that generates a first prediction model, which is a neural network that predicts the deterioration level of a catalyst in a second plant different from the first plant, based on the first prediction model generated by the model generation unit; and a deterioration prediction unit that predicts the deterioration level in the second plant based on the first prediction model generated by the model generation unit .

The differential pressure prediction device according to the present disclosure includes the deterioration prediction device described above and a differential pressure estimation unit that estimates the differential pressure between the input and output of the air preheater based on the degree of deterioration predicted by the deterioration prediction device.

The plant according to the present disclosure is a plant in which a catalyst is used, and includes a device in which the catalyst deteriorates, and the deterioration prediction device described above that predicts the degree of deterioration of the catalyst.

The deterioration prediction method of the present disclosure includes generating a first prediction model which is a neural network that predicts a degree of deterioration of a catalyst in a second plant different from the first plant, the first prediction model being learning data including first data related to a catalyst in past operation of a first plant and second data related to the state of the past operation, the learning data having planned values, plant data, and catalyst data after poisoning as inputs and catalyst deterioration data as output, and predicting the degree of deterioration in the second plant based on the generated first prediction model .

The differential pressure prediction method of the present disclosure includes: generating a first prediction model which is a neural network that predicts a degree of catalyst deterioration in a second plant different from the first plant, in which weighting of data connections between nodes is determined based on learning data including first data related to a catalyst in past operation of a first plant and second data related to the state of the past operation, the learning data having planned values, plant data, and catalyst data after poisoning as inputs and catalyst deterioration data as output; predicting the degree of deterioration in the second plant based on the generated first prediction model; and estimating a differential pressure between an input and an output of an air preheater based on the predicted degree of deterioration .

A program according to the present disclosure causes a computer to execute the following steps: generate a first prediction model, which is a neural network that determines weighting of data connections between nodes based on learning data including first data related to a catalyst in past operations of a first plant and second data related to the state of the past operations, the learning data having planned values, plant data, and catalyst data after poisoning as inputs and catalyst deterioration data as output, and predicts the degree of deterioration of a catalyst in a second plant different from the first plant; and predict the degree of deterioration in the second plant based on the generated first prediction model .

A program for causing a computer to execute the processing of a configuration according to the present disclosure configures, as hardware, a model generation unit that determines weighting of data connections between nodes based on learning data including first data related to a catalyst in past operations of a first plant and second data related to the state of the past operations, and that generates a first prediction model, which is a neural network that predicts the deterioration level of a catalyst in a second plant different from the first plant, based on the learning data that inputs planned values, plant data, and catalyst data after poisoning and outputs catalyst deterioration data, and a deterioration level prediction unit that predicts the deterioration level in the second plant based on the first prediction model generated by the model generation unit .

The prediction device, plant, prediction method, program, and configuration program according to the embodiments of the present disclosure allow the plant user to know the degree of catalyst deterioration without performing complex processing.

FIG. 1 is a diagram illustrating an example of a configuration of a plant according to a first embodiment of the present disclosure. FIG. 1 is a diagram illustrating an example of a configuration of a sensor device according to a first embodiment of the present disclosure. 1 is a diagram illustrating an example of a configuration of a deterioration prediction device according to a first embodiment of the present disclosure. FIG. 2 is a diagram illustrating an example of past performance data in various plants in the first embodiment of the present disclosure. 4A to 4C are diagrams for explaining generation of a first model and a second model in the first embodiment of the present disclosure. FIG. 11 is a diagram showing an example of data input from a first model and a second model in the first embodiment of the present disclosure. FIG. 2 is a first diagram showing a processing flow of the deterioration prediction device according to the first embodiment of the present disclosure. FIG. 2 is a second diagram showing the processing flow of the deterioration prediction device according to the first embodiment of the present disclosure. FIG. 11 is a diagram showing an example of a comparison result of examples in the first embodiment of the present disclosure. FIG. 11 is a diagram illustrating an example of a configuration of a differential pressure prediction device according to a second embodiment of the present disclosure. FIG. 11 is a diagram illustrating an example of a configuration of a plant according to a second embodiment of the present disclosure. FIG. 11 is a diagram showing a process flow of the differential pressure prediction device according to the second embodiment of the present disclosure. FIG. 11 is a diagram for explaining the processing of the differential pressure prediction device according to the second embodiment of the present disclosure. FIG. 13 is a diagram showing an example of a comparison result of examples in the second embodiment of the present disclosure. FIG. 1 is a schematic block diagram illustrating a configuration of a computer according to at least one embodiment.

First Embodiment
Hereinafter, the embodiments will be described in detail with reference to the drawings.
The deterioration prediction system according to the first embodiment of the present disclosure predicts a deterioration state of a denitration device 20 provided in a plant 1, based on operation data of the plant 1 to be predicted. The deterioration prediction system includes a data server device 50 and a deterioration prediction device 40. The data server device 50 is a device that stores data of the plant 1. The data of the plant 1 includes, for example, fuel properties, operation data, and the like. The deterioration prediction device 40 predicts a deterioration state of the denitration device 20 provided in the plant 1, based on the data stored in the data server device 50.

(Plant configuration)
The configuration of a plant 1 (an example of a second plant, an example of a plant) according to the first embodiment will be described.
The plant 1 according to the first embodiment is a plant for predicting the degree of deterioration of the catalyst performance with respect to the total operation time of the plant 1 .
As shown in FIG. 1 , the plant 1 includes a boiler 10 , a denitration device 20 (an example of a device in which catalyst deterioration occurs), and a sensor device 30 .

The boiler 10 is a boiler that uses coal or other fuel. When the fuel is burned, the boiler 10 discharges the combustion gas to the denitrification device 20.

The denitration device 20 is a device that reduces the NOx concentration in the combustion gas by decomposing NOx contained in the combustion gas. For stable operation of the plant 1, it is preferable to keep the denitration rate of the denitration device 20 approximately constant.
For example, the denitration rate of the denitration device 20, which tends to decrease due to deterioration of the catalyst performance caused by poisoning components, is kept constant by injecting ammonia (NH ₃ ) at the inlet of the denitration device 20. The poisoning components are components that enter from the outside and are different from the main components of the catalyst.
Poisoning components enter the denitration device 20 from the outside. Therefore, in the denitration device 20, as the poisoning components increase, the active sites decrease. As a result, the value of the reaction rate constant K, which indicates the level (quality) of the catalyst performance, decreases. In other words, the performance of the catalyst deteriorates.
The plant 1 according to the first embodiment is the subject of prediction of the degree of deterioration of the catalyst performance.

2, the sensor device 30 includes a first sensor 301, a second sensor 302, a third sensor 303, and a fourth sensor 304. The sensor device 30 is provided at the inlet of the denitrification device 20.
The first sensor 301 detects the dust concentration at the inlet of the denitration device 20 .
The second sensor 302 detects the NOx concentration at the inlet of the denitration device 20 .
The third sensor 303 detects the concentration of SOx at the inlet of the denitration device 20 .
The fourth sensor 304 detects the O 2 (oxygen) concentration at the inlet of the denitration device 20 .
The data measured by the sensor device 30 is transmitted to the data server device 50 via a network such as the Internet. As a result, data of the plant 1 is accumulated in the data server device 50.

(Configuration of Deterioration Prediction Device)
The deterioration prediction device 40 is a device that predicts the degree of deterioration of the catalyst performance with respect to the total operation time of the plant 1 .
As shown in FIG. 3 , the deterioration prediction device 40 includes a memory unit 401, a planned value acquisition unit 402, a plant data acquisition unit 403, a first model generation unit 404 (an example of a model generation unit), a second model generation unit 405 (an example of a model generation unit), a post-poisoning data acquisition unit 406, and a catalyst deterioration level acquisition unit 407 (an example of a deterioration level prediction unit).

The storage unit 401 stores various information necessary for the processing of the deterioration prediction device 40 .
For example, the storage unit 401 stores past performance data D1 for various plants (an example of a first plant).
Specifically, the storage unit 401 stores, for example, the planned values, plant data, and post-poisoning catalyst data shown in FIG.
It should be noted that the storage unit 401 stores data that changes over time in association with the time.

As shown in FIG. 4, the planned values include catalyst performance, catalyst specifications, and equipment specifications.
Examples of catalyst performance include the initial value K0 of the reaction rate constant, the ratio (K/K0) of the reaction rate constant K after a certain operating time has elapsed to the initial value K0 of the reaction rate constant (the degree of deterioration of the reaction rate constant, i.e., the degree of deterioration of the catalyst), etc. The initial value K0 of the reaction rate constant is a value given in advance.

Examples of catalyst specifications include the initial specific surface area of the catalyst, the initial pore volume of the catalyst, and the initial composition of the catalyst (e.g., TiO2 (titania), WO3 (tungsten trioxide), V2O5 (vanadium pentoxide), SiO2 (silica), etc.). The catalyst specifications are obtained, for example, by sampling and analyzing the actual catalyst.

Examples of device specifications include flow rate, dust concentration at the inlet of the denitration device 20, NOx concentration at the inlet of the denitration device 20, SOx concentration at the inlet of the denitration device 20, and O2 (oxygen) concentration at the inlet of the denitration device 20.
The flow rate is a value determined for each device. The dust concentration at the inlet of the denitration device 20, the NOx concentration at the inlet of the denitration device 20, the SOx concentration at the inlet of the denitration device 20, and the O2 concentration at the inlet of the denitration device 20 are values measured by a sensor device 30 provided at the inlet of the denitration device 20. The dust concentration at the inlet of the denitration device 20, the NOx concentration at the inlet of the denitration device 20, the SOx concentration at the inlet of the denitration device 20, and the O2 concentration at the inlet of the denitration device 20 may be calculated from the fuel properties. The fuel properties here are data indicating the amount of each component in the ash that flows into the catalyst, and are data obtained by analyzing fuel for each plant and held.

As shown in FIG. 4, the plant data includes fuel properties and operation data.
An example of the fuel property is the amount of fly ash components such as SiO2 and Al2O3 that flow into the catalyst.
Moreover, an example of the operation data is the operation time of the plant.

As shown in FIG. 4, the catalyst data after poisoning includes the catalyst properties after poisoning and the catalyst composition after poisoning.
Examples of the catalyst properties after poisoning include the specific surface area of the catalyst after poisoning and the pore volume of the catalyst after poisoning.
Examples of the catalyst composition after poisoning include the value obtained by normalizing the proportion of TiO2 in the catalyst after poisoning with the proportion of WO3 (TiO2/WO3), and the value obtained by normalizing the proportion of SiO2 in the catalyst after poisoning with the proportion of WO3 (SiO2/WO3).
The catalyst data after poisoning is data on the catalyst inside the denitration device 20, and is difficult to obtain while the plant is in operation. Therefore, the catalyst data after poisoning is obtained during maintenance and inspection of the plant 1, etc.

More specifically, the memory unit 401 stores the catalyst deterioration degree (i.e., the ratio (K/K0)) determined for past plant operation.

Furthermore, for example, the storage unit 401 stores data used when predicting the degree of deterioration of the catalyst in the plant 1 using a model.
Specifically, the storage unit 401 stores a first model (an example of a second prediction model) generated using past performance data D1, a second model (an example of a first prediction model) generated using past performance data D1, catalyst performance of the plant 1, catalyst specifications of the plant 1, equipment specifications of the plant 1, and fuel properties of the plant 1. The first model is a model that predicts catalyst data after poisoning in the plant 1. The second model is a model that predicts the degree of catalyst deterioration from the catalyst data after poisoning predicted by the first model. Each of the first model and the second model is a trained model that has been trained using past performance data in various plants.
An example of the catalytic performance of plant 1 is the initial value K0 of the reaction rate constant of plant 1.
Examples of the catalyst specifications for plant 1 include the initial specific surface area of the catalyst in plant 1, the initial pore volume of the catalyst in plant 1, and the like.
Furthermore, an example of the equipment specifications of the plant 1 is the flow rate in the plant 1.
Further, an example of the fuel properties of the plant 1 is the amount of components in the ash in the plant 1 that flow into the catalyst (for example, the amount of TiO2, SiO2, etc. that flow into the catalyst).
The generation of the first model and the generation of the second model will be described later.

The planned value acquisition unit 402 acquires the planned value of the plant 1 .
For example, the planned value acquisition unit 402 reads out the catalyst performance of the plant 1 , the catalyst specifications of the plant 1 , and the equipment specifications of the plant 1 from the storage unit 401 .
Specifically, the planned value acquisition unit 402 reads out from the memory unit 401 the initial value K0 of the reaction rate constant of plant 1, the initial specific surface area of the catalyst, the initial pore volume of the catalyst, the initial composition of the catalyst, the flow rate, etc.

In addition, for example, the planned value acquisition unit 402 acquires from the sensor device 30 the detection values of the dust concentration detected by the sensor device 30, the NOx concentration at the inlet of the denitration device 20, the SOx concentration at the inlet of the denitration device 20, and the O2 (oxygen) concentration at the inlet of the denitration device 20.
In addition, the detection values acquired by the planned value acquisition unit 402 may be stored in the memory unit 401, and when determining the degree of deterioration of a catalyst in another plant, the stored detection values may be used as performance data.

The plant data acquisition unit 403 acquires plant data for plant 1 from the data server device 50 and records the acquired plant data in the memory unit 401.

(Generation of the first model)
The first model generation unit 404 generates a first model. Examples of the first model include machine learning models such as a neural network, a random forest, and a support vector machine. In the following description, the generation of the first model will be described using a neural network as an example.

The first model generation unit 404 generates a trained first model by machine learning the modeled neural network using past performance data (planned values, plant data, and catalyst data after poisoning) from various plants as training data.
The neural network is, for example, a folded neural network having an input layer, a middle layer, and an output layer. The learning data here is data in which the planned value, the plant data, and the catalyst data after poisoning are each associated with each other in a one-to-one manner.

Specifically, the first model generation unit 404 divides a plurality of learning data (planned values, plant data, and post-poisoning catalyst data) into training data, evaluation data, and test data. As shown in FIG. 5, the first model generation unit 404 inputs the planned values and plant data of the training data into the neural network. The neural network (the first model before learning in FIG. 5) outputs the post-poisoning catalyst data. The first model generation unit 404 changes the weighting of the data coupling between nodes by performing backpropagation according to the output every time the planned values and plant data of the training data are input into the neural network and the post-poisoning catalyst data is output from the neural network (i.e., changes the model of the neural network). Next, the first model generation unit 404 inputs the planned values and plant data of the evaluation data into the neural network of the model changed by the planned values and plant data of the training data. The neural network outputs the post-poisoning catalyst data according to the input planned values and plant data. The first model generation unit 404 changes the weighting of the data coupling between nodes based on the output of the neural network as necessary. The neural network generated by the first model generating unit 404 in this manner is the trained first model. Next, the first model generating unit 404 inputs the planned value of the test data and the plant data to the trained neural network of the first model as a final check. The trained neural network of the first model outputs post-poisoning catalyst data corresponding to the planned value of the input test data and the plant data. When the post-poisoning catalyst data output by the trained neural network of the first model for all test data is within a predetermined error range with respect to the planned value of the input test data and the post-poisoning catalyst data associated with the plant data, the trained neural network of the first model is determined to be a desired model. In addition, when the post-poisoning catalyst data output by the trained neural network of the first model for even one of the test data is not within a predetermined error range with respect to the planned value of the input test data and the post-poisoning catalyst data associated with the plant data, the trained first model is generated using new training data.
The generation of a trained model by the first model generation unit 404 described above is repeated until a desired trained first model is obtained.
The first model generation unit 404 writes the generated trained first model into the storage unit 401 .

(Generation of the second model)
The second model generation unit 405 generates a second model.
For example, the second model generation unit 405 generates a trained second model by machine learning the modeled neural network using past performance data (planned values, plant data, catalyst data after poisoning, and catalyst deterioration data) from various plants as training data.
The neural network is, for example, a folded neural network having an input layer, a middle layer, and an output layer. The learning data here is data in which the planned value, the plant data, the catalyst data after poisoning, and the catalyst deterioration data are each associated with each other in a one-to-one manner.

Specifically, the second model generation unit 405 divides a plurality of learning data (planned values, plant data, catalyst data after poisoning, and catalyst deterioration data) into training data, evaluation data, and test data. As shown in FIG. 5, the second model generation unit 405 inputs the planned values, plant data, and catalyst data after poisoning of the training data to a neural network. The neural network outputs catalyst deterioration data. Then, similarly to the case where the first model generation unit 404 generates the first model, the second model generation unit 405 determines that the trained neural network of the second model is a desired model when the catalyst deterioration data output by the trained neural network of the second model for all test data is within a predetermined error range with respect to the catalyst deterioration data associated with the input planned values, plant data, and catalyst data after poisoning. In addition, if the catalyst deterioration data output by the neural network of the trained second model for any one of the test data is not within a specified error range with respect to the catalyst deterioration data associated with the planned value of the input test data, the plant data, and the catalyst deterioration data after poisoning, a trained second model is generated using new training data.
The generation of a trained model by the second model generation unit 405 described above is repeated until a desired trained second model is obtained.
The second model generation unit 405 writes the generated trained second model into the storage unit 401 .

The post-poisoning data acquisition unit 406 acquires post-poisoning catalyst data based on the planned values, the plant data, and the learned first model.

For example, the post-poisoning data acquisition unit 406 reads out from the memory unit 401 the catalyst performance of plant 1 (e.g., the initial value K0 of the reaction rate constant of plant 1), the catalyst specifications of plant 1 (e.g., the initial specific surface area of the catalyst in plant 1, the initial pore volume of the catalyst in plant 1, etc.), the equipment specifications of plant 1 (e.g., the flow rate in plant 1, etc.), the fuel properties of plant 1 (e.g., the amount of ash components flowing into the catalyst in plant 1), and the trained first model.
In addition, the post-poisoning data acquisition unit 406 acquires information indicating the dust concentration at the inlet of the denitration device 20, the NOx concentration at the inlet of the denitration device 20, the SOx concentration at the inlet of the denitration device 20, and the O2 concentration at the inlet of the denitration device 20 from the sensor device 30 during operation of the plant 1.
As shown in FIG. 6 , the post-poisoning data acquisition unit 406 inputs information indicating the catalyst performance of plant 1, the catalyst specifications of plant 1, the equipment specifications of plant 1, the fuel properties of plant 1, the dust concentration at the inlet of the denitration device 20 acquired from the sensor device 30, the NOx concentration at the inlet of the denitration device 20, the SOx concentration at the inlet of the denitration device 20, and the O2 concentration at the inlet of the denitration device 20, into the trained first model.
Then, the post-poisoning data acquisition unit 406 acquires post-poisoning catalyst data (for example, catalyst properties after poisoning, catalyst composition after poisoning) output by the trained first model.
This post-poisoning catalyst data acquired by the post-poisoning data acquisition unit 406 is post-poisoning catalyst data obtained by prediction when the plant 1 is operated for the time indicated by the plant data under the conditions indicated by the planned values and the plant data.

The catalyst deterioration level acquisition unit 407 acquires catalyst deterioration data based on the planned value, the plant data, the catalyst data after poisoning, and the learned second model.

For example, the catalyst deterioration level acquisition unit 407 reads out the catalyst specifications of plant 1 (e.g., the initial specific surface area of the catalyst in plant 1, the initial pore volume of the catalyst in plant 1, etc.), the equipment specifications of plant 1 (e.g., the flow rate in plant 1, etc.), and the learned second model from the memory unit 401.
In addition, the catalyst deterioration level acquisition unit 407 acquires information indicating the dust concentration at the inlet of the denitration device 20, the NOx concentration at the inlet of the denitration device 20, the SOx concentration at the inlet of the denitration device 20, and the O2 concentration at the inlet of the denitration device 20 from the sensor device 30 during operation of the plant 1.
In addition, the catalyst deterioration level acquisition unit 407 acquires the post-poisoning catalyst data predicted by the post-poisoning data acquisition unit 406 from the post-poisoning data acquisition unit 406 .
As shown in FIG. 6 , the catalyst deterioration level acquisition unit 407 inputs information indicating the catalyst specifications of plant 1, the equipment specifications of plant 1, the fuel properties of plant 1, the dust concentration at the inlet of the denitration device 20 acquired from the sensor device 30, the NOx concentration at the inlet of the denitration device 20, the SOx concentration at the inlet of the denitration device 20, and the O2 concentration at the inlet of the denitration device 20, the catalyst data after poisoning predicted by the post-poisoning data acquisition unit 406, and information indicating the operating time, into the learned second model.
Then, the catalyst deterioration level acquisition unit 407 acquires the catalyst deterioration data output by the trained second model.
The catalyst deterioration data acquired by the catalyst deterioration level acquisition unit 407 is catalyst deterioration data obtained by prediction when the plant 1 is operated under the conditions indicated by the planned values and plant data for the period of time indicated by the plant data.

(Model learning process)
The following describes the operation of the deterioration prediction device 40. First, the learning process of the machine learning model used by the deterioration prediction device 40 to predict catalyst deterioration will be described.
The manager of the deterioration prediction device 40, etc., acquires catalyst deterioration data of the catalyst of the denitration device 20 equipped in a plant other than the plant 1 that is the target of prediction in advance during maintenance and inspection of the plant. The catalyst deterioration data is obtained by measuring the NOx concentration of the gas at the input end and the NOx concentration of the gas at the output end of the denitration device 20. In addition, the manager of the deterioration prediction device 40, etc., acquires post-poisoning catalyst data including catalyst properties and catalyst composition by destructive inspection in advance during maintenance and inspection of the plant other than the plant 1 that is the target of prediction. The post-poisoning catalyst data is data of the catalyst inside the denitration device 20, and is difficult to acquire while the plant is in operation.

The first model generation unit 404 performs machine learning on the first model, which is a machine learning model, using past performance data (planned values, plant data, and catalyst data after poisoning) from various plants as learning data. The planned values and plant data are acquired from the data server device 50.
Specifically, the first model generation unit 404 generates training data in which the planned value and plant data of the performance data are used as input samples, and the catalyst data after poisoning is used as an output sample. The first model generation unit 404 trains the first model based on the training data so as to output a predicted value of the catalyst state after poisoning by inputting the planned value and the plant data.
The first model generation unit 404 writes the trained first model into the storage unit 401 .

The second model generation unit 405 trains the second model, which is a machine learning model, by machine learning past performance data (planned values, plant data, catalyst data after poisoning, and catalyst deterioration data) from various plants as learning data.
Specifically, the second model generation unit 405 generates training data using the planned value, plant data, and post-poisoning catalyst data from among the performance data as input samples and the catalyst deterioration data as an output sample. The second model generation unit 405 trains the second model based on the training data so as to output a predicted value of the catalyst deterioration data by inputting the planned value, plant data, and post-poisoning catalyst data.
The second model generation unit 405 writes the trained second model into the storage unit 401 .

(Catalyst Deterioration Prediction Processing)
Next, a process of predicting catalyst deterioration by the deterioration prediction device 40 will be described using the process flow of the deterioration prediction device 40 shown in Figures 7 and 8. The deterioration prediction device 40 predicts catalyst data after poisoning based on the planned value and plant data, and predicts catalyst deterioration data using the predicted catalyst data after poisoning.

(Processing to predict catalyst data after poisoning)
First, a process in which the deterioration prediction device 40 predicts catalyst data after poisoning based on the planned value, the plant data, and the learned first model will be described with reference to FIG.

The post-poisoning data acquisition unit 406 reads out the catalyst performance of plant 1 (e.g., the initial value K0 of the reaction rate constant of plant 1), the catalyst specifications of plant 1 (e.g., the initial specific surface area of the catalyst in plant 1, the initial pore volume of the catalyst in plant 1, etc.), the equipment specifications of plant 1 (e.g., the flow rate in plant 1, etc.), the fuel properties of plant 1 (e.g., the amount of ash components flowing into the catalyst in plant 1), and the trained first model from the memory unit 401 (step S1).

During operation of the plant 1, the post-poisoning data acquisition unit 406 acquires information indicating the dust concentration at the inlet of the denitration device 20, the NOx concentration at the inlet of the denitration device 20, the SOx concentration at the inlet of the denitration device 20, and the O2 concentration at the inlet of the denitration device 20 from the sensor device 30 (step S2).

The post-poisoning data acquisition unit 406 inputs the information indicating the catalyst performance of plant 1, the catalyst specifications of plant 1, the equipment specifications of plant 1, the fuel properties of plant 1, the dust concentration at the inlet of the denitration device 20, the NOx concentration at the inlet of the denitration device 20, the SOx concentration at the inlet of the denitration device 20, and the O2 concentration at the inlet of the denitration device 20, which are read from the memory unit 401, and information indicating the operating time, into the trained first model (step S3).

The post-poisoning data acquisition unit 406 acquires the post-poisoning catalyst data (e.g., the catalyst properties after poisoning, the catalyst composition after poisoning) output by the trained first model (step S4).

(Processing to predict catalyst deterioration data)
Next, a process in which the deterioration prediction device 40 predicts catalyst deterioration data based on the planned value, the plant data, the post-poisoning catalyst data, and the learned second model will be described with reference to FIG.

The catalyst deterioration level acquisition unit 407 reads out the catalyst specifications of plant 1 (e.g., the initial specific surface area of the catalyst in plant 1, the initial pore volume of the catalyst in plant 1, etc.), the equipment specifications of plant 1 (e.g., the flow rate in plant 1, etc.), and the trained second model from the memory unit 401 (step S11).

During operation of the plant 1, the catalyst deterioration level acquisition unit 407 acquires information indicating the dust concentration at the inlet of the denitration device 20, the NOx concentration at the inlet of the denitration device 20, the SOx concentration at the inlet of the denitration device 20, and the O2 concentration at the inlet of the denitration device 20 from the sensor device 30 (step S12).

The catalyst deterioration level acquisition unit 407 acquires the post-poisoning catalyst data predicted by the post-poisoning data acquisition unit 406 from the post-poisoning data acquisition unit 406 (step S13).

As shown in FIG. 6, the catalyst deterioration level acquisition unit 407 inputs the following information into the trained second model: the catalyst specifications of plant 1, the equipment specifications of plant 1, the fuel properties of plant 1, the dust concentration at the inlet of the denitration device 20 acquired from the sensor device 30, the NOx concentration at the inlet of the denitration device 20, the SOx concentration at the inlet of the denitration device 20, and the O2 concentration at the inlet of the denitration device 20 (step S14).

The catalyst deterioration level acquisition unit 407 acquires the catalyst deterioration data output by the trained second model (step S15).

(Example)
For a certain plant, a prediction of the degree of catalyst deterioration was performed by the deterioration prediction device 40 according to the first embodiment of the present disclosure described above, and the prediction was compared with an actual measurement of the degree of catalyst deterioration in the plant.
We prepared 127 sets of data for 11 plants with different manufacturing years and operating hours, of which 70% was used as learning data and the remaining 30% was used as accuracy verification data.
FIG. 9 shows the results of a comparison between the predicted degree of catalyst deterioration and the actual measurement of the degree of catalyst deterioration.
9, the horizontal axis represents operation time, and the vertical axis represents the degree of catalyst deterioration, which is calculated by dividing the reaction rate constant K at each operation time by the initial reaction rate constant K0.
The accuracy of the prediction of the degree of catalyst deterioration was found to be 0.05, with the root mean squared error (RMSE).

The plant 1 according to the first embodiment of the present disclosure has been described above.
The deterioration prediction device (40) includes model generation units (404, 405) and a deterioration degree prediction unit (407).
The model generation unit generates a first prediction model that predicts the degree of deterioration of a catalyst in a second plant (1) different from the first plant based on learning data including first data related to a catalyst in past operation of the first plant and second data related to the state of the past operation.
A deterioration degree prediction unit (407) predicts the deterioration degree in the second plant (1) based on the first prediction model generated by the model generation unit (404, 405).

This deterioration prediction device (40) can predict the degree of deterioration of the catalyst in the second plant (1) simply by inputting data about the second plant (1) into the first prediction model. As a result, the user of the second plant (1) can know the degree of deterioration of the catalyst without performing complex analysis of the catalyst.

Second Embodiment
The deterioration prediction system according to the second embodiment of the present disclosure predicts a deterioration state of a denitration device 20 included in a plant 1 to be predicted based on operation data of the plant 1, and predicts a differential pressure between an input and an output of an air preheater 60 (described later) from the predicted deterioration state. The deterioration prediction system includes a data server device 50 and a differential pressure prediction device 70.

(Configuration of differential pressure prediction device)
The differential pressure prediction device 70 is a device that predicts the differential pressure of the air preheater 60 based on the predicted degree of deterioration of the catalyst performance.
As shown in FIG. 10 , the differential pressure prediction device 70 includes a scale-up factor estimation unit 701 , a leak estimation unit 702 , a blockage degree estimation unit 703 , and a differential pressure estimation unit 704 in addition to the components of the deterioration prediction device 40 .

The scale-up factor estimation unit 701 obtains a correction coefficient for correcting the difference between the performance characteristics of the catalyst obtained in the laboratory and the performance characteristics of the catalyst in the actual plant 1. This correction coefficient is the scale-up factor.
For example, performance characteristics of a catalyst are obtained in a laboratory for various combinations of gas temperature, gas concentration, and oxygen concentration. The obtained performance characteristics of the catalyst are then expressed by an equation including the gas temperature, gas concentration, and oxygen concentration. However, in the actual plant 1, the performance of the catalyst is reduced compared to the performance characteristics of the catalyst obtained in the laboratory. The scale-up factor estimation unit 701 determines a scale-up factor for correcting the equation showing the performance characteristics of the catalyst obtained in the laboratory so that the equation can be applied to the actual plant 1. The scale-up factor is determined by an empirical rule obtained from the performance characteristics of the catalyst obtained in the laboratory and the performance characteristics of the catalyst in the actual plant 1.

The leak estimation unit 702 estimates the amount of leaked ammonia based on an empirical formula in which a scale-up factor is applied to an equation showing the performance characteristics of the catalyst obtained in a laboratory, the NOx concentration at the inlet of the denitration device 20, the NOx concentration at the outlet of the denitration device 20, and catalyst deterioration data. This involves determining the performance characteristics of the catalyst from the empirical formula, and estimating the amount of unreacted ammonia from the performance characteristics and the NOx concentrations at the input and output of the denitration device 20.

The degree of blockage estimation unit 703 estimates the degree of blockage in the air preheater 60 using an empirical formula calculated based on differential pressure data when the differential pressure in the air preheater 60 in the actual plant 1 is increasing and operating data of the plant 1.
For example, when the pressure difference of the air preheater 60 in the actual plant 1 is increasing, the pressure difference between the input and output of the element of the air preheater 60 is monitored and the pressure difference is calculated. The closing speed is calculated from the pressure difference, and an empirical formula is obtained using the operating data of the plant 1 at that time.
This is based on the idea that the leaked ammonia concentration can be estimated from operational data, and if this leaked ammonia is assumed to be precipitated in the air preheater 60 as acid ammonium sulfate (NH4HSO4), then the degree of blockage in the elements of the air preheater 60 can be estimated since there is a correlation between the amount of acid ammonium sulfate and the degree of blockage in the air preheater 60.

The differential pressure estimation unit 704 estimates the differential pressure between the input and output of the air preheater 60 based on the amount of leaked ammonia estimated by the leak estimation unit 702 and the degree of blockage in the element of the air preheater 60 estimated by the blockage degree estimation unit 703.

(Plant configuration)
The configuration of a plant 1 according to the second embodiment will be described.
As shown in FIG. 11 , the plant 1 according to the second embodiment includes a boiler 10, a denitration device 20, a sensor device 30, and an air preheater 60.
The air preheater 60 is a device for preheating the air for combustion to improve the combustion efficiency of the boiler. If ammonia leaks from the denitrification device 20, the leaked ammonia reacts with sulfur trioxide (SO3) in the combustion gas in the air preheater 60 to produce acid ammonium sulfate. The precipitation of acid ammonium sulfate is a cause of blockage. In order to remove this acid ammonium sulfate, it is necessary to stop the plant 1 and wash the air preheater 60 with water.

(Process for predicting pressure difference in air preheater 60)
Next, a process in which the differential pressure prediction device 70 predicts the differential pressure between the input and output of the air preheater 60 will be described with reference to FIGS. 12 and 13. FIG.
The deterioration prediction device 40 executes the processes shown in steps S11 to S15 to obtain catalyst deterioration data.

The scale-up factor estimation unit 701 calculates a scale-up factor that corrects the difference between the performance characteristics of the catalyst determined in the laboratory and the performance characteristics of the catalyst in the actual plant 1 (step S21).

The leak estimation unit 702 estimates the amount of leaked ammonia based on an empirical formula in which a scale-up factor is applied to an equation showing the performance characteristics of the catalyst obtained in a laboratory, the NOx concentration at the inlet of the denitration device 20, the NOx concentration at the outlet of the denitration device 20, and catalyst deterioration data (step S22).

The blockage degree estimation unit 703 estimates the degree of blockage in the air preheater 60 using an empirical formula calculated based on the differential pressure data when the differential pressure in the air preheater 60 in the actual plant 1 is increasing and the operating data of the plant 1 (step S23).

The differential pressure estimation unit 704 estimates the differential pressure between the input and output of the air preheater 60 based on the amount of leaked ammonia estimated by the leak estimation unit 702 and the degree of blockage in the element of the air preheater 60 estimated by the blockage degree estimation unit 703 (step S24).
The differential pressure estimation unit 704 may notify the estimated differential pressure to a person involved in catalyst installation work or a person in charge of the plant 1. Based on this notification, a catalyst replacement work plan or operation support for the plant 1 may be performed.

(Example)
For a certain plant, the differential pressure between the input and output of the air preheater 60 was estimated by the differential pressure prediction device 70 according to the second embodiment of the present disclosure described above.
FIG. 14 shows the results of a comparison between the predicted and measured differential pressures in the plant.
In Fig. 14, the horizontal axis represents operation time, and the vertical axis represents the differential pressure between the input and output of the air preheater 60. In Fig. 14, the air preheater 60 is indicated as AH (Air Heater). The air preheater 60 has been washed once with water.
As can be seen from FIG. 14, the prediction of the differential pressure between the input and output of the air preheater 60 by the differential pressure prediction device 70 obtained results close to the actually measured values.

The plant 1 according to the second embodiment of the present disclosure has been described above.
The differential pressure prediction device (70) includes a deterioration prediction device (40), a scale-up factor estimation section (701), a leak estimation section (702), a degree of blockage estimation section (703), and a differential pressure estimation section (704).
The scale-up factor estimation unit (701) determines a scale-up factor for correcting the difference between the performance characteristics of the catalyst determined in a laboratory and the performance characteristics of the catalyst in the actual plant (1). The leak estimation unit (702) estimates the amount of leaked ammonia based on an empirical equation in which the scale-up factor is applied to an equation indicating the performance characteristics of the catalyst obtained in a laboratory, the NOx concentration at the inlet of the denitration device (20), the NOx concentration at the outlet of the denitration device (20), and catalyst deterioration data. The clogging degree estimation unit (703) estimates the clogging degree in the air preheater 60 using an empirical equation determined based on differential pressure data when the differential pressure in the air preheater (60) in the actual plant (1) is increasing and operation data of the plant (1). A differential pressure estimation unit (704) estimates a differential pressure between the input and output of the air preheater (60) based on the amount of leaked ammonia estimated by the leak estimation unit (702) and the degree of blockage in the element of the air preheater (60) estimated by the blockage degree estimation unit (703).

This differential pressure prediction device (70) estimates the differential pressure between the input and output of the air preheater (60) when leaking ammonia occurs in the denitrification device (20). As a result, it is possible to plan catalyst replacement work and provide operational support for the plant 1 at the appropriate time.

The order of the processes in the embodiments of the present disclosure may be changed as long as appropriate processing is performed.

Note that the memory unit 401 and other storage devices (including registers and latches) in each embodiment of the present invention may be provided anywhere within the range in which appropriate information is transmitted and received. Furthermore, there may be multiple memory units 401 and other storage devices within the range in which appropriate information is transmitted and received, and data may be stored in a distributed manner.

In each embodiment of the present disclosure, the first model generation unit 404 has been described as generating a trained first model as software and storing it in the storage unit 401. In addition, in each embodiment of the present disclosure, the second model generation unit 405 has been described as generating a trained second model as software and storing it in the storage unit 401.
However, in another embodiment of the present disclosure, each of the trained first model and the trained second model may be realized as hardware.
For example, the first model generation unit 404 may write a configuration program that realizes the processing performed by the trained first model stored in the memory unit 401 into programmable hardware such as an FPGA (Field-Programmable Gate Array).
Also, for example, the second model generation unit 405 may write a configuration program that realizes the processing performed by the learned second model stored in the memory unit 401 into programmable hardware such as an FPGA (Field-Programmable Gate Array).

Although the embodiments of the present disclosure have been described, the deterioration prediction device 40, the differential pressure prediction device 70, and other control devices may have a computer device inside. The above-mentioned process steps are stored in a computer-readable recording medium in the form of a program, and the above-mentioned process is performed by the computer reading and executing the program. Specific examples of computers are shown below.
FIG. 10 is a schematic block diagram illustrating a computer configuration according to at least one embodiment.
As shown in FIG. 10, the computer 5 includes a CPU 6, a main memory 7, a storage 8, and an interface 9.
For example, the deterioration prediction device 40, the differential pressure prediction device 70, and the other control devices are each implemented in the computer 5. The operations of each of the above-mentioned processing units are stored in the storage 8 in the form of a program. The CPU 6 reads out the program from the storage 8, loads it in the main memory 7, and executes the above-mentioned processing in accordance with the program. The CPU 6 also secures storage areas in the main memory 7 corresponding to each of the above-mentioned storage units in accordance with the program.

Examples of storage 8 include HDD (Hard Disk Drive), SSD (Solid State Drive), magnetic disk, magneto-optical disk, CD-ROM (Compact Disc Read Only Memory), DVD-ROM (Digital Versatile Disc Read Only Memory), and semiconductor memory. Storage 8 may be an internal medium directly connected to the bus of computer 5, or an external medium connected to computer 5 via interface 9 or a communication line. In addition, when this program is distributed to computer 5 via a communication line, computer 5 that receives the program may expand the program in main memory 7 and execute the above-mentioned process. In at least one embodiment, storage 8 is a non-transitory tangible storage medium.

The program may also realize some of the functions described above. Furthermore, the program may be a file that can realize the functions described above in combination with a program already recorded in the computer device, a so-called differential file (differential program).

Although several embodiments of the present disclosure have been described, these embodiments are merely examples and do not limit the scope of the disclosure. Various additions, omissions, substitutions, and modifications may be made to these embodiments without departing from the spirit of the disclosure.

<Additional Notes>
The deterioration prediction device 40, the plant 1, the deterioration prediction method, the program, and the program for causing a computer to execute configuration processing described in each embodiment of the present disclosure can be understood, for example, as follows.

(1) A deterioration prediction device (40) according to a first aspect,
a model generating unit (404, 405) that generates a first prediction model for predicting a deterioration level of a catalyst in a second plant (1) different from the first plant, based on learning data including first data related to a catalyst in a past operation of the first plant and second data related to a state of the past operation;
a deterioration degree prediction unit (407) for predicting the deterioration degree in the second plant (1) based on the first prediction model generated by the model generation unit (404, 405);
Equipped with.

This deterioration prediction device (40) can predict the degree of deterioration of the catalyst in the second plant (1) simply by inputting data about the second plant (1) into the first prediction model. As a result, the user of the second plant (1) can know the degree of deterioration of the catalyst without performing complex analysis of the catalyst.

(2) A deterioration prediction device (40) according to a second aspect is the deterioration prediction device (40) of (1),
The learning data includes data on a catalyst after poisoning in a past operation of the first plant and data on the deterioration degree in a past operation of the first plant,
The model generation unit (404, 405)
a first model generation unit (404) that generates a second prediction model that predicts data related to the catalyst after poisoning in the second plant (1) based on the learning data;
a second model generation unit (405) that generates the first prediction model for predicting the deterioration degree in the second plant (1) based on data related to the catalyst after poisoning in the second plant (1) predicted by the second prediction model;
Equipped with
The deterioration degree prediction unit (407)
The deterioration degree in the second plant (1) is predicted based on the first prediction model and the second prediction model.

This deterioration prediction device (40) makes it possible to predict data related to catalysts in equipment that is difficult to predict for the second plant (1). As a result, the user of the second plant (1) can easily determine which data affects the degree of deterioration of the catalyst in the second plant (1).

(3) A differential pressure prediction device (70) according to a third aspect includes:
A deterioration prediction device (40) according to the first or second aspect,
a differential pressure estimation unit (704) for estimating a differential pressure between an input and an output of an air preheater (60) based on the deterioration degree predicted by the deterioration prediction device (40);
Equipped with.

This differential pressure prediction device (70) estimates the differential pressure between the input and output of the air preheater (60) when leaking ammonia occurs in the denitrification device (20). As a result, it becomes possible to plan catalyst replacement work and provide operational support for the plant 1 at the appropriate time.

(4) A differential pressure prediction device (70) according to a fourth aspect is the differential pressure prediction device (70) according to (3),
a blockage degree estimation unit (703) for estimating a blockage degree in the air preheater (60) based on the deterioration degree predicted by the deterioration prediction device (40);
Equipped with
The differential pressure estimation unit (704)
A pressure difference between the input and output of the air preheater (60) is estimated based on the degree of blockage estimated by the blockage degree estimation unit (703).

This differential pressure prediction device (70) clarifies the configuration of the differential pressure prediction device (70) and estimates the differential pressure between the input and output of the air preheater (60) when leaking ammonia occurs in the denitrification device (20). As a result, it becomes possible to plan catalyst replacement work and provide operational support for the plant 1 at the appropriate time.

(5) A differential pressure prediction device (70) according to a fifth aspect is a differential pressure prediction device (70) according to (3) or (4),
a leak estimation unit (702) that estimates an amount of leak ammonia based on the degree of deterioration predicted by the deterioration prediction device (40);
Equipped with
The differential pressure estimation unit (704)
A pressure difference between the input and output of the air preheater (60) is estimated based on the amount of leaked ammonia estimated by the leak estimation unit (702).

This differential pressure prediction device (70) clarifies the configuration of the differential pressure prediction device (70) and estimates the differential pressure between the input and output of the air preheater (60) when leaking ammonia occurs in the denitrification device (20). As a result, it becomes possible to plan catalyst replacement work and provide operational support for the plant 1 at the appropriate time.

(6) A differential pressure prediction device (70) according to a fifth aspect is any one of the differential pressure prediction devices (70) according to (3) to (5),
a scale-up factor estimation unit (701) for determining a scale-up factor based on the deterioration degree predicted by the deterioration prediction device (40);
Equipped with
The differential pressure estimation unit (704)
The scale-up factor estimation unit (701) estimates a pressure difference between the input and output of the air preheater (60) based on the scale-up factor calculated.

This differential pressure prediction device (70) clarifies the configuration of the differential pressure prediction device (70) and estimates the differential pressure between the input and output of the air preheater (60) when leaking ammonia occurs in the denitrification device (20). As a result, it becomes possible to plan catalyst replacement work and provide operational support for the plant 1 at the appropriate time.

(7) A plant (1) according to a seventh aspect,
A plant in which a catalyst is used,
a device (20) for deactivating the catalyst;
A deterioration prediction device (40) according to (1) or (2) for predicting a deterioration degree of the catalyst;
Equipped with.

With this plant (1), the degree of catalyst deterioration in the plant (1) can be predicted simply by inputting data about the plant (1) into the first prediction model. As a result, a user of the plant (1) can know the degree of catalyst deterioration without performing complex analyses on the catalyst.

(8) A deterioration prediction method according to an eighth aspect,
generating a first prediction model for predicting a deterioration degree of a catalyst in a second plant different from the first plant based on learning data including first data related to a catalyst in a past operation of the first plant and second data related to the state of the past operation;
predicting the deterioration degree in the second plant based on the generated first prediction model;
Includes.

This deterioration prediction method makes it possible to predict the degree of deterioration of the catalyst in the second plant (1) simply by inputting data about the second plant (1) into the first prediction model. As a result, a user of the second plant (1) can know the degree of deterioration of the catalyst without performing complex analyses of the catalyst.

(9) A program according to a tenth aspect,
On the computer,
generating a first prediction model for predicting a deterioration degree of a catalyst in a second plant different from the first plant based on learning data including first data related to a catalyst in a past operation of the first plant and second data related to the state of the past operation;
predicting the deterioration degree in the second plant based on the generated first prediction model;
Execute the command.

This program makes it possible to predict the degree of catalyst deterioration in the second plant (1) simply by inputting data about the second plant (1) into the first prediction model. As a result, a user of the second plant (1) can know the degree of catalyst deterioration without performing complex analyses on the catalyst.

(10) A program for causing a computer to execute a configuration process according to an eleventh aspect,
a model generating unit (404, 405) for generating a first prediction model for predicting a deterioration degree of a catalyst in a second plant (1) different from the first plant, based on learning data including first data related to a catalyst in a past operation of the first plant and second data related to the state of the past operation;
a deterioration degree prediction unit (407) for predicting the deterioration degree in the second plant (1) based on the first prediction model generated by the model generation unit (404, 405);
Each of the above is configured as hardware.

By using a program for causing a computer to execute the processing of this configuration, the degree of catalyst deterioration in the second plant (1) can be predicted simply by inputting data about the second plant (1) into the first prediction model. As a result, a user of the second plant (1) can know the degree of catalyst deterioration without performing complex analysis of the catalyst.

1... Plant 5... Computer 6... CPU
Reference Signs List 7: Main memory 8: Storage 9: Interface 10: Boiler 20: Denitrification device 30: Sensor device 40: Deterioration prediction device 50: Data server device 60: Air preheater 70: Differential pressure prediction device 301: First sensor 302: Second sensor 303: Third sensor 304: Fourth sensor 401: Memory unit 402: Plan value acquisition unit 403: Plant data acquisition unit 404: First model generation unit 405: Second model generation unit 406: Post-poisoning data acquisition unit 407: Catalyst deterioration degree acquisition unit 701: Scale-up factor estimation unit 702: Leak estimation unit 703: Blockage degree estimation unit 704: Differential pressure estimation unit

Claims (11)

a model generating unit that generates a first prediction model, which is a neural network that predicts a deterioration level of a catalyst in a second plant different from the first plant, by determining weighting of data connections between nodes based on learning data including first data related to a catalyst in a past operation of the first plant and second data related to a state of the past operation, the learning data having a planned value, plant data, and catalyst deterioration data after poisoning as inputs and an output of catalyst deterioration data;
a deterioration degree prediction unit that predicts the deterioration degree in the second plant based on the first prediction model generated by the model generation unit;
A deterioration prediction device comprising : The model generation unit
a weighting of data connections between nodes is determined based on the learning data having the planned value and the plant data as input and the post-poisoning catalyst data as output, a second prediction model is generated to predict data related to the post-poisoning catalyst in the second plant , and a first prediction model is generated to predict the deterioration degree in the second plant based on the data related to the post-poisoning catalyst in the second plant predicted by the second prediction model,
The deterioration degree prediction unit is
predicting the deterioration degree in the second plant based on the first prediction model and the second prediction model;
The deterioration prediction device according to claim 1 . The deterioration prediction device according to claim 1 or 2,
A differential pressure estimation unit that estimates a differential pressure between an input and an output of an air preheater based on the deterioration degree predicted by the deterioration prediction device;
A differential pressure prediction device comprising: a blockage degree estimation unit that estimates a blockage degree in the air preheater based on the deterioration degree predicted by the deterioration prediction device;
Equipped with
The differential pressure estimation unit is
A pressure difference between an input and an output of the air preheater is estimated based on the degree of blockage estimated by the blockage degree estimation unit.
The differential pressure prediction device according to claim 3 . a leak estimation unit that estimates an amount of leaked ammonia based on the degree of deterioration predicted by the deterioration prediction device;
Equipped with
The differential pressure estimation unit is
A pressure difference between an input and an output of the air preheater is estimated based on the amount of the leaked ammonia estimated by the leak estimation unit.
The differential pressure prediction device according to claim 3 or 4. a scale-up factor estimation unit that calculates a scale-up factor based on the deterioration degree predicted by the deterioration prediction device;
Equipped with
The differential pressure estimation unit is
A differential pressure between an input and an output of the air preheater is estimated based on the scale-up factor calculated by the scale-up factor estimation unit.
The differential pressure prediction device according to any one of claims 3 to 5. A plant in which a catalyst is used,
A device in which the catalyst degradation occurs;
A deterioration prediction device according to claim 1 or 2, which predicts a deterioration level of the catalyst;
The plant is provided with: generating a first prediction model, which is a neural network that predicts a degree of catalyst deterioration in a second plant different from the first plant, by determining weighting of data connections between nodes based on learning data including first data related to a catalyst in a past operation of the first plant and second data related to the state of the past operation, the learning data having planned values, plant data, and catalyst deterioration data as inputs and catalyst deterioration data as output;
predicting the deterioration degree in the second plant based on the generated first prediction model;
A deterioration prediction method comprising : generating a first prediction model, which is a neural network that predicts a degree of catalyst deterioration in a second plant different from the first plant, by determining weighting of data connections between nodes based on learning data including first data related to a catalyst in a past operation of the first plant and second data related to the state of the past operation, the learning data having planned values, plant data, and catalyst deterioration data as inputs and catalyst deterioration data as output;
predicting the deterioration degree in the second plant based on the generated first prediction model;
Estimating a differential pressure between an input and an output of the air preheater based on the predicted deterioration degree;
A differential pressure prediction method comprising : On the computer,
generating a first prediction model, which is a neural network that predicts a degree of catalyst deterioration in a second plant different from the first plant, by determining weighting of data connections between nodes based on learning data including first data related to a catalyst in a past operation of the first plant and second data related to the state of the past operation, the learning data having planned values, plant data, and catalyst deterioration data as inputs and catalyst deterioration data as output;
predicting the deterioration degree in the second plant based on the generated first prediction model;
A program that executes the following. a model generation unit that generates a first prediction model, which is a neural network that predicts a degree of catalyst deterioration in a second plant different from the first plant, based on learning data including first data related to a catalyst in past operations of a first plant and second data related to the state of the past operations, and which determines a weighting of data connections between nodes based on the learning data, which inputs planned values, plant data, and catalyst data after poisoning and outputs catalyst deterioration data, and a deterioration degree prediction unit that predicts the degree of deterioration in the second plant based on the first prediction model generated by the model generation unit, and a configuration process that causes a computer to configure, as hardware, each of the following:

Images