JP2023184210A - Plant management device and plant management method - Google Patents

Abstract

To predict results of processing performed by a chemical plant by reflecting characteristics of unit operation.SOLUTION: A chemical plant management device includes: a storage part which stores information of a plurality of different unit operations included in processing processes in a chemical plant, log data information related to the processing of the unit operations, and condition information related to a condition in each of the unit operations; an input part which inputs sensor information acquired from a sensor device related to a first unit operation in the chemical plant; and a calculation part which creates a prediction model for predicting results of the processing in the first unit operation based on the log data information and the condition information related to the first unit operation, and stores the prediction model into the storage part in association with the first unit operation.SELECTED DRAWING: Figure 4

Description

The present invention relates to a plant management device and a plant management method.

Conventionally, there is a technology that uses AI (Artificial Intelligence) or the like to create a model for predicting the quality of products produced by a plant. For example, Patent Document 1 describes, “a process data processing unit that performs predetermined processing on process data obtained from a chemical plant, and a process data processing unit that performs predetermined processing on process data obtained from the chemical plant, or Among the data, the chemical a predictive model creation unit that creates a predictive model that has learned the characteristics of the process data obtained from the plant; Find the value according to the

WO2021/157667

Model analysis using AI may be performed based on log data within the plant in order to analyze sensor information related to plant operation to understand or predict the situation within the plant.

On the other hand, there are various processes and unit operations included in each process in a plant, and the sensor information to be analyzed is information in any of the unit operations. Each unit operation within a plant may have its own operating conditions and methods.

However, in general, AI is data-driven and has no concept of process or unit operation constraints or special operating conditions that differ from normal operating conditions. The inventor discovered that there is a problem in that the operating conditions and methods specific to the operation are not taken into account, which affects analysis efficiency and analysis accuracy. For example, calculations may diverge to include data unrelated to the sensor information to be analyzed, making it impossible to perform efficient analysis, or analysis accuracy may be reduced because the basic theory of each unit operation is not taken into consideration. An example is putting it away.
Therefore, a system is desired that can create a model for analyzing sensor information to be analyzed while taking into account constraints such as analysis policies and analysis conditions specific to unit operations.

In order to achieve the above object, one of the representative plant management devices of the present invention collects information on a plurality of different unit operations included in a treatment process in a chemical plant and log data related to the processing of the unit operations. a storage unit that stores information and condition information regarding conditions in each of the unit operations; an input unit that inputs sensor information acquired from a sensor device related to a first unit operation in the chemical plant; , create a prediction model for predicting the result of processing in the first unit operation based on the log data information and the condition information related to the first unit operation, and and a calculation unit that stores data in the storage unit in association with a first unit operation.
Further, one of the representative plant management methods of the present invention inputs sensor information obtained from a sensor device related to a first unit operation included in a plurality of different unit operations included in a treatment process in a chemical plant. step, the sensor information, log data information related to the processing of the first unit operation, and condition information that is information regarding conditions in the first unit operation. The method includes the steps of creating a prediction model for predicting a processing result, and storing the prediction model in a predetermined storage device in association with the first unit operation.

According to the present invention, it is possible to predict the results of processing by a chemical plant by reflecting the characteristics of unit operations. Problems, configurations, and effects other than those described above will be made clear by the following description of the embodiments.

Explanatory diagram for reaction prediction of a chemical plant in an embodiment Explanatory diagram of a specific example of reaction prediction Diagram explaining the differences due to the presence or absence of condition information Configuration diagram showing the configuration of the plant management device Explanatory diagram for operation of plant management equipment Explanatory diagram for specific examples of grouping Flowchart showing the behavior when creating a predictive model Flowchart showing the operation of the predictive model during operation Explanatory diagram of automatic plant operation An explanatory diagram when using the batch method

Hereinafter, typical embodiments for carrying out the present invention will be described with reference to the drawings as appropriate.
FIG. 1 is an explanatory diagram of reaction prediction in a chemical plant in an embodiment. The plant management device disclosed in the embodiment predicts, for example, the result of a process performed using a reactor. In FIG. 1, sensor data is acquired using sensors regarding the raw materials to be introduced into the reactor and the state of the reactor.

Here, unit operations will be explained. A unit operation is a procedural element that defines an independent process, consisting of algorithms necessary for initiating, organizing, and controlling a phase. For example, reactions, fractional distillation, stirring, heat transfer, separation, etc. can be treated as unit operations. A plant process includes one or more of these unit operations. Examples of the process include, for example, a production process of a specific product, or a part of the production process. Note that unit operations and process divisions may be determined depending on the equipment configuration and operation method of the chemical plant, and specific processing contents.

The disclosed plant management device predicts processing results using condition information of unit operations in addition to sensor data.
The condition information indicates constraints such as an analysis policy and analysis conditions specific to a unit operation. The condition information includes the type of unit operation, a predetermined mathematical model, the type of process, operating conditions, and the like.
The type of unit operation is, for example, reaction or fractional distillation. The mathematical model is, for example, a reaction rate equation or a heat transfer calculation equation. The type of process is, for example, the type of product produced in the process. The operating conditions are conditions related to temperature, flow rate, etc. Examples include threshold values such as maximum, minimum, and standard deviation in sensor data such as temperature and flow rate, and unique values.

The plant management device generates a data set to be given to the prediction model by adding data, deleting sensor data, correcting sensor data, etc. based on the condition information. Here, since the type and period of sensor data required for analysis differ for each unit operation, by generating a dataset to be given to the prediction model based on condition information, only the necessary data can be used and higher performance can be achieved. It becomes possible to perform accurate and efficient predictions. The predictive model then uses the given data set to predict the outcome of the reactor process (eg, the quality of the product produced by the reactor).
In this way, by adding, deleting, and correcting data based on condition information to the data group acquired from the sensor, it becomes possible to predict reactions that take into account constraints such as analysis policies and analysis conditions specific to unit operations.

FIG. 2 is an explanatory diagram of a specific example of reaction prediction. For example, when predicting the quality of product A, the plant management device uses, in addition to the operation data of product A, condition information of unit operations used in the production process of product A as an explanatory variable. Therefore, more accurate model creation and more accurate prediction can be performed than when performing data analysis using AI using only the operational data of product A as an explanatory variable.

As the operation data for product A, log data from when product A was generated in the past may be used. As the condition information, reaction kinetics, reaction rate, etc. regarding product A can be used.

Similarly, the quality of product B and product C can be predicted by using condition information as an explanatory variable in addition to operation data.

The creation of a predictive model used for AI data analysis will be explained. The plant management device creates a prediction model for predicting the result of processing in a unit operation based on sensor data, log data related to the unit operation, and condition information. Then, the created prediction model is stored in a predetermined storage unit in association with a unit operation.

For example, in the case of a chemical plant that produces a single product, a large amount of operational data regarding a specific unit operation can be obtained, and a predictive model can be created from this operational data.
On the other hand, in a plant that handles a wide variety of products, if you try to analyze information on each unit operation for each product or product number, the data will become complex and huge, reducing the efficiency of analysis, and there is a possibility that the necessary data will be insufficient. There is.

Therefore, the disclosed plant management device groups unit operations with similar characteristics with respect to the unit operations to be analyzed, and compares the conditions of similar unit operations and the created ones with respect to the unit operations corresponding to the sensor information to be analyzed. It allows analysis to be performed based on the model. Here, the conditions of similar unit operations and the reason for using a created model will be explained. Analysis policies and analysis conditions associated with each unit operation may be common or similar depending on the type of unit operation, the equipment configuration of the chemical plant, the properties of the target raw materials, intermediates, and products, etc. May contain conditions. Therefore, even if the condition information or log data is related to a unit operation that is managed as a different unit operation from the unit operation to be analyzed, it is difficult to create a predictive model for the unit operation to be analyzed or to create a dataset for prediction. It may be available. For example, analysis policies and analysis conditions specific to unit operations may be included in different bases or different processes, but if unit operations of the same type include common or similar analysis policies and analysis conditions. There is. Furthermore, even unit operations of different types may include common or similar analysis policies and analysis conditions, depending on the type of feature quantity that is important in the prediction model and analysis, and the equipment configuration. Further, depending on the nature of the substance to be processed in each unit operation, common or similar analysis policies and analysis conditions may be included.

Specifically, the plant management device creates a plurality of characteristic groups each including one or more unit operations having characteristics within a predetermined range, based on the characteristics of the unit operations. Then, among the plurality of characteristic groups, a characteristic group that includes a unit operation related to the input sensor information is identified, and the sensor information and log data information related to one or more unit operations included in the characteristic group are A prediction model for predicting a processing result in a unit operation is created based on the condition information. In addition, as another method of creating a model, if the unit operation related to the input sensor information is the first unit operation, regarding other unit operations included in the characteristic group that includes the first unit operation. , a predictive model regarding the first unit operation may be created using a plurality of already created learning models. For example, a predictive model regarding the first unit operation can be created by performing ensemble learning on a plurality of existing learning models.

In FIG. 2, product A and product B are included in the same group Gr1. Note that although only product C belongs to group Gr2, only one unit operation may belong to one group in this way.
Grouping includes the type of unit operation, type of process, information on the equipment configuration in the chemical plant, reaction characteristics, information to identify whether it is a batch process or continuous process, information on the base to identify the location of the chemical plant, and information on the unit operation. Any characteristic can be used depending on the purpose of the analysis, such as the optimum feature amount under the operating conditions.
Regarding grouping, we have explained the case of creating one group for multiple unit operations, but it is also possible to create a group for each type of characteristic for each of multiple characteristic types related to unit operations. good. That is, as an example, when focusing on two characteristic types A and B, grouping for characteristic type A (pattern 1) and grouping for characteristic type B (pattern 2) are performed separately. . In this case, once sensor data to be analyzed is input, similar groups are identified for each type, and condition information, log data, or creation related to unit operations included in the identified multiple similar groups. A prediction model or a data set to be input to the prediction model may be created based on the completed prediction model. At this time, a grouping pattern to be referred to may be selected depending on the purpose of the analysis.

FIG. 3 is an explanatory diagram of the difference depending on the presence or absence of condition information. If only sensor data is used, the area selected by AI may be too large or too small for the necessary and sufficient data range. For example, a situation may occur in which the reactor internal temperature and stirring speed are appropriately selected, but the raw material composition and catalyst flow rate are omitted from the selection, and the raw material flow rate that does not need to be considered is included. In this way, it is difficult to improve the prediction accuracy of a prediction model that is trained using data that has excess or deficiency.

On the other hand, if condition information is used, a data set with a necessary and sufficient data range can be selected by reflecting the type of unit operation, whether the operation method is continuous or batch, data regarding startup and shutdown times, etc. Prediction models can be trained. Therefore, the prediction accuracy of the prediction model can be improved. Furthermore, by using condition information to select data to be input into a prediction model, it contributes to improving the efficiency of analysis and improving the accuracy of analysis by using only necessary data.

FIG. 4 is a configuration diagram showing the configuration of the plant management device. As shown in FIG. 4, the plant management device includes a calculation device 11, a main storage device 12, a user interface 13, a display section 14, a communication interface 15, and an auxiliary storage device 20.

The arithmetic device 11 is, for example, a CPU (Central Processing Unit), and operates as the arithmetic unit described in the claims.
The main storage device 12 is, for example, RAM (Random Access Memory). The arithmetic device 11 controls the operation of the plant management device by loading various programs and data into the main storage device 12 and sequentially executing the programs.

The user interface 13 is, for example, a keyboard, and accepts operational input from the administrator. Specifically, the user interface 13 is used for inputting product data and the like.
The display unit 14 is, for example, a liquid crystal display, and is used to display various information to the administrator.

The communication interface 15 communicates with the equipment of the chemical plant via the data integration infrastructure. In FIG. 4, chemical plants are constructed at multiple locations, such as base L1 and base L2. The chemical plants at each site are equipped with various plant equipment and sensor devices. For example, base L1 employs a continuous system and has a set of sensor devices and plant equipment for each unit operation. One or more unit operations form a process, and a plurality of processes form a production line for an object.

The communication interface 15 can receive sensor data from sensor devices at the base and can send control instructions and the like to plant equipment. That is, the communication interface 15 operates as an input unit described in the claims regarding sensor data.

The auxiliary storage device 20 is a storage unit that stores various data, and operates as a storage unit described in the claims. The auxiliary storage device 20 stores log data 21, a condition information database 22, a unit operation database 23, a predictive model database 24, a process model database 25, reference information 26, abnormality related information 27, and the like.

The log data 21 is log data related to processing of unit operations, and is, for example, operational data of a chemical plant. Each piece of log data may be stored in association with information on a related unit operation.
The condition information database 22 is a database in which condition information regarding conditions for each unit operation is registered.
The condition information may be a representation of a physical/chemical phenomenon related to a unit operation to be analyzed using a mathematical formula, theory, or method. Physical and chemical phenomena expressed using mathematical formulas, theories, and methods are stored as mathematical models. Examples include mathematical models showing reaction rate equations, Arrhenius equations, heat transfer equations, heat transfer area equations, distillation column plate number equations, pressure loss equations, stirring power equations, and the like.

The unit operation database 23 is a database in which information on a plurality of different unit operations included in processing steps in a chemical plant is registered.
The predictive model database 24 is a database in which predictive models for predicting processing results in unit operations are registered.
The process model database 25 is a database in which process models for simulating processes including unit operations on a computer are registered.
The reference information 26 is reference information indicating that an abnormality has occurred regarding each unit operation.
The abnormality related information 27 is information in which operating conditions when an abnormality occurs in a unit operation are stored in association with the unit operation.

The calculation device 11 creates a prediction model by referring to the sensor data acquired from the sensor device, the log data 21, and the condition information database 22, and registers the prediction model in the prediction model database 24 in association with a unit operation. .

Furthermore, the arithmetic device 11 can create a plurality of characteristic groups each including one or more unit operations having characteristics within a predetermined range, based on the characteristics of the unit operations. Then, among the plurality of characteristic groups, a characteristic group that includes a unit operation related to the input sensor data is identified, and the sensor data and log data and condition information related to one or more unit operations included in the characteristic group are combined. Based on this, a predictive model can be created.

For example, information on the type of unit operation may be used as the characteristic of the unit operation. In this case, the arithmetic device 11 creates a plurality of characteristic groups based on information on the type of unit operation.

Furthermore, characteristic groups can be created such that two or more unit operations included in different types of processes belong to the same group. For example, if "stirring" is included as a unit operation in the step of producing product A and the step of producing product B, a characteristic group including "stirring" of product A and stirring of product B is created. Different processes may contain common analysis policies and operating methods, so by making two or more unit operations included in different types of processes belong to the same group, it is possible to improve the efficiency of plants that handle a wide variety of products. However, it becomes easier to secure the data necessary for analysis.

Further, when information and condition information of unit operations at a plurality of different bases are stored in the memory, a characteristic group may be created such that two or more unit operations at different bases belong to the same characteristic group. For example, a characteristic group is created that includes agitation when product A is produced at base L1 and agitation when product A is produced at base L2. Since analysis policies and operating methods may be common even at different bases, by ensuring that two or more unit operations at different bases belong to the same characteristic group, even in plants that handle a wide variety of products. , it becomes easier to secure the data necessary for analysis.

Furthermore, characteristic groups may be created based on typical characteristic amounts in unit operations. This is because when important feature quantities are common, common analysis policies and driving methods are included in the creation of predictive models. For example, characteristic groups may be created such that unit operations in which temperature is important in operating conditions belong to the same characteristic group.

The calculation device 11 is also capable of updating condition information. Once performance data that is a result of unit operation processing is added to the log data 21, condition information can be updated based on the added log data 21 and the prediction model.

Furthermore, the calculation device 11 can also output operating conditions for a unit operation based on the created prediction model and sensor data. For example, a data set created based on condition information is input into a prediction model, and information on the results of prediction is output based on target quality information and information on recommendations for changing operating conditions. Good too. Alternatively, the optimum conditions may be searched based on the prediction model, sensor data, and target quality information.

Further, the calculation device 11 can output information for predicting the result of processing in a unit operation by inputting information on operating conditions into a process model and performing calculations. In this way, by simulating the process using the process model, various evaluations can be performed before the actual process is performed.

The arithmetic unit 11 specifies suitable operating conditions for the unit operation based on information predicting the results of processing in the unit operation, and issues control instructions to plant equipment related to the unit operation based on the identified operating conditions. It is also possible to create a control instruction containing the information and send the control instruction to the plant equipment. By transmitting control instructions in this way, automatic plant operation can be realized.

Further, the user interface 13 can receive input of arbitrary operating conditions for unit operations, and the calculation device 11 inputs information on the received operating conditions into the process model and performs calculations to obtain processing results for the unit operations. It is possible to determine whether the output result is abnormal based on the reference information 26. When determining that there is an abnormality, the arithmetic unit 11 stores the operating condition in the abnormality related information 27 in association with the unit operation. In this way, by simulating the occurrence of an abnormality, it is possible to accumulate conditions under which an abnormality occurs.

FIG. 5 is an explanatory diagram of the operation of the plant management device. The plant management device performs operations in the order of unit operation understanding S101, objective variable setting S102, data selection S103, feature amount addition S104, learning period setting S105, and software sensor analysis execution S106.

In unit operation understanding S101, the AI learns the conditions registered in the condition information database 22. Specifically, learning by AI may be realized by calculation by the calculation device 11.
Objective variable setting S102 accepts designation of an analysis target. For example, if the quality of a product is specified as an analysis target, the quality of the product becomes the objective variable. In this case, the predictive model becomes a software sensor that obtains the quality of the product through software calculations.

In data selection S103, AI automatically selects data that is closely related to the analysis target from both the data analysis aspect and condition information.
In feature amount addition S104, the AI automatically adds feature amounts effective for soft sensing of the analysis target based on the condition information.
In learning period setting S105, the AI automatically selects an effective analysis period for soft sensing to be analyzed based on the condition information.
In soft sensor analysis execution S106, the AI learns or predicts the target analysis target using the selected data and period.

FIG. 6 is an explanatory diagram of a specific example of grouping.
Grouping based on unit operations includes groupings such as ``reactions with similar theoretical models are grouped together'' and ``fractional distillation, general distillation, and vacuum distillation are integrated and grouped by distillation category.'' It is possible.
In grouping based on characteristics, it is possible to "group by fluid (solid, liquid, gas)." For example, it is possible to classify "unit operation α, liquid" into Gr11, and "unit operation α, gas" into Gr12. It is also possible to "group by equipment used." For example, it is possible to group the reactions based on whether the reaction is carried out in a continuous mode or in a batch mode.

Next, the operation of the plant management device will be explained.
FIG. 7 is a flowchart showing operations when creating a prediction model.
The calculation device 11 first executes a sensor data input step of inputting sensor data acquired from a sensor device of the plant (step S201).

Thereafter, the arithmetic device 11 uses the sensor data, the log data 21, and the condition information in the condition information database 22 to generate a data set for use in creating a predictive model (step S202).

The arithmetic device 11 executes machine learning using the generated data set and creates a prediction model for predicting the result of processing in a unit operation (step S203). With the above configuration, it is possible to create a model for analyzing sensor information to be analyzed while taking into account constraints such as analysis policies and analysis conditions specific to unit operations.
Then, the created predictive model is stored in the predictive model database 24 of the auxiliary storage device 20 in association with the unit operation (step S204), and the process ends.
If the number of sensor data accumulated for a specific unit operation is sufficient, only the sensor data for the unit operation can be input in step S201 to create a prediction model. On the other hand, even if the number of sensor data accumulated for a specific unit operation is insufficient, a predictive model can be created by using data accumulated for other unit operations in the group to which the unit operation belongs. be.

FIG. 8 is a flowchart showing the operation of the prediction model during operation.
The computing device 11 first executes a sensor data input step of inputting sensor data acquired from a sensor device of the plant (step S301). Here, if the sensor information to be analyzed is related to an arbitrary first unit operation, the input sensor data (sensor information) is determined based on the condition information related to the first unit operation. It's okay. Thereby, data for creating a data set necessary for analysis can be efficiently and reliably acquired.

After that, the calculation device 11 uses the sensor data and the condition information in the condition information database 22 to generate a data set to be given to the prediction model (step S302). Since the type and period of sensor data required for analysis differ for each unit operation, by generating a dataset to be given to the prediction model based on condition information, only the necessary data is used, resulting in higher accuracy and efficiency. This makes it possible to make accurate predictions.

The arithmetic device 11 executes prediction using the prediction model using the generated data set as an explanatory variable (step S303). At this time, the predictive model used may be a predictive model created only from data of the target unit operation, or a predictive model created using data of a group to which the target unit operation belongs.
Then, the prediction result is output (step S304), and the process ends.

FIG. 9 is an explanatory diagram of automatic operation of the plant. The plant management device generates a data set using the sensor data and condition information acquired by the sensor device. By feeding this data set to a prediction model and simulating the reactor state, it is possible to predict what state the reactor will actually be in. The plant management device identifies suitable operating conditions for a unit operation based on the prediction results, generates control instructions for the plant, and transmits them to the plant equipment. By transmitting control instructions in this way, automatic plant operation can be realized.

Although FIG. 4 shows a state in which the plant at the base L1 employs a continuous system, the plant may also use a batch system. FIG. 10 shows a case where the base L1 adopts the batch method. Since a batch system is adopted, at base L1, multiple unit operations are performed on one plant facility, and then the facility is transferred to the facility for the next process. The sensor device is installed in association with plant equipment. That is, one sensor device may be used in multiple unit operations. Therefore, it is necessary to send the output of the sensor device to the plant management device with information that can specify which unit operation the sensor output corresponds to. The other configurations and operations are the same as those in FIG. 4, so their explanation will be omitted.

In addition, in the explanation so far, a method of creating a prediction model by dividing unit operations into a plurality of groups has been exemplified, but other methods can also be used. For example, the computing device 11 identifies a second unit operation that is different from the first unit operation and is related to the first unit operation, and further includes information on log data related to the second unit operation. It is also possible to create a prediction model for predicting the result of processing in the first unit operation based on the condition information related to the second unit operation and the second unit operation. As an example, by searching for similar unit operations for a certain unit operation and creating a predictive model using log data and condition information accumulated for similar unit operations, the amount of data that can be used to create a predictive model can be increased. Can be done.

As described above, the disclosed plant management device includes information on a plurality of different unit operations included in a treatment process in a chemical plant, information on log data related to the processing of the unit operations, and conditions for each of the unit operations. an auxiliary storage device 20 as a storage unit that stores condition information regarding the sensor; a communication interface 15 as an input unit that inputs sensor information acquired from a sensor device related to a first unit operation in a chemical plant; information, the information of the log data related to the first unit operation, and the condition information, create a prediction model for predicting the result of the processing in the first unit operation, and create a prediction model for predicting the result of the processing in the first unit operation, and a calculation device 11 as a calculation unit that stores the data in the storage unit in association with the first unit operation.
By employing such a configuration and operation, the plant management device can predict the results of processing by the chemical plant, reflecting the characteristics of unit operations.

Further, the calculation unit creates a plurality of characteristic groups each including one or more of the unit operations having characteristics in a predetermined range based on the characteristics of the unit operation, and among the plurality of characteristic groups, the input unit identifying a first characteristic group that includes the first unit operation related to the sensor information input to the sensor information, and relating to the sensor information and one or more unit operations included in the first characteristic group; A prediction model for predicting a processing result in the first unit operation can be created based on the log data information and the condition information.
Therefore, even in a plant that handles a wide variety of products in small quantities, a predictive model can be created efficiently.

Further, the condition information may include information on a predetermined mathematical model related to each of the unit operations.
Further, the characteristic of the unit operation may be information on the type of the unit operation, and the calculation section may create the plurality of characteristic groups based on the information on the type of the unit operation.
Further, the characteristic group may include two or more unit operations included in different types of processes in the same group.
Further, the storage unit stores information on the unit operations at a plurality of different bases and the condition information, and the characteristic group includes two or more unit operations at different bases in the same group. You can.
Further, the characteristic of the unit operation is information indicating a representative feature amount in the unit operation, and the calculation unit creates the plurality of characteristic groups based on the information indicating the representative feature amount. It may be.
In this way, arbitrary conditions can be used as the condition information, and unit operations can be divided into arbitrary groups.

Further, the storage unit further stores first performance data that is a result of the processing of the first unit operation, and the calculation unit combines the created prediction model and the first performance data. The condition information regarding the first unit operation may be updated based on the condition information.
In this way, by updating the condition information using data accumulated through operations, it is possible to improve the accuracy of prediction by the prediction model.

Further, the calculation unit may output the first operating condition in the first unit operation based on the created prediction model and the sensor information.
Furthermore, the storage unit stores a process model that simulates a process including the first unit operation on a computer, and the calculation unit inputs information about the first operating conditions into the process model. The calculation may output information that predicts the result of processing in the first unit operation.
Furthermore, the arithmetic unit specifies a second operating condition in the first unit operation based on information that predicts a result of processing in the first unit operation, and specifies a second operating condition in the first unit operation, and The control instruction to the equipment may include creating a control instruction including the second operating condition and transmitting the control instruction to the plant equipment.
In this way, predictions made by the predictive model can be used for outputting appropriate operating conditions, process simulation, and automatic operation.

Further, the storage unit stores reference information that is reference information indicating that an abnormality has occurred regarding each of the unit operations, and operating conditions when the abnormality occurs in the unit operations, in association with the unit operations. The input unit is capable of accepting an input of an arbitrary second operating condition in the first unit operation, and the calculation unit is configured to store abnormality-related information that is information about the second operating condition. outputting the processing result in the first unit operation by inputting and calculating the information into the process model, determining whether or not the output result is abnormal based on the reference information, and determining whether the output result is abnormal based on the reference information. If determined, the second operating condition may be stored in the abnormality related information in association with the first unit operation.
By employing such a configuration and operation, conditions under which an abnormality occurs can be accumulated by simulating an abnormal state.

Further, the calculation unit identifies a second unit operation that is different from the first unit operation and is related to the first unit operation, and further specifies the second unit operation that is related to the second unit operation. A prediction model for predicting a result of processing in the first unit operation may be created based on log data information and the condition information related to the second unit operation.
In this way, similar unit operations can be identified using any method without being limited to grouping, and a predictive model can be created from information related to different unit operations.

In addition, from another perspective, the disclosed plant management device includes information on a plurality of unit operations included in a treatment process in a chemical plant, information on log data related to processing of the unit operations, and conditions for each of the unit operations. A plurality of characteristic groups each including one or more unit operations having characteristics in a predetermined range are created based on the auxiliary storage device 20 as a storage unit that stores condition information regarding and the characteristics of the unit operations. The computing device 11 includes a processing unit 11 and a communication interface 15 serving as an input unit for inputting sensor information acquired from a sensor device related to a first unit operation in a chemical plant. Among the characteristic groups, a first characteristic group including the first unit operation related to the sensor information input to the input unit is identified, and the sensor information and the first characteristic group included in the first characteristic group are identified. It also operates as a calculation unit that predicts the result of processing in the first unit operation based on the log data information and the condition information related to the above unit operation.
Therefore, the plant management device can predict the results of processing by the chemical plant by reflecting the characteristics of the unit operations.

Further, the disclosed plant management method includes a step of inputting sensor information obtained from a sensor device related to a first unit operation included in a plurality of different unit operations included in a treatment process in a chemical plant; , for predicting the result of processing in the first unit operation based on log data information related to the processing of the first unit operation and condition information that is information regarding conditions in the first unit operation. and storing the predictive model in a predetermined storage device in association with the first unit operation.
Therefore, the plant management method can predict the results of processing by a chemical plant by reflecting the characteristics of unit operations.

Note that the present invention is not limited to the above-described embodiments, and includes various modifications. For example, the embodiments described above are described in detail to explain the present invention in an easy-to-understand manner, and the present invention is not necessarily limited to having all the configurations described. Furthermore, it is possible not only to delete a configuration but also to replace or add a configuration.

11: Arithmetic device, 12: Main storage device, 13: User interface, 14: Display unit, 15: Communication interface, 20: Auxiliary storage device, 21: Log data, 22: Condition information database, 23: Unit operation database, 24 : Prediction model database, 25: Process model database, 26: Standard information, 27: Abnormality related information

Claims (16)

a storage unit that stores information on a plurality of different unit operations included in a treatment process in a chemical plant, information on log data related to processing of the unit operations, and condition information regarding conditions in each of the unit operations;
an input unit for inputting sensor information obtained from a sensor device related to a first unit operation in a chemical plant;
creating a prediction model for predicting a processing result in the first unit operation based on the sensor information, the log data information related to the first unit operation, and the condition information; A plant management device comprising: a calculation unit that stores a predictive model in the storage unit in association with the first unit operation. The plant management device according to claim 1,
The arithmetic unit is
creating a plurality of characteristic groups each including one or more of the unit operations based on the characteristics of the unit operation;
Among the plurality of characteristic groups, a first characteristic group including the first unit operation related to the sensor information input to the input section is identified, and the first characteristic group is associated with the sensor information and the first characteristic group. A plant management device that creates a prediction model for predicting a processing result in a first unit operation based on information of the log data related to one or more of the included unit operations and the condition information. . The plant management device according to claim 1,
The condition information includes information on a predetermined mathematical model related to each of the unit operations. The plant management device according to claim 2,
The characteristic of the unit operation is information on the type of the unit operation,
The plant management device, wherein the calculation unit creates the plurality of characteristic groups based on information on the type of the unit operation. The plant management device according to claim 4,
Further, in the plant management device, the characteristic group includes two or more unit operations included in different types of processes in the same group. The plant management device according to claim 2,
The storage unit stores information on the unit operation and the condition information at a plurality of different bases,
Further, in the plant management device, the characteristic group includes two or more unit operations at different bases in the same group. The plant management device according to claim 2,
The characteristic of the unit operation is information indicating a representative feature amount in the unit operation,
The plant management device, wherein the calculation unit creates the plurality of characteristic groups based on information indicating the representative feature amount. The plant management device according to claim 1,
The storage unit further stores first performance data that is a result of the processing of the first unit operation,
The plant management device, wherein the calculation unit updates the condition information regarding the first unit operation based on the created prediction model and the first performance data. The plant management device according to claim 1,
The plant management device, wherein the calculation unit outputs a first operating condition in a first unit operation based on the created prediction model and the sensor information. The plant management device according to claim 9,
The storage unit stores a process model that simulates a process including the first unit operation on a computer,
The calculation unit outputs information for predicting a result of processing in the first unit operation by inputting information on the first operating conditions into the process model and performing calculations. The plant management device according to claim 10,
The calculation unit specifies a second operating condition in the first unit operation based on information predicting a result of processing in the first unit operation,
A plant management device that creates a control instruction to plant equipment related to the first unit operation, which includes the second operating condition, and transmits the control instruction to the plant equipment. The plant management device according to claim 10,
The storage unit stores reference information that is reference information indicating that an abnormality has occurred regarding each of the unit operations, and information that stores operating conditions when an abnormality occurs in the unit operations in association with the unit operations. It stores information related to certain anomalies,
The input unit is capable of accepting input of an arbitrary second operating condition in the first unit operation,
The calculation unit outputs the processing result in the first unit operation by inputting information on the second operating condition into the process model and calculating it, and determines whether the output result is abnormal based on the reference information. A plant management device that determines whether or not the operation is abnormal, and stores the second operating condition in association with the first unit operation in the abnormality-related information when it is determined that the second operating condition is abnormal. The plant management device according to claim 1,
The arithmetic unit is
identifying a second unit operation that is different from the first unit operation and that is related to the first unit operation;
Further, a prediction model for predicting the result of processing in the first unit operation based on information of the log data related to the second unit operation and the condition information related to the second unit operation. A plant management device that creates The plant management device according to claim 1,
Sensor information acquired from a sensor device related to the first unit operation that is input to the input section is determined based on the condition information related to the first unit operation. a storage unit that stores information on a plurality of unit operations included in a treatment process in a chemical plant, information on log data related to processing of the unit operations, and condition information regarding conditions in each of the unit operations;
a processing unit that creates a plurality of characteristic groups each including one or more of the unit operations having characteristics in a predetermined range, based on the characteristics of the unit operation;
an input unit for inputting sensor information obtained from a sensor device related to a first unit operation in a chemical plant;
Among the plurality of characteristic groups, a first characteristic group including the first unit operation related to the sensor information input to the input section is identified, and the first characteristic group is associated with the sensor information and the first characteristic group. A plant management device comprising: a calculation unit that predicts a result of processing in a first unit operation based on information of the log data related to one or more of the included unit operations and the condition information. inputting sensor information obtained from a sensor device related to a first unit operation included in a plurality of different unit operations included in a treatment process in the chemical plant;
The result of the processing in the first unit operation based on the sensor information, log data information related to the processing of the first unit operation, and condition information that is information regarding conditions in the first unit operation. creating a predictive model for predicting the
A plant management method comprising the step of storing the prediction model in a predetermined storage device in association with the first unit operation.

Images