Control system, control method, and control program

Abstract

To reduce the load on the plant's proper operation. [Solution] An example of a control system comprises at least one processor. The at least one processor determines the operation data by repeatedly calculating, based on observation data showing the actual values ​​of the plant, control target data showing the predicted values ​​of the controlled object in the plant and operation data showing the operation values ​​of the control device of the plant, using a given computational model.

Description

Technical Field

[0001] One aspect of the present disclosure relates to a control system, a control method, and a control program.

Background Art

[0002] Patent Document 1 describes a control device that constructs model predictive control directly using a process simulator. This control device includes a process prediction means for predicting the behavior of a process controlled variable based on each measured value of a disturbance sensor and a process output sensor, a process operation amount command value, and a predicted value of a process disturbance; a step response selection means for selecting a current step response time series based on the process operation amount command value and the measured value of the process disturbance; and a process optimization means for determining an optimal operation amount command value over a predetermined period using at least the predicted process prediction value and the selected step response time series as inputs.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] It is desired to reduce the load associated with the proper operation of a plant.

Means for Solving the Problems

[0005] A control system according to one aspect of the present disclosure includes at least one processor. The at least one processor repeatedly performs a process of calculating, by a given calculation model, control target data indicating a predicted value of a control target in the plant and operation data indicating an operation value of a control device of the plant based on observation data indicating actual values of the plant, and determines the operation data.

[0006] A control method relating to one aspect of this disclosure is performed by a control system comprising at least one processor. This control method includes the step of determining the operation data by repeatedly calculating, based on observation data showing actual values ​​of the plant, control target data showing predicted values ​​of the controlled object in the plant and operation data showing the operation values ​​of the control device of the plant, using a given computational model.

[0007] A control program relating to one aspect of this disclosure causes a computer to perform a step of determining the operation data by repeatedly calculating, based on observation data showing the actual values ​​of the plant, control target data showing the predicted values ​​of the controlled object in the plant and operation data showing the operation values ​​of the control device of the plant, using a given computational model.

[0008] In this respect, the controlled data and operation data are updated through repeated processing by the computational model, and the operation data is ultimately determined. Therefore, the load required for the proper operation of the plant can be reduced. [Effects of the Invention]

[0009] According to one aspect of this disclosure, the load required for the proper operation of the plant can be reduced. [Brief explanation of the drawing]

[0010] [Figure 1] This figure shows an example of the application of a control system. [Figure 2] This is a diagram showing an example of the optimization unit. [Figure 3] This figure shows an example of costs related to plant stability. [Figure 4] This figure shows an example of the implementation of the optimization unit. [Figure 5] This figure shows another example of the optimization implementation. [Figure 6] This diagram shows a typical hardware configuration for a computer that functions as a control system. [Figure 7]This is a flowchart illustrating an example of the control system's operation. [Figure 8] Here is another example of how the control system operates, shown in a flowchart. [Figure 9] This figure shows an example of a user interface. [Modes for carrying out the invention]

[0011] The embodiments described herein will be described in detail below with reference to the attached drawings. In the description of the drawings, the same or equivalent elements are denoted by the same reference numerals, and redundant descriptions are omitted.

[0012] [System Overview] Figure 1 shows an example of the application of the control system 1 according to an embodiment. The control system 1 is a computer system that supports the operation of plant 2. In one example, the control system 1 generates and outputs operation data indicating operations related to plant 2. The control system 1 may output its operation data to plant 2 for the automatic operation of plant 2. Alternatively, the control system 1 may provide its operation data to the operator of plant 2 to support the operator of said plant 2. The control system 1 may be constructed as a component of plant 2, or it may be installed outside of plant 2.

[0013] Plant 2 is a production facility related to industrial activities and comprises a group of devices for producing a given product. Examples of Plant 2 include, but are not limited to, petroleum-related plants such as petroleum refining plants and petrochemical plants. A petroleum-related plant includes, for example, a group of devices for the purpose of petroleum refining or the manufacture of petrochemical products. Such a group of devices includes, for example, at least one of an atmospheric distillation unit, a hydrorefining unit, a catalytic reforming unit, a catalytic cracking unit, a hydrocracking unit, and a desulfurization unit.

[0014] Plant 2 is equipped with a control device 3. The control device 3 is a device that controls the controlled object 4. The controlled object 4 refers to the object that is controlled in the operation of Plant 2. The controlled object 4 may be equipment that makes up Plant 2. Examples of the controlled object 4 include production equipment (tanks, processing furnaces, etc.), industrial machinery, electrical circuits, sensors, communication networks, infrastructure equipment (water systems, smart grids, etc.), and mobile objects (cars, robots, ships, aircraft, etc.). Alternatively, the controlled object 4 may be a product produced by Plant 2. The control device 3 may control physical parameters (temperature, pressure, height, etc.) related to the equipment within Plant 2. If the controlled object 4 is a product, the control device 3 may control the equipment within Plant 2 so that the physical parameters (e.g., concentration, mass, content, etc.) related to that product meet a given standard.

[0015] Traditionally, plants operate through constant monitoring and operation by operators. Therefore, plant operation largely depends on the operator's skill. In recent years, with the decline in skilled operators, the transfer of skills necessary for plant operation has become a challenge. For example, control system 1 outputs operation data indicating the operation values ​​of control device 3 to control device 3, and control device 3 controls the controlled object 4 based on that operation data. This automated operation allows plant 2 to be operated appropriately without being dependent on the operator's skill. Alternatively, control system 1 may provide the operation data to the operator. The operator can then operate control device 3 by referring to that operation data, reducing the operator's workload. Furthermore, it becomes possible for less experienced operators to operate control device 3 in the same way as skilled operators. Regardless of whether the operation of control device 3 is automatic or manual, control system 1 can reduce the workload on plant 2 operation. In addition, control system 1 can contribute to the production efficiency of plant 2 as much as or more than skilled operators.

[0016] [System Configuration] As shown in FIG. 1, the control system 1 is connected to the control device 3 and the database 20 via a given communication network. The communication network may be constructed using at least one of the Internet and the intranet. The communication network may be constructed by a wired network, a wireless network, or a combination thereof.

[0017] The database 20 is a device that stores various data used in the control system 1. The database 20 may be a component of the control system 1, or may be provided in a computer system separate from the control system 1. The database 20 may be a component of the plant 2, or may be provided outside the plant 2. The database 20 stores the observation data indicating the performance values of the plant 2. The observation data is continuously or intermittently recorded by one or more sensors in the plant 2 and stored in the database 20. In one example, the observation data is time-series data for a set of one or more physical parameters. Examples of the physical parameters indicated by the observation data include, but are not limited to, pressure, temperature, and concentration. At least a part of the physical parameters indicated by the observation data may be the same as at least a part of the physical parameters of the operation data, or may be the same as at least a part of the physical parameters of the control target data described later.

[0018] FIG. 1 also shows the functional configuration of the control system 1. In one example, the control system 1 includes an optimization unit 10 as a functional component. The optimization unit 10 is a functional module that repeatedly calculates control target data and operation data using a given calculation model based on observation data, and finally determines the operation data. It can be said that the operation data determined by the optimization unit 10 is the optimal operation data indicating the optimal operation value of the control device 3. In the present disclosure, note that the "optimal operation data (operation value)" refers to operation data (operation value) estimated to be optimal, and is not necessarily the actually optimal operation data (operation value). Hereinafter, in order to distinguish the operation data finally determined by the optimization unit 10 from the operation data calculated during the iterative process, it is also referred to as "optimal operation data".

[0019] The observation data is electronic data indicating the actual performance value of the plant 2. The "actual performance value of the plant" is a value based on actual measurements over a time span from a certain past time point to a given reference time. The actual performance value of the plant may be an actual measured value, a value obtained by performing processing such as outlier removal and interpolation on the measured value, or a relative value from these values at a certain time point. In the present disclosure, the reference time is, for example, the current time.

[0020] The control target data is electronic data indicating the predicted value of the control target 4 in the plant 2. The "predicted value of the control target" is a value indicating the state of the control target 4 over a given time span after a given reference time. The predicted value of the control target may be represented by a relative value or a statistic from the predicted value at a certain time point. In one example, the control target data is time-series data for a set of one or more physical parameters. In another example, the control target data is the average value of a certain physical parameter over a certain time span. The control target data may indicate at least one of physical parameters related to the devices in the plant 2 (such as temperature, pressure, height, etc.) and physical parameters related to the product (such as concentration, mass, content, etc.).

[0021] Operational data is electronic data representing the operational values ​​of the control device 3. "Operational values ​​of the control device" refers to values ​​based on the operation of the control device 3 over a given time interval after a given reference time. The operational values ​​of the controlled object may be expressed as relative values ​​from the operational values ​​at a given point in time, or as statistical values ​​of the operational values. Therefore, the operational values ​​represented by operational data are also predicted values. The time interval of the operational data may be the same as or different from the time interval of the controlled object data. In one example, the operational data is time-series data of a set of one or more physical parameters. In another example, the operational data is the average value of a certain physical parameter over a certain time interval. Examples of physical parameters represented by operational data include, but are not limited to, flow rate, set temperature, and set pressure.

[0022] In one example, the calculation model includes a prediction model 11 and a decision model 12. The prediction model 11 is a calculation model that calculates the controlled data based on observed data and the most recent operational data in each iteration. The prediction model 11 may accept future disturbances in addition to observed and operational data. Future disturbances are data that indicates disturbance factors not included in the operational data, such as the amount of inflow into plant 2 or the outside temperature. Future disturbances are predicted values. Future disturbances may be set by the user or predicted by the prediction model 11. Alternatively, the observed values ​​may be set as future disturbances, assuming that the currently observed disturbances will continue. The decision model 12 is a calculation model that calculates operational data based on the most recent controlled data in each iteration. The decision model 12 may accept the most recent operational data in each iteration, in addition to or instead of the most recent controlled data.

[0023] Figure 2 shows an example of the optimization unit 10 using a prediction model 11 and a decision model 12. In this figure, the reference time for the observed data, operation data, and controlled data is the current time. The optimization unit 10 inputs the observed data and initial operation data into the prediction model 11 to calculate the controlled data, and then inputs this controlled data into the decision model 12 to calculate the operation data. This series of processes using the prediction model 11 and the decision model 12 corresponds to one loop. In the second loop, the optimization unit 10 inputs the observed data and the calculated operation data into the prediction model to calculate new controlled data. The optimization unit 10 inputs this controlled data into the decision model 12 to calculate new operation data. This operation data is used in the third loop.

[0024] In short, the optimization unit 10 calculates new target data based on the observed data read from the database 20 and the most recent operation data in each iteration. The "most recent operation data" in the i-th loop is the operation data obtained in the (i-1)th loop. The optimization unit 10 calculates new operation data based on this new target data, i.e., the most recent target data in each iteration. In this iteration, the optimization unit 10 uses the observed data read from the database 20 in each loop. The fact that new target data and new operation data are calculated in each loop means that the target data and operation data are updated by the iteration. The optimization unit 10 repeats the process using the prediction model 11 and the decision model 12 until a given termination condition is met, and finally determines the obtained operation data as the optimal operation data.

[0025] In one example, the prediction model 11 may include a neural network that accepts observed data and manipulated data as inputs and outputs controlled data. In this case, the prediction model 11 is a trained model. An example of a neural network is a recurrent neural network (RNN), but it is not limited to this. Suppose the number of physical parameters for the observed data, manipulated data, and controlled data are p, q, and r, respectively. In this case, the neural network may accept p physical parameters or time series data of the observed data and q physical parameters or time series data of the manipulated data as inputs and output one physical parameter or time series data of the controlled data. In this example, the prediction model 11 predicts the entire controlled data by including r neural networks corresponding to the r physical parameters or time series data of the controlled data. Alternatively, a single neural network may accept p physical parameters or time series data of the observed data and q physical parameters or time series data of the manipulated data as inputs and output r physical parameters or time series data of the controlled data.

[0026] In machine learning to obtain the predictive model 11, training data corresponding to the observed data, controlled data, and manipulated data are prepared in advance. A given machine learning system generates the predictive model 11 using this training data. In one example, the machine learning system performs the following processing for each record of the training data: The machine learning system inputs the observed and manipulated data shown in that record into the machine learning model and obtains an estimate of the controlled data output from the neural network. Based on the error between this estimate and the correct answer for the controlled data shown in that record, the machine learning system updates the parameters within the machine learning model using a method such as backpropagation. For example, the machine learning system updates the weights of the neural network. The machine learning system continues machine learning until a given termination condition is met. For example, the machine learning system may evaluate the performance of the machine learning model using validation data and terminate machine learning if the evaluation meets a given criterion. Alternatively, the termination condition may be set based on the error, or based on the number of records processed, i.e., the number of learning iterations. The machine learning system provides the machine learning model at the time machine learning is terminated as the predictive model 11.

[0027] The machine learning system may be part of control system 1, or it may be a separate computer system from control system 1. Since trained models are portable between computer systems, control system 1 can utilize a predictive model 11 generated on another computer system.

[0028] In one example, decision model 12 includes nonlinear optimization using a cost function to calculate the cost for plant 2. Decision model 12 applies at least one of the controlled data and the operational data to its cost function to calculate the cost, and then calculates the operational data based on this cost. The cost is a value that indicates the evaluation of at least one of the controlled data and the operational data. In one example, decision model 12 calculates new operational data through nonlinear optimization to reduce its cost. This calculation can be described as a process of searching for more optimal operational data. In other examples, decision model 12 may be optimized to increase its cost.

[0029] In one example, decision model 12 uses a function to calculate the cost of at least one of the profitability and stability of plant 2. In this disclosure, the cost related to profitability is also referred to as the profitability cost, and the cost related to stability is also referred to as the stability cost. "Plant profitability" is an indicator of how efficiently profits can be generated by operating the plant. A low profitability cost means high profitability, and a high profitability cost means low profitability. "Plant stability" is an indicator of the plant's ability to maintain a state in which proper operation is guaranteed, or to quickly return the plant to such a state. A low stability cost means high stability, and a high stability cost means low stability.

[0030] In one example, profitability costs are set based on the profit calculated from the cost of the products produced by Plant 2, the cost of raw materials, and the utility costs required to operate the processing steps of the manufacturing equipment.

[0031] In one example, the stability cost is determined based on at least one of the following: the degree of deviation of the controlled data from a target setpoint for the controlled object 4, and the amount of variation in the operational data. Figure 3 shows an example of the stability cost. Examples (a) and (b) in Figure 3 are graphs of the stability cost based on the degree of deviation of the controlled data. In these graphs, the vertical axis shows the value of a certain physical parameter related to the controlled object 4 as the controlled value, and the horizontal axis shows the time axis from the reference point. Example (c) in Figure 3 is a graph of the stability cost based on the amount of variation in the operational data. In this graph, the vertical axis shows the value of a certain physical parameter as the operational value, and the horizontal axis shows the time axis from the reference point.

[0032] In example (a) of Figure 3, the set value for the controlled object 4 is represented by a control range W defined by a given upper and lower limit. In this example, the optimization unit 10 sets measurement points at given time intervals from the reference time and calculates the deviation Ep of the controlled object data from the control range W at each measurement point. The given time intervals may be uniform or different. If the controlled object value is within the control range W at a given measurement point, the deviation at that measurement point is 0. In one example, the cost function is the sum of the squares of the individual deviations, and the optimization unit 10 calculates the cost using this cost function. In example (a), the control range W is fixed, but the control range W may vary along the time axis.

[0033] In example (b) of Figure 3, the set value for the controlled object 4 is represented by a given reference value R. In this example, the optimization unit 10 sets measurement points at given time intervals from the reference time and calculates the deviation degree Eq of the controlled object data from the reference value R at each measurement point. The given time intervals may be uniform or different. In one example, the cost function is the sum of the squares of the individual deviation degrees, and the optimization unit 10 calculates the cost using this cost function. In example (b), the reference value R is fixed, but the reference value R may fluctuate along the time axis.

[0034] As shown in the examples (a) and (b) of Figure 3, in one example, the optimization unit 10 calculates the degree of deviation of the controlled object data from the set value for the controlled object 4 at one or more measurement points on the time axis, and calculates the cost based on the multiple degree of deviation. The optimization unit 10 may also calculate the cost based on a cost function other than the sum of squares.

[0035] In example (c) of Figure 3, the optimization unit 10 sets measurement points at given time intervals from the reference time and calculates the amount of change Vc of the operation data between adjacent or any two measurement points. The measurement points may be set at the reference time. The given time intervals may be uniform or different. In one example, the cost function is the sum of the squares of the individual changes, and the optimization unit 10 calculates the cost using this cost function.

[0036] Figures 4 and 5 both show examples of the implementation of the optimization unit 10, which takes into account at least one of the profitability and stability of plant 2.

[0037] In the implementation 31 shown in Figure 4, the optimization unit 10 includes one prediction model 11 and one decision model 12. The decision model 12 calculates costs using a cost function that considers at least one of the profitability cost group and the stability cost group. The optimization unit 10 receives the input of observation data Da and repeatedly processes using the prediction model 11 and the decision model 12 to determine the optimal operation data Dc. The optimization unit 10 may further input the observation data Da and the optimal operation data Dc into the prediction model 11 to determine the controlled target data Dt corresponding to the optimal operation data Dc. This controlled target data Dt represents the predicted value for the controlled target 4 when the control device 3 is operated based on the operation data Dc. In implementation 31, the optimal operation data Dc is obtained through such a simple iterative process.

[0038] In the implementation 32 shown in Figure 5, the optimization unit 10 comprises one prediction model 11 and two decision models 12. These two decision models 12 are distinguished as the first decision model 12a and the second decision model 12b. In implementation 32, the sets of profitability costs and stability costs are divided into a first cost group and a second cost group. The first decision model 12a includes nonlinear optimization using a first cost function for the first cost group. The second decision model 12b includes nonlinear optimization using a second cost function for the second cost group. In one example, the first cost function corresponds to at least profitability costs, and the second cost function corresponds to at least stability costs. The first cost group may be the group of profitability costs, and the second cost group may be the group of stability costs. Alternatively, the first cost group may consist of the group of profitability costs and a part of the group of stability costs, and the second cost group may consist of the remainder of the group of stability costs. Assume that the cost group related to stability consists of the degree of deviation of the controlled data from the control range (Example 3(a)), the degree of deviation of the controlled data from the reference value (Example 3(b)), and the amount of variation in the operational data (Example 3(c)). In this case, the first cost group may consist of the profitability cost group and the degree of deviation of the controlled data from the control range, and the second cost group may consist of the degree of deviation of the controlled data from the reference value and the amount of variation in the operational data.

[0039] In implementation 32, the optimization unit 10 receives the input of observation data Da and repeatedly processes using the prediction model 11 and the first decision model 12a to determine the control target data Dr, which is used as a setting value for the control target 4. In this disclosure, this iterative process is also referred to as the first process or first optimization. The optimization unit 10 sets the control target data Dr as a setting value for calculating the cost in the second decision model 12b. For example, the optimization unit 10 sets the control target data Dr as a reference value. The control target data Dr may be the reference value R in example (b) of Figure 3.

[0040] Next, the optimization unit 10 receives the input of the observed data Da and repeatedly processes using the prediction model 11 and the second decision model 12b to determine the optimal operation data Dc. In this disclosure, this iterative process is also referred to as the second process or second optimization. In the second process, the optimization unit 10 calculates the cost in the second decision model 12b based on the set value of the controlled data Dr and the second cost function, and calculates the operation data based on this cost. The optimization unit 10 may input the observed data Da and the optimal operation data Dc into the prediction model 11 to determine the controlled data Dt corresponding to the optimal operation data Dc. The controlled data Dt is expected to approximate the set value of the controlled data Dr. In implementation 32, the cost calculation is divided into multiple cost functions, so that each cost function can be simplified. As another example, the optimization unit 10 may include one or more prediction models and three or more decision models. The optimization unit 10 receives input from at least one of the following: observed data and controlled data output from other prediction models, and repeatedly processes data using the prediction model and the decision model. Each prediction model outputs controlled data, and each decision model outputs operation data as needed.

[0041] Figure 6 shows a typical hardware configuration of a computer 100 that functions as a control system 1. For example, computer 100 includes a processor 101, main memory 102, auxiliary storage 103, network interface 104, and device interface 105. These components are connected via a bus 106. The processor 101 is an electronic circuit that runs the operating system and application programs. Examples of processors 101 include CPUs, GPUs, FPGAs, and ASICs. The main memory 102 consists of, for example, ROM and RAM, and stores instructions and various data executed by the processor 101. The auxiliary storage 103 consists of, for example, a hard disk or flash memory, and generally stores more data than the main memory 102. The auxiliary storage 103 stores control programs that allow at least one computer to function as a control system 1. The network interface 104 consists of, for example, a network card or wireless communication module, and sends and receives data to and from an external device 110 via a communication network. Examples of external devices 110 include a database 20 and a control device 3. The device interface 105 is an interface for direct connection to an external device 120, such as a USB. Examples of external devices 120 include input devices such as keyboards, mice, and touch panels, and output devices such as monitors and speakers.

[0042] Each functional module of the control system 1 is implemented by loading a control program onto the processor 101 or main memory 102 and executing that program. The control program includes code for implementing each functional module of the control system 1. The processor 101 operates the network interface 104 or external device 120 according to the control program and reads and writes data to the main memory 102 or auxiliary storage 103. Through this process, each functional module of the control system 1 is implemented. The data or database required for the processing may be stored in the main memory 102 or auxiliary storage 103.

[0043] The control system 1 may consist of one or more computers. When multiple computers are used, these computers are connected to each other via a communication network to logically constitute a single control system 1.

[0044] The control program may be provided by being permanently recorded on a tangible recording medium such as a CD-ROM, DVD-ROM, or semiconductor memory. Alternatively, the control program may be provided via a communication network as a data signal superimposed on a carrier wave.

[0045] [System operation] The operation of the control system 1 will be explained with reference to Figures 7 and 8, and the control method according to the embodiment will also be explained. Figure 7 is a flowchart showing an example of the operation of the control system 1 as processing flow S1. Figure 8 is a flowchart showing another example of the operation of the control system 1 as processing flow S2. Processing flows S1 and S2 correspond to the implementations 31 and 32 described above, respectively.

[0046] First, let's explain the processing flow S1. In step S11, the optimization unit 10 acquires observation data. In one example, the optimization unit 10 accesses the database 20 and reads observation data for a given time range from a certain point in the past to a reference time. The reference time and time range may be set in advance by the control system 1 or specified by the user of the control system 1.

[0047] In step S12, the optimization unit 10 sets the initial operation data. As described above, the optimization unit 10 inputs the observation data and operation data into the prediction model 11 to calculate the data to be controlled, but the operation data is not provided by the decision model 12 in the first processing by the prediction model 11. Therefore, it is necessary to set the initial operation data for the processing of the first loop. In one example, the optimization unit 10 sets the initial operation data for a given time range after the reference time. The optimization unit 10 may also set the initial operation data to indicate that the operation value is 0 for the entire time range. An operation value of 0 means that no operation is performed. In another example, optimized operation data may be set as the initial operation data.

[0048] In step S13, the optimization unit 10 inputs the observed data and the most recent operational data into the prediction model 11 to calculate the data to be controlled. The most recent operational data is the initial operational data in the first loop, and in the second loop and beyond, it is the operational data calculated by the decision model 12.

[0049] In step S14, the optimization unit 10 inputs the calculated controlled target data, i.e., the most recent controlled target data, into the decision model 12 to calculate the operation data. Depending on the cost calculation method, the optimization unit 10 may also input the operation data obtained in the previous loop into the decision model 12.

[0050] In step S15, the optimization unit 10 determines whether to terminate the optimization of the operation data. The optimization unit 10 may determine to terminate the optimization if the combination of the prediction model 11 and the decision model 12 is executed a given number of times, and to continue the optimization otherwise. Alternatively, the optimization unit 10 may determine to terminate the optimization if the amount of change in cost calculated in the decision model 12 (for example, the difference between the cost between two consecutive loops) becomes smaller than a given threshold, and to continue the optimization otherwise.

[0051] If optimization is to continue (NO in step S15), the process returns to step S13 and the next loop process is executed. In step S13, the optimization unit 10 inputs the observed data and the operation data calculated in step S14 into the prediction model 11 to calculate new control target data. In step S14, the optimization unit 10 inputs this new control target data into the decision model 12 to calculate new operation data.

[0052] If optimization is to be terminated (YES in step S15), the process proceeds to step S16. In step S16, the optimization unit 10 outputs optimal operation data. In one example, the optimization unit 10 transmits the operation data to the control device 3. The control device 3 operates based on the operation data to control the controlled object 4. In this automated operation, the optimization unit 10 may terminate its instructions to the control device 3 when the plant 2 satisfies the given conditions. The optimization unit 10 may output the operation data to the control device 3 and also output the operation data to a given output device. The plant 2 operator can perceive the optimal operation data and understand how the control device 3 will operate. Alternatively, the optimization unit 10 may output the operation data to an output device without transmitting it to the control device 3, thereby supporting manual operation by the plant 2 operator. The operator can operate the control device 3 by referring to the operation data. Regardless of whether the operation of the control device 3 based on the operation data is automatic or manual, it can be expected that the controlled object 4 will reach the desired state as a result of the operation data.

[0053] Next, we will explain the processing flow S2. In step S21, the optimization unit 10 acquires observation data. This process is the same as in step S11.

[0054] In step S22, the optimization unit 10 sets the initial operation data. This process is the same as in step S12.

[0055] In step S23, the optimization unit 10 inputs the observed data and the most recent operational data into the prediction model 11 to calculate the data to be controlled. The most recent operational data is the initial operational data in the first loop, and in the second loop and beyond, it is the operational data calculated by the first decision model 12a.

[0056] In step S24, the optimization unit 10 inputs the calculated controlled target data, i.e., the most recent controlled target data, into the first decision model 12a to calculate the operation data. Depending on the cost calculation method, the optimization unit 10 may also input the operation data obtained in the previous loop into the first decision model 12a.

[0057] In step S25, the optimization unit 10 determines whether to terminate the first optimization. The optimization unit 10 may determine to terminate the first optimization if the combination of the prediction model 11 and the first decision model 12a has been executed a given number of times, and to continue the first optimization otherwise. Alternatively, the optimization unit 10 may determine to terminate the first optimization if the amount of change in cost calculated in the first decision model 12a falls below a given threshold, and to continue the first optimization otherwise.

[0058] If the first optimization is to be continued (NO in step S25), the process returns to step S23 and the next loop process is executed. In step S23, the optimization unit 10 inputs the observed data and the operation data calculated in step S24 into the prediction model 11 to calculate new control target data. In step S24, the optimization unit 10 inputs this new control target data into the first decision model 12a to calculate new operation data.

[0059] If the first optimization is terminated (YES in step S25), the process proceeds to step S26. In step S26, the optimization unit 10 sets the set value of the controlled object in the second decision model 12b. In one example, the optimization unit 10 inputs the operation data and observation data finally obtained by the first optimization into the prediction model 11 to calculate the controlled object data. The optimization unit 10 sets this controlled object data in the second decision model 12b.

[0060] In step S27, the optimization unit 10 sets the initial operation data. The optimization unit 10 may set the initial operation data to indicate that the operation value is 0 over the entire given time interval. Alternatively, the optimization unit 10 may set the operation data finally obtained in the first optimization as the initial operation data.

[0061] In step S28, the optimization unit 10 inputs the observed data and the most recent operational data into the prediction model 11 to calculate the data to be controlled. The most recent operational data is the initial operational data in the first loop, and in the second loop and beyond, it is the operational data calculated by the second decision model 12b.

[0062] In step S29, the optimization unit 10 inputs the calculated controlled target data, i.e., the most recent controlled target data, into the second decision model 12b to calculate the operation data. Depending on the cost calculation method, the optimization unit 10 may also input the operation data obtained in the previous loop into the second decision model 12b.

[0063] In step S30, the optimization unit 10 determines whether to terminate the second optimization. The optimization unit 10 may determine to terminate the second optimization if the combination of the prediction model 11 and the second decision model 12b is executed a given number of times, and to continue the second optimization otherwise. Alternatively, the optimization unit 10 may determine to terminate the second optimization if the amount of change in cost calculated in the second decision model 12b becomes smaller than a given threshold, and to continue the second optimization otherwise.

[0064] If the second optimization is to be continued (NO in step S30), the process returns to step S28 and the next loop process is executed. In step S28, the optimization unit 10 inputs the observed data and the operation data calculated in step S29 into the prediction model 11 to calculate new control target data. In step S29, the optimization unit 10 inputs this new control target data into the second decision model 12b to calculate new operation data.

[0065] If the second optimization is to be terminated (YES in step S30), the process proceeds to step S31. In step S31, the optimization unit 10 outputs the optimal operation data. This process is the same as in step S16.

[0066] Both processing flows S1 and S2 can be repeatedly executed at a given time interval. The given time intervals may be uniform or different. For example, before the next processing flow S1 or S2 is executed, the observation data for that time interval is stored in the database 20, and the reference time advances by the amount of that time interval. The control system 1 executes the next processing flow S1 or S2 according to the elapsed time.

[0067] The optimization unit 10 may output the controlled object data corresponding to the optimal operation data to the output device. In this disclosure, this controlled object data is also referred to as "optimal controlled object data." In this disclosure, "optimal controlled object data" refers to the controlled object data that is estimated to be optimal, and should be noted that it is not necessarily the optimal controlled object data in reality. The optimization unit 10 inputs the observation data and the optimal operation data into the prediction model 11 to calculate the optimal controlled object data. The operator of plant 2 can perceive the optimal controlled object data and understand how the state of the controlled object 4 will change.

[0068] An example of outputting controlled data will be explained with reference to Figure 9. Figure 9 is a diagram showing an example of a user interface that displays controlled data. For example, the optimization unit 10 displays the user interface 200 on the display device. The user interface 200 shows the time-dependent change of one physical parameter related to the controlled object 4. The vertical axis of the graph shows the controlled value, which is the value of that physical parameter. The lower limit line 201 and the upper limit line 202 indicate the management range of the controlled value. It is desirable that the controlled value is within this management range. The horizontal axis of the graph shows the time axis, and the baseline line 203 indicates the reference time. Graph 211 shows the time-series data of the actual values ​​of the physical parameter from 11:25 to the reference time of 11:55, and is obtained from observation data. Graph 212 shows the time-series data of the predicted values ​​of the physical parameter after the reference time, and is obtained from the optimal controlled data. Graph 213 shows the predicted values ​​of the physical parameter when the control device 3 is not operated after the reference time. Graph 213 is obtained, for example, by inputting observation data and operation data indicating that the operation value after the reference time is 0 into the prediction model 11. Graph 212 shows the first controlled target data corresponding to the optimal operation data. Graph 213 can be said to show an example of second controlled target data corresponding to comparison operation data different from the optimal operation data. The optimization unit 10 may display graphs 211, 212, and 213 simultaneously or switch between them. From the results shown on the user interface 200, it can be seen that if the control device 3 is operated based on the optimal operation data, the controlled target 4 can be transitioned to an appropriate state, but if that operation is not performed, such appropriate control cannot be achieved. The optimization unit 10 may further display the optimal operation data on the user interface. That is, at least one processor may simultaneously or switch between displaying a user interface on the display device that shows the determined operation data, the first controlled target data corresponding to the determined operation data, and the second controlled target data corresponding to comparison operation data different from the determined operation data.

[0069] [effect] As described above, a control system relating to one aspect of this disclosure comprises at least one processor. The at least one processor determines the operation data by repeatedly calculating, using a given computational model, control target data indicating predicted values ​​of the controlled object in the plant and operation data indicating the operation values ​​of the control device of the plant, based on observation data indicating actual values ​​of the plant.

[0070] A control method relating to one aspect of this disclosure is performed by a control system comprising at least one processor. This control method includes the step of determining the operation data by repeatedly calculating, based on observation data showing actual values ​​of the plant, control target data showing predicted values ​​of the controlled object in the plant and operation data showing the operation values ​​of the control device of the plant, using a given computational model.

[0071] A control program relating to one aspect of this disclosure causes a computer to perform a step of determining the operation data by repeatedly calculating, based on observation data showing the actual values ​​of the plant, control target data showing the predicted values ​​of the controlled object in the plant and operation data showing the operation values ​​of the control device of the plant, using a given computational model.

[0072] A plant relating to one aspect of this disclosure comprises the above-described control system and control device.

[0073] In this respect, the controlled data and operation data are updated through repeated processing by the computational model, and the operation data is ultimately determined. Therefore, the load required for the proper operation of the plant can be reduced. In the technology described in Patent Document 1 above, it is necessary to conduct tests on the actual machine or process simulator in advance to obtain the step response time series, which places a burden on the preparation and requires improvement of the accuracy of the process simulator. In one respect of this disclosure, it is not necessary to prepare the step response time series, so the load required to construct the computational model can be reduced and the simulator can be quickly updated in accordance with the operating conditions of the plant.

[0074] In control systems relating to other aspects, the computational model may include a prediction model that calculates controlled data based on observational data and operational data, and a decision model that calculates operational data based on controlled data, and at least one processor may repeatedly process using the prediction model and the decision model to determine the operational data. By iterative processing using such two types of models, the operational data can be estimated with high accuracy.

[0075] In control systems relating to other aspects, the predictive model may include a neural network that accepts observation data and operation data as inputs and outputs controlled data. By employing this neural network, controlled data for obtaining operation data can be obtained automatically, so operation data can be determined without having to prepare controlled data in advance.

[0076] In control systems relating to other aspects, the decision model may include nonlinear optimization using a cost function to calculate plant costs, and at least one processor may calculate operational data based on the costs calculated by the cost function. By employing this nonlinear optimization, operational data can be estimated with high accuracy.

[0077] In control systems relating to other aspects, at least one processor may calculate a stability cost related to plant stability as a cost related to the plant. In this case, operational data that takes plant stability into account can be obtained.

[0078] In control systems relating to other aspects, at least one processor may calculate the degree of deviation of the controlled object data from the setpoint for the controlled object as a stability cost. By using this degree of deviation, operational data that takes plant stability into account can be obtained.

[0079] In control systems relating to other aspects, the set values ​​for the controlled object are defined by upper and lower limits, forming a control range, and at least one processor may calculate the degree of deviation of the controlled object data from the control range. By considering this degree of deviation, it is possible to obtain operation data that takes into account the stability of the plant while allowing for margin in plant operation.

[0080] In control systems relating to other aspects, the set value for the controlled object is a reference value, and at least one processor may calculate the degree of deviation of the controlled object data from the reference value. By using this degree of deviation, operational data can be obtained to further stabilize the controlled object.

[0081] In control systems relating to other aspects, at least one processor may calculate the degree of deviation at each of multiple measurement points on the time axis and calculate a cost based on multiple degree of deviation. Using this cost, operational data that takes into account the stability of the plant over a given time span can be obtained.

[0082] In control systems relating to other aspects, at least one processor may calculate the amount of variation in the operation data as a stability cost. By considering this amount of variation, operation data that smooths the operation of the control device can be obtained.

[0083] In control systems relating to other aspects, at least one processor may calculate the amount of variation between measurement points based on multiple measurement points on the time axis, and calculate the cost based on the multiple amounts of variation. By considering the amount of variation at a given interval along the time axis, operation data that smooths the operation of the control device over a given time width can be obtained.

[0084] In control systems relating to other aspects, at least one processor may calculate a profitability cost related to the plant's profitability as a cost related to the plant. In this case, operational data that takes the plant's profitability into account can be obtained.

[0085] In control systems relating to other aspects, the decision model may include a first decision model that includes nonlinear optimization using a first cost function and a second decision model that includes nonlinear optimization using a second cost function, and at least one processor may repeatedly perform a first process using a prediction model and the first decision model, and a second process using the prediction model and the second decision model to determine the operational data. By dividing the decision model, the cost function in each decision model can be simplified.

[0086] In control systems relating to other aspects, at least one processor may repeat the first process to determine the control target data to be used as a set value for the control target, set the determined control target data in the second decision model, and repeat the second process using the prediction model and the second decision model to determine the operation data. Since the target in the second decision model is automatically set by the set value, there is no need to prepare that target. This reduces the load on the proper operation of the plant.

[0087] In control systems relating to other aspects, at least one processor may calculate the cost in the second decision model based on the determined controlled data and a second cost function, and then calculate the operation data based on the calculated cost. In this case, the target for calculating the cost in the second decision model can be automatically set.

[0088] In control systems relating to other aspects, at least one processor may simultaneously or alternately display a user interface on a display device showing a first controlled object data corresponding to determined operation data and a second controlled object data corresponding to comparison operation data different from the determined operation data. This user interface can inform the user how the predicted value of the controlled object differs when the determined operation data is used versus when it is not.

[0089] In control systems relating to other aspects, at least one processor may output determined operation data to a control device, and the control device may control the object being controlled based on the determined operation data. This process enables the plant to be controlled appropriately and automatically.

[0090] [Differentiation] The embodiments described above have been explained in detail. However, the disclosure is not limited to the embodiments described above. The disclosure can be modified in various ways without departing from its essence.

[0091] In this disclosure, the expression “transmits” data or information from the first computer to the second computer, or any equivalent expression, means a transmission to ultimately deliver the data or information to the second computer. Note that this expression may also mean that another computer or communication device relays the data or information during the transmission.

[0092] In this disclosure, the expression "at least one processor executes a first process, a second process, ... and the nth process," or a corresponding expression, refers to a concept that includes cases where the entity executing the n processes from the first process to the nth process changes along the way. In other words, this expression refers to a concept that includes both cases where all n processes are executed by the same processor and cases where the processor changes at an arbitrary rate for the n processes.

[0093] The processing steps of the method executed by at least one processor are not limited to the examples in the embodiments. For example, some of the steps (processes) described above may be omitted, or each step may be performed in a different order. Also, any two or more of the steps described above may be combined, or some of the steps may be modified or deleted. Alternatively, other steps may be performed in addition to each of the above steps. [Explanation of symbols]

[0094] 1...Control system, 2...Plant, 3...Control device, 4...Controlled object, 10...Optimization unit, 11...Prediction model, 12...Decision model, 12a...First decision model, 12b...Second decision model, 20...Database, 200...User interface.

Claims

1. Equipped with at least one processor, The at least one processor calculates first operation data indicating the operation value of the control device that controls the controlled object by repeating the first process and the second process. The first processing of the i-th loop involves inputting the first observation data and the first operation data calculated in the (i-1)th loop into a prediction model including a neural network to calculate first controlled target data that indicates the predicted value of the controlled target in the i-th loop. The second process of the i-th loop involves inputting the first control target data calculated in the i-th loop into the first decision model to calculate the first operation data in the i-th loop. Control system.

2. The at least one processor further calculates the second operation data by repeating the third process and the fourth process, The repetition of the third process and the fourth process is performed after the repetition of the first process and the second process is completed. The at least one processor uses the result obtained by repeating the first process and the second process in repeating the third process and the fourth process. The third process in the j-th loop involves inputting the second observation data and the second operation data calculated in the (j-1)th loop into the prediction model to calculate second control target data that indicates the predicted value of the control target in the j-th loop. The fourth process of the j-th loop involves inputting the second control target data calculated in the j-th loop into a second decision model different from the first decision model, thereby calculating the second operation data in the j-th loop. The control system according to claim 1.

3. The initial value of the second operation data is set to zero, or set based on the first operation data. The control system according to claim 2.

4. The first decision model is configured to minimize a cost function corresponding to at least the profitability cost, The second decision model described above is configured to minimize a cost function that corresponds to at least the stability cost. The control system according to claim 2 or 3.

5. The at least one processor further calculates the third operation data by repeating the fifth process and the sixth process, The repetition of the fifth process and the sixth process is performed after the repetition of the third process and the fourth process is completed. The at least one processor uses the result obtained by repeating the third process and the fourth process in repeating the fifth process and the sixth process. The fifth process in the k-th loop involves inputting the third observation data and the third operation data calculated in the (k-1)th loop into the prediction model to calculate third control target data that indicates the predicted value of the control target in the k-th loop. The sixth process of the k-th loop involves inputting the third control target data calculated in the k-th loop into a third decision model different from the first decision model and the second decision model, thereby calculating the third operation data in the k-th loop. The control system according to any one of claims 2 to 4.

6. The at least one processor is configured to further use one or more additional decision models in addition to the first decision model and the second decision model, The first decision model, the second decision model, and the one or more additional decision models constitute a series of decision models. The at least one processor, for each specific decision model among the one or more additional decision models, (a) After the iterations for the preceding decision model in the sequence are completed, the prediction process and the decision process are repeated: (b) The results obtained from the iteration of the preceding decision model are used in the iteration of the specific decision model, (c) Calculate operation data using the specific decision model described above. The control system according to any one of claims 2 to 5.

7. The second observation data is identical to the first observation data. The control system according to any one of claims 2 to 6.

8. The result calculated by repeating the first process and the second process is the first control target data, The above results are used to calculate the cost in the second decision model. The control system according to any one of claims 2 to 7.

9. The at least one processor calculates the first operation data based on the first controlled object data by nonlinear optimization using a cost function, The first decision model includes the nonlinear optimization, A control system according to any one of claims 1 to 8.

10. The at least one processor is configured to simultaneously or in a switching manner display a user interface on a display device showing a first predicted value corresponding to the calculated first operation data and a second predicted value corresponding to comparison operation data different from the calculated first operation data. A control system according to any one of claims 1 to 9.

11. The comparison operation data corresponds to a state in which the controlled object is not operated. The control system according to claim 10.

12. The at least one processor is configured to provide the first operation data to the operator, The control system according to any one of claims 1 to 11.

13. The controlled object is at least one of the following: a plant, an oil-related plant, a refining plant, a petrochemical plant, a group of equipment for petroleum refining, production equipment for petrochemical products, an atmospheric distillation unit, a hydrorefining unit, a catalytic reforming unit, a catalytic cracking unit, a hydrocracking unit, a desulfurization unit, production equipment, a tank, a processing furnace, industrial machinery, an electrical circuit, a sensor, a communication network, infrastructure equipment, a water supply system, a smart grid, a mobile body, an automobile, a robot, a ship, or an aircraft. The control system according to any one of claims 1 to 12.

14. The at least one processor outputs the calculated first operation data to the control device, The control device controls the controlled object based on the calculated first operation data. The control system according to any one of claims 1 to 13.

15. The first observation data corresponds to a first time, The first operation data corresponds to a second time following the first time, A control system according to any one of claims 1 to 14.

16. The first controlled data corresponds to a third time following the first time, The control system according to claim 15.

17. The predicted value is a predicted value relating to the state of the controlled object. The control system according to any one of claims 1 to 16.

18. A control system according to any one of claims 1 to 17, The control device and, A plant equipped with these features.

19. A control method for calculating operation data indicating an operation value of a control device using the control system described in any one of claims 1 to 17.

20. A control program that causes one or more computers to function as a control system according to any one of claims 1 to 17.

Images