Control system, cloud controller, and edge controller

Abstract

To immediately and appropriately compensate for the effects of disturbances in a cloud-based control system.SOLUTION: A control system 10 is equipped with a feedback control unit that calculates a feedback operation amount for a controlled object based on the controlled variable of a controlled object 520 and the target value of the controlled variable, a disturbance candidate value generation unit that generates one or more disturbance candidate values representing the value of disturbances affecting the controlled variable of the controlled object 520, a feed-forward control unit that calculates the feed-forward operation amount to suppress the influence of the disturbance indicated by each disturbance candidate value, an operation amount calculation unit that calculates the operation amount for each disturbance candidate value based on the feedback operation amount and feed-forward operation amount, a disturbance value measurement unit that measures the disturbance value to obtain the actual disturbance value, an operation amount selection unit that selects the operation amount corresponding to the disturbance candidate value with the minimum distance to the actual disturbance value, and an operation unit that operates the selected operation amount on the controlled object 520.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to a control system, a cloud controller, and an edge controller that remotely control a control target. [Background technology]

[0002] In control systems, disturbances caused by unpredictable changes in the state of the equipment or environmental factors can be measured using sensors attached to the site or the equipment being controlled, but feedforward control can be used to compensate for these disturbances. Feedforward control measures disturbances using sensors and adds a manipulated variable in advance to offset the effect of the disturbance on the system, thereby minimizing the effect of the disturbance on the system output.

[0003] For such control systems, there is a growing demand to achieve asset-lightness at the site by running on-premise control systems that have traditionally been implemented using dedicated hardware and operating systems in a cloud environment. When a control system is cloud-based, a cloud controller that determines the amount of operation for the controlled object is generally placed on the cloud. Also, an edge controller that operates actuators and sensors is placed at the site (edge) where the controlled object is installed.

[0004] In this type of configuration, the communication delays that occur in the wired and wireless networks between the cloud controller and the edge controller are larger than in on-premise control systems. As a result, the cloud controller cannot immediately measure disturbances that occur at the edge, and therefore cannot properly compensate for the effects of the disturbances. Patent Documents 1 and 2 describe technologies that address this issue.

[0005] The remote control system described in Patent Document 1 is a remote control system that controls a control object located at a remote location via a communication path with a communication time delay, and is equipped with a communication disturbance estimation means that estimates a communication disturbance in the communication path based on a control signal or a signal equivalent to the control signal transmitted via the communication path and a response signal or a signal equivalent to the response signal transmitted from the control object located at a remote location, and a compensation value generation means that generates a compensation value that compensates for the communication delay based on the communication disturbance estimated by the communication disturbance estimation means, and is characterized in that the communication delay in the remote control system is compensated for by the compensation value generated by the compensation value generation means.

[0006] Furthermore, the remote control device described in Patent Document 2 includes a measurement value receiving unit that receives measurement values ​​related to equipment from a local control device that controls the equipment; a calculation unit that calculates a control value to be used for controlling the equipment when a control delay including a communication delay occurs between the local control device and the equipment, using a model that calculates a control value corresponding to the received measurement value and delay amount from the delay amount corresponding to the control delay and the measurement value; and a control value transmitting unit that transmits the calculated control value to the local control device. [Prior art documents] [Patent documents]

[0007] [Patent Document 1] Re-table No. 2006-046500 [Patent Document 2] Japanese Patent Application Publication No. 2023-174330 Summary of the Invention [Problem to be solved by the invention]

[0008] The control system described in Patent Document 1 compensates for the effects of disturbances using a disturbance observer that estimates the disturbances from the system output. However, because feedback control is performed after the effects of the disturbances appear in the system output, the effects of the disturbances cannot be compensated for immediately, and the effects of the disturbances end up appearing in the system output.

[0009] In the remote control device described in Patent Document 2, the remote control device transmits a control amount calculated for each candidate communication delay amount to a local control device. The local control device measures the actual value of the communication delay and selects a control amount to compensate for the communication delay. However, in a cloud-based control system, the cloud controller cannot immediately measure disturbances occurring at the edge and cannot appropriately compensate for the effects of the disturbances.

[0010] The present invention has been made in consideration of the above background, and aims to provide a control system, a cloud controller, and an edge controller that instantly and appropriately compensate for the effects of disturbances in a cloud-based control system. [Means for solving the problem]

[0011] In order to solve the above-mentioned problems, the control system according to the present invention includes a control variable measurement unit that measures a control variable of a controlled object, a feedback control unit that calculates a feedback manipulation variable for the controlled object based on the control variable and a target value of the control variable, a disturbance candidate value generation unit that generates one or more disturbance candidate values ​​that are values ​​of disturbances that affect the controlled variable of the controlled object, a feedforward control unit that calculates, for each of the disturbance candidate values, a feedforward manipulation variable that suppresses the influence of the disturbance indicated by the disturbance candidate value, a manipulation variable calculation unit that calculates a manipulation variable for each of the disturbance candidate values ​​based on the feedback manipulation variable and the feedforward manipulation variable, a disturbance value measurement unit that measures the value of the disturbance and sets it as an actual disturbance value, a manipulation variable selection unit that selects a manipulation variable corresponding to the disturbance candidate value that is the smallest distance from the actual disturbance value, and an operation unit that operates the selected manipulation variable for the controlled object. [Effects of the Invention]

[0012] According to the present invention, it is possible to provide a control system, a cloud controller, and an edge controller that can immediately and appropriately compensate for the effects of disturbances in a cloud-based control system. Problems, configurations, and effects other than those described above will become clear from the description of the following embodiments. [Brief explanation of the drawings]

[0013] [Figure 1] 1 is an overall configuration diagram of a control system according to an embodiment of the present invention; [Figure 2] FIG. 2 is a functional block diagram of a cloud controller according to the present embodiment. [Figure 3] FIG. 3 is a data configuration diagram of a disturbance actual measurement value database according to the present embodiment. [Figure 4] 10 is an example of a disturbance candidate value generated by a disturbance candidate value generating unit according to the present embodiment. [Figure 5] 10 is an example of a feedforward manipulated variable (W) calculated by the feedforward control unit according to the present embodiment. [Figure 6] 10 is an example of an operation amount calculated by an operation amount calculation unit according to the present embodiment. [Figure 7] FIG. 2 is a functional block diagram of an edge controller according to the present embodiment. [Figure 8] FIG. 3 is a data configuration diagram of an operation amount history database according to the present embodiment. [Figure 9] FIG. 4 is a data configuration diagram of a disturbance actual measurement value corresponding manipulated variable table according to the present embodiment. [Figure 10] 10 is an example of learning data for a distance model 231 according to the present embodiment. [Figure 11] 10 is a flowchart of an operation amount calculation process according to the present embodiment. [Figure 12] 10 is a flowchart of an operation process according to the present embodiment. [Figure 13] 10 is a flowchart of a manipulation amount selection process according to the present embodiment. [Figure 14] FIG. 10 is a hardware configuration diagram illustrating an example of a computer that realizes the functions of the cloud controller and the edge controller according to the above-described embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0014] <Control system overview> An outline of a control system in an embodiment of the present invention will be described below. The control system includes an edge controller installed near (on-site) the facility / device to be controlled, and a cloud controller located in a remote location.

[0015] The cloud controller receives from the edge controller the control variables, which are the state variables of the equipment / device to be controlled, and the actual disturbance measurements. Disturbances are factors that cause changes in the control variables and cannot be accurately predicted. Examples of disturbances include environmental factors such as the temperature around the equipment, state variables of raw materials in production equipment (such as temperature and raw material component ratios), and state variables of the materials and electricity used by the equipment / device. Disturbances can be measured using sensors, and there is not necessarily one type of disturbance.

[0016] The cloud controller calculates a feedback manipulated variable based on the received controlled variable and target value of the controlled variable. The cloud controller also generates one or more candidate values ​​for the disturbance (disturbance candidate value) and calculates a feedforward manipulated variable for each disturbance candidate value to suppress the influence of the disturbance. The cloud controller then calculates a manipulated variable based on the feedback manipulated variable and the feedforward manipulated variable and transmits it to the edge controller together with the disturbance candidate value.

[0017] The edge controller selects the manipulated variable corresponding to the disturbance candidate value that is closest to the actual disturbance value from the received disturbance candidate values, and operates and controls the equipment to be controlled.The edge controller also measures the manipulated variable and disturbance value and sends them to the cloud controller.Note that the disturbance candidate value and the actual disturbance value are not a single value, but a set of one or more values. According to such a control system, it becomes possible to perform the operation (control) of the manipulated variable that is most suitable for the current disturbance without being affected by communication delays.

[0018] <Control system configuration> FIG. 1 is an overall configuration diagram of a control system 10 according to this embodiment. The control system 10 includes a cloud controller 100 and an edge controller 200. The cloud controller 100 and the edge controller 200 are connected via a network 580. The edge controller 200 operates and controls facilities / equipment that are control targets 520 installed at edges 510 (sites). Sensors 570 that measure disturbances and transmit the measurement results to the edge controller 200 are arranged at the edges 510 and the control targets 520.

[0019] The following description will be given taking as an example a control system 10 that controls a heater to maintain the temperature of the water in an aquarium within a specific range. The controlled variable in this control system 10 is the water temperature, and the manipulated variable is the power (W) applied to the heater. Note that the manipulated variable may be another manipulated variable such as starting a pump or opening and closing a valve, or there may be multiple manipulated variables. There may also be multiple controlled variables.

[0020] Note that the control system 10 need not be a water temperature control system, but may also be a railway interlocking control system, a generator control system, a rolling control system, a chemical process control system, or the like. Disturbances in water temperature control include outside air temperature, solar radiation, changes in water volume, and inflow water temperature. Other disturbances include sensor errors and jitter on the time axis during sample / hold. As will be described later, the disturbances in this embodiment are the outside air temperature, solar radiation, and the state of the heating button (see FIG. 3). The actual disturbance value and the candidate disturbance value are a set of the outside air temperature, solar radiation, and the state of the heating button.

[0021] Cloud Controller Configuration 2 is a functional block diagram of the cloud controller 100 according to this embodiment. The cloud controller 100 is a computer, and includes a control unit 110, a storage unit 130, and a communication unit 180. The communication unit 180 includes a communication device, and is capable of transmitting and receiving data to and from the edge controller 200. Note that the cloud controller 100 may be a virtual machine running on a physical computer.

[0022] The storage unit 130 is configured to include storage devices such as a read-only memory (ROM), a random access memory (RAM), and a solid-state drive (SSD). The storage unit 130 stores a disturbance actual measurement value database 140, a disturbance candidate value generation model 131, and a program 138. The program 138 includes a description of the processing of the control unit 110, including the manipulated variable calculation processing (see FIG. 11) described below. Note that the various storage contents of the storage unit 130 may be stored in an external storage device such as a cloud server and read as needed.

[0023] <Cloud controller: memory unit: disturbance measurement value database> 3 is a data configuration diagram of the disturbance actual value database 140 according to this embodiment. The disturbance actual value database 140 is, for example, data in a tabular format, and is time-series data of disturbance actual values, which are measurements of disturbances measured by the edge 510. Each row (record) of the disturbance actual value database 140 includes columns (attributes) of the outside air temperature, amount of solar radiation, and state of the heating button measured at the measurement time.

[0024] The outside temperature and solar radiation amount are the outside temperature and solar radiation amount at the edge 510 at the time of measurement. The heating button is a button operated by an operator at the edge 510. When the heating button is operated, power that is a predetermined value greater than the operation amount calculated by the edge controller 200 (see the operation amount selection unit 214 described below) is supplied to the heater. The state of the heating button is either "1" indicating that it is pressed, or "0" indicating that it is not pressed. The disturbance measurement value database 140 may also include other disturbance measurement values ​​such as water volume and inflow water temperature. In the following, the disturbance measurement value and the disturbance candidate value are defined as a set of the outside temperature, solar radiation amount, and the state of the heating button.

[0025] <Cloud controller: memory unit: disturbance candidate value generation model> Returning to Fig. 2, the description of the storage unit 130 will be continued. The disturbance candidate value generation model 131 is a machine learning model (generative model) generated using records including attributes excluding the measurement time from the disturbance actual measurement value database 140 as learning data. The disturbance candidate value generation model 131 is a generative model using, for example, the TabDDPM learning framework, and is used to generate (calculate) sets of outside air temperature, amount of solar radiation, and state of the heating button (disturbance candidate values) based on random numbers. The disturbance candidate value generation model 131 may also be other generative models such as Generative Adversarial Networks, Variational Autoencoders, or Gaussian Mixture Models.

[0026] Cloud Controller: Control Unit The control unit 110 is configured to include a CPU (Central Processing Unit) and is provided with a target setting unit 111, a controlled variable receiving unit 112, a disturbance actual measurement value receiving unit 113, a disturbance candidate value learning unit 114, a disturbance candidate value generating unit 115, a feedback control unit 116, a feedforward control unit 117, and an operation variable calculating unit 118. The control unit 110 may be configured to include a GPU (Graphics Processing Unit), an NPU (Neural (network) Processing Unit), an FPGA (Field Programmable Gate Array), an ASIC (Application Specific Integrated Circuit), etc.

[0027] <Cloud Controller: Control Unit: Target Setting Unit> The target setting unit 111 sets a target value for the water temperature, which is a control amount of the controlled object 520. The target value may be a single value or may have a range of values. The target value may be specified by a manager of the controlled object 520, or may be specified by an external device / system of the control system 10.

[0028] Cloud controller: control unit: control amount receiver, disturbance actual value receiver The control amount receiving unit 112 receives the control amount transmitted by the edge controller 200 . The disturbance actual measurement value receiving unit 113 receives the disturbance actual measurement value transmitted by the edge controller 200 and the measurement time of the disturbance actual measurement value, and stores them in the disturbance actual measurement value database 140 (see FIG. 3).

[0029] <Cloud controller: control unit: disturbance candidate value learning unit> The disturbance candidate value learning unit 114 generates the disturbance candidate value generation model 131 using records including attributes excluding the measurement time in the disturbance actual measurement value database 140 as learning data. The disturbance candidate value learning unit 114 generates the disturbance candidate value generation model 131 when the number of records in the disturbance actual measurement value database 140 reaches a predetermined value. The predetermined value is, for example, an integer multiple of 1000. The disturbance candidate value learning unit 114 generates the first disturbance candidate value generation model 131 when the number of records reaches 1000. Subsequently, every time the number of records increases by 1000, the disturbance candidate value learning unit 114 generates the disturbance candidate value generation model 131 using all records in the disturbance actual measurement value database 140. The disturbance candidate value learning unit 114 may generate the disturbance candidate value generation model 131 using a predetermined number of recent records instead of all records.

[0030] <Cloud controller: control unit: disturbance candidate value generation unit> The disturbance candidate value generation unit 115 generates one or more pairs of disturbance candidate values, which are the outside temperature, the amount of solar radiation, and the state of the heating button, using the disturbance candidate value generation model 131. FIG. 4 shows an example of disturbance candidate values ​​generated by the disturbance candidate value generation unit 115 according to this embodiment. As shown in FIG. 4, the disturbance candidate value generation unit 115 generates, for example, five pairs of the outside temperature, the amount of solar radiation, and the state of the heating button. It is desirable that the disturbance candidate value generation unit 115 generate a plurality of disturbance candidate values.

[0031] When there is no disturbance candidate value generation model 131, the disturbance candidate value generation unit 115 generates one or more disturbance candidate values ​​according to a predetermined distribution. For example, the disturbance candidate value generation unit 115 may generate a candidate value for each disturbance according to a uniform distribution whose endpoints are the maximum and minimum values ​​of the actual measurement values, or may use another distribution such as a Gaussian distribution. The disturbance candidate value generation unit 115 may also generate a candidate value according to the distribution of the actual measurement values ​​of each disturbance stored in the disturbance actual measurement value database 140. It is desirable that the disturbance candidate value generation unit 115 generate a plurality of disturbance candidate values.

[0032] As described above, the cloud controller 100 (control system 10) includes the disturbance candidate value generator 115 that generates one or more disturbance candidate values, which are values ​​of disturbances that affect the control amount of the control target 520. The disturbance candidate value generating unit 115 generates disturbance candidate values ​​using a disturbance candidate value generating model 131, which is a machine learning model generated using actual disturbance values ​​(see actual disturbance value database 140) as learning data. The disturbance candidate value generating unit 115 generates disturbance candidate values ​​so as to follow the distribution of the actual disturbance values. The disturbance candidate value generating unit 115 generates disturbance candidate values ​​so as to follow a predetermined distribution (for example, a uniform distribution or a Gaussian distribution).

[0033] <Cloud controller: control unit: feedback control unit> Returning to Fig. 2, the description of the control unit 110 continues. Based on the received control variable (water temperature), the feedback control unit 116 calculates a feedback manipulated variable, which is a manipulated variable (power of the heater) that will bring the control variable to a target value. The feedback control unit 116 calculates the feedback manipulated variable, for example, by PID control.

[0034] <Cloud controller: control unit: feedforward control unit> The feedforward control unit 117 calculates a feedforward manipulated variable, which is a manipulated variable to be added to a manipulated variable to suppress the influence of the disturbance for each disturbance candidate value generated by the disturbance candidate value generation unit 115. In this embodiment, the feedforward manipulated variable is the sum of manipulated variables that suppress the influence of the outside air temperature, the amount of solar radiation, and pressing of the heating button. FIG. 5 is an example of the feedforward manipulated variable (W) calculated by the feedforward control unit 117 according to this embodiment. A feedforward manipulated variable is calculated for each disturbance candidate value shown in FIG. 4.

[0035] <Cloud controller: control unit: operation amount calculation unit> 2, the description of the control unit 110 continues. The operation amount calculation unit 118 calculates an operation amount based on the feedback operation amount and the feedforward operation amount, and transmits the calculated operation amount to the edge controller 200. For example, the operation amount calculation unit 118 sets the sum of the feedback operation amount and the feedforward operation amount as the operation amount.

[0036] 6 is an example of the manipulated variable calculated by the manipulated variable calculation unit 118 according to this embodiment. The feedback manipulated variable calculated by the feedback control unit 116 is 1000 (W), and the manipulated variable is the value obtained by adding 1000 to the feedforward manipulated variable (see FIG. 5) for each disturbance candidate value.

[0037] As described above, the cloud controller 100 (control system 10) includes the feedback control unit 116 that calculates the feedback manipulated variable for the control target 520 based on the controlled variable and the target value of the controlled variable. The cloud controller 100 (control system 10) includes a feedforward control unit 117 that calculates, for each disturbance candidate value, a feedforward manipulated variable that suppresses the influence of the disturbance indicated by the disturbance candidate value. The cloud controller 100 (control system 10) includes a manipulated variable calculation unit 118 that calculates a manipulated variable for each disturbance candidate value based on a feedback manipulated variable and a feedforward manipulated variable.

[0038] <Edge controller configuration> 7 is a functional block diagram of the edge controller 200 according to this embodiment. The edge controller 200 is a computer, and includes a control unit 210, a storage unit 230, and a communication unit 280. The communication unit 280 includes a communication device, and is capable of transmitting and receiving data to and from the cloud controller 100, the control target 520, and the sensor 570.

[0039] The storage unit 230 is configured to include storage devices such as a ROM, a RAM, an SSD, etc. The storage unit 230 stores a disturbance measurement value database 240, an operation amount history database 250, a distance model 231, and a program 238. The program 238 includes a description of the processing of the control unit 210, including the operation processing (see FIG. 12) described later.

[0040] <Edge controller: memory unit: disturbance measurement value database> The disturbance actual value database 240 stores the disturbance actual value measured by the sensor 570. The configuration of the disturbance actual value database 240 is similar to the disturbance actual value database 140 of the cloud controller 100 (see FIG. 3).

[0041] <Edge controller: memory unit: operation amount history database> 8 is a data configuration diagram of the operation amount history database 250 according to this embodiment. The operation amount history database 250 is, for example, data in a tabular format, and is time-series data of the operation amount (electric power of the heater) of the operation on the control target 520. Each row (record) of the operation amount history database 250 includes an operation time and a column (attribute) of the operation amount at that operation time.

[0042] <Edge controller: memory unit: distance model> Returning to Fig. 7, the description of the storage unit 230 continues. The distance model 231 is a machine learning model generated using learning data in which two actual disturbance measurements are used as explanatory variables and the difference between the manipulated variables corresponding to the two actual disturbance measurements is used as a target variable. The distance model 231 is a distance model that employs, for example, the Siamese Networks architecture, and is used when calculating the distance between two disturbance values ​​(a set of outside air temperature, amount of solar radiation, and state of the heating button). The two disturbance values ​​are the actual disturbance measurements and a candidate disturbance value transmitted by the cloud controller 100 (see manipulated variable selection unit 214, described later).

[0043] Edge controller: control unit The control unit 210 is configured to include a CPU, and is equipped with a control amount measurement and transmission unit 211, a disturbance value measurement and transmission unit 212, a distance learning unit 213, an operation amount selection unit 214, and an operation unit 215. The control unit 210 may be configured to include a GPU, an NPU, an FPGA, an ASIC, etc.

[0044] <Edge controller: control unit: control amount measurement transmission unit> The control amount measurement transmission unit 211 (control amount measurement unit) measures the control amount (water temperature) of the control target 520 and transmits it to the cloud controller 100. The control amount measurement transmission unit 211 measures and transmits the control amount at a predetermined timing, for example, at a predetermined cycle.

[0045] <Edge controller: control unit: disturbance value measurement transmission unit> The disturbance value measurement transmission unit 212 (disturbance value measurement unit) acquires the actual disturbance value measured by the sensor 570, stores it in the actual disturbance value database 240 together with the measurement time, and transmits it to the cloud controller 100. The disturbance value measurement transmission unit 212 measures, stores, and transmits the value at a predetermined timing, for example, at a predetermined cycle.

[0046] As described above, the edge controller 200 (control system 10) includes a control amount measurement unit (control amount measurement transmission unit 211) that measures the control amount of the control target 520. The edge controller 200 (control system 10) also includes a disturbance value measurement unit (disturbance value measurement transmission unit 212) that measures the value of the disturbance and sets it as an actual disturbance value.

[0047] <Edge controller: control unit: distance learning unit> The distance learning unit 213 generates a distance model 231. More specifically, the distance learning unit 213 generates a disturbance actual measurement value corresponding manipulated variable table 270 (see FIG. 9 described later) that indicates the disturbance actual measurement value and the manipulated variable at that disturbance actual measurement value.

[0048] 9 is a data configuration diagram of the manipulated variable table 270 for actual disturbance values ​​according to this embodiment. The outside temperature, amount of solar radiation, and heating button in the manipulated variable table 270 for actual disturbance values ​​correspond to the outside temperature, amount of solar radiation, and heating button in the actual disturbance value database 240. The manipulated variables are the manipulated variables in the manipulated variable history database 250 (see FIG. 8), and are the manipulated variables at the most recent operation time after and following the measurement of the outside temperature, amount of solar radiation, and heating button. In other words, the manipulated variables are the manipulated variables for the actual disturbance value and the operation of the control target 520 immediately after the measurement of the actual disturbance value, and are manipulated variables corresponding to the actual disturbance value.

[0049] Next, the distance learning unit 213 generates learning data 275 for two records in the disturbance actual measurement value-compatible manipulated variable table 270, using two disturbance actual measurement values ​​as explanatory variables and the difference between the manipulated variables of the records as a response variable. FIG. 10 shows an example of learning data 275 for the distance model 231 according to this embodiment. The first row of the learning data 275 is learning data 275 generated based on the first and second rows of the disturbance actual measurement value-compatible manipulated variable table 270. The distance learning unit 213 generates the distance model 231 using the learning data 275.

[0050] The distance learning unit 213 generates a distance model 231 when the number of records in the disturbance actual value database 240 and the manipulated variable history database 250 reaches a predetermined value. The predetermined value is, for example, an integer multiple of 1000. The distance learning unit 213 generates the first distance model 231 when the number of records in the disturbance actual value database 240 and the manipulated variable history database 250 reaches 1000. Subsequently, every time the number of records increases by 1000, the distance learning unit 213 generates a disturbance candidate value generation model 131 using all records in the disturbance actual value database 140. The distance learning unit 213 may generate the distance model 231 using a predetermined number of recent records instead of all records.

[0051] The distance learning unit 213 may generate the distance model 231 when the number of items in the learning data 275 exceeds an integer multiple of 1000. If the number of records in the disturbance measurement value database 240 and the operation amount history database 250 is X, the number of combinations ( X C2) of the training data 275 can be generated. The distance learning unit 213 may generate the distance model 231 when the number of combinations exceeds an integral multiple of 1000.

[0052] <Edge controller: control unit: operation amount selection unit> Returning to Fig. 7, the explanation of the control unit 210 will be continued. The manipulated variable selection unit 214 (manipulated variable reception unit) selects a manipulated variable corresponding to the disturbance candidate value that is closest in distance to the latest actual disturbance value and is equal to or less than a predetermined distance from among the disturbance candidate values ​​transmitted to the cloud controller 100. More specifically, the manipulated variable selection unit 214 refers to the disturbance actual measurement value database 240 to acquire the latest actual disturbance value. Next, the manipulated variable selection unit 214 selects the disturbance candidate value that is closest in distance to the acquired actual disturbance value. The manipulated variable selection unit 214 calculates the distance between the actual disturbance value and the disturbance candidate value using a distance model 231.

[0053] When there is no distance model 231, the manipulated variable selection unit 214 calculates the distance between the actual disturbance value and the candidate disturbance value using a known distance function. The distance function may be, for example, the Mahalanobis distance, but other distance functions such as the Euclidean distance or the Manhattan distance may also be used. Next, if the distance between the selected disturbance candidate value and the actual disturbance measurement value is equal to or less than a predetermined distance, the manipulated variable selector 214 selects the manipulated variable corresponding to the disturbance candidate value.If the distance is greater than the predetermined distance, the manipulated variable selector 214 selects the previous manipulated variable.

[0054] As described above, the edge controller 200 (control system 10) includes the manipulated variable selector 214 that selects the manipulated variable corresponding to the disturbance candidate value that has the smallest distance from the actual disturbance measurement value. The manipulated variable selection unit 214 calculates the distance between the actual disturbance value and the disturbance candidate value using a distance model 231, which is a machine learning model generated using learning data 275 (see FIG. 10) in which explanatory variables are two actual disturbance values ​​and a response variable is the difference between the manipulated variables corresponding to the actual disturbance values. The manipulated variable selection unit 214 acquires the disturbance candidate value that is the shortest distance from the actual disturbance value, and selects the manipulated variable corresponding to the disturbance candidate value. If the distance between the actual disturbance value and the disturbance candidate value exceeds a predetermined value, the manipulated variable selector 214 selects the immediately previous manipulated variable.

[0055] <Edge controller: control unit: operation unit> The operation unit 215 performs an operation on the controlled object 520 with the operation amount selected by the operation amount selection unit 214 .

[0056] As described above, the edge controller 200 (control system 10) includes the operation unit 215 that operates the control target 520 with a selected operation amount.

[0057] <Operation amount calculation process> 11 is a flowchart of the operation amount calculation process according to this embodiment. The cloud controller 100 starts the operation amount calculation process when it receives the control amount transmitted by the edge controller 200. The operation amount calculation process is repeatedly executed when the control amount is received. It is assumed that the target value of the control amount has already been set at the start of the operation amount calculation process.

[0058] In step S11, the control amount receiving unit 112 receives the control amount transmitted by the edge controller 200 and outputs it to the feedback control unit . In step S12, the feedback control unit 116 calculates a feedback manipulated variable based on the target value and the controlled variable.

[0059] In step S13, if the disturbance candidate value generation model 131 is present (step S13→present), the operation amount calculation unit 118 proceeds to step S14. If the disturbance candidate value generation model 131 is not present (step S13→absent), the operation amount calculation unit 118 proceeds to step S15.

[0060] In step S14, the manipulated variable calculation unit 118 calculates disturbance candidate values ​​using the disturbance candidate value generation model 131. In step S15, the manipulated variable calculation unit 118 calculates a disturbance candidate value using a predetermined distribution.

[0061] In step S16, the feedforward control unit 117 calculates a feedforward manipulated variable for the disturbance candidate value calculated in steps S14 and S15. In step S17, the manipulated variable calculation unit 118 calculates a manipulated variable based on the feedback manipulated variable and the feedforward manipulated variable, and transmits the calculated manipulated variable to the edge controller 200 together with the disturbance candidate value.

[0062] <<Operation processing>> 12 is a flowchart of the operation processing according to this embodiment. The edge controller 200 starts the operation processing upon receiving the disturbance candidate value and the manipulated variable transmitted by the cloud controller 100. The operation processing is repeatedly executed upon receiving the disturbance candidate value and the manipulated variable.

[0063] In step S21, the manipulated variable selection unit 214 receives the disturbance candidate value and the manipulated variable transmitted by the cloud controller 100. In step S22, the manipulated variable selector 214 refers to the disturbance actual measurement value database 240 to acquire the latest disturbance actual measurement value.

[0064] In step S23, the operation amount selection unit 214 executes an operation amount selection process to acquire the operation amount. Details of the operation amount selection process will be described later with reference to FIG. In step S24, the operation unit 215 executes the operation of the controlled object 520 using the operation amount acquired in step S23.

[0065] <Operation amount selection process> Fig. 13 is a flowchart of the operation amount selection process according to this embodiment. Details of step S23 (see Fig. 12) will be described with reference to Fig. 13. Note that the operation process (see Fig. 12) is a process that is repeatedly executed, and the operation amount selection process is also repeatedly executed.

[0066] In step S31, the manipulated variable selection unit 214 starts the process of repeating steps S32 to S34 for each disturbance candidate value received from the cloud controller 100. In step S32, if there is a distance model 231 (step S32→YES), the operation amount selection unit 214 proceeds to step S33. If there is no distance model 231 (step S32→NO), the operation amount selection unit 214 proceeds to step S34.

[0067] In step S33, the manipulated variable selection unit 214 uses the distance model 231 to calculate the distance between the actual disturbance value acquired in step S22 and the disturbance candidate value. In step S34, the manipulated variable selection unit 214 calculates the distance between the actual disturbance value acquired in step S22 and the disturbance candidate value using a distance function.

[0068] In step S35, the operation amount selection unit 214 acquires the smallest distance (minimum distance) among the distances calculated in steps S33 and S34. In step S36, if the minimum distance is equal to or less than the predetermined value (step S36→YES), the operation amount selection unit 214 proceeds to step S37. If the minimum distance is greater than the predetermined value (step S36→NO), the operation amount selection unit 214 proceeds to step S38.

[0069] In step S37, the manipulated variable selection unit 214 selects the manipulated variable corresponding to the disturbance candidate value with the smallest distance. In step S38, the operation amount selection unit 214 selects the operation amount selected in the previous operation amount selection process. The operation amount selection unit 214 selects the previous operation amount by obtaining the latest operation amount with reference to the operation amount history database 250 (see FIG. 8). In step S39, the operation amount selection unit 214 stores the selected operation amount and the current time in the operation amount history database 250.

[0070] As described above, the control amount measurement unit (control amount measurement transmission unit 211), feedback control unit 116, disturbance candidate value generation unit 115, feedforward control unit 117, operation amount calculation unit 118, disturbance value measurement unit (disturbance value measurement transmission unit 212), operation amount selection unit 214, and operation unit 215 perform repetitive processing.

[0071] <Control system features> The cloud controller 100 generates one or more disturbance candidate values ​​and calculates a feedforward manipulated variable for each disturbance candidate value. The cloud controller 100 calculates the manipulated variable based on the feedback manipulated variable and the feedforward manipulated variable, and transmits the manipulated variable together with the disturbance candidate value to the edge controller 200. The edge controller 200 selects, from the disturbance candidate values, the manipulated variable corresponding to the disturbance candidate value that has the smallest distance from the actual disturbance value, and operates and controls the control target 520.

[0072] According to such a control system 10, feedback control and feedforward control, which have conventionally been performed at the edge 510 (on-site), are now performed by the cloud controller 100. This allows the control device installed at the edge 510 to be miniaturized, resulting in cost reduction. In particular, when there are multiple edges 510 and control targets 520, the cost reduction effect is significant.

[0073] The cloud controller 100 calculates the manipulated variable taking into account feedback control and feedforward control, and the edge controller 200 selects the manipulated variable appropriate for the current disturbance. This makes it possible to operate the control target 520 in a way that is most appropriate for the current disturbance (actual disturbance measurement value) without being affected by communication delays.

[0074] <<Variation: Processing Order>> In the above-described embodiment, the control amount measurement transmission unit 211 and the disturbance value measurement transmission unit 212 measure the control amount and the disturbance value at a predetermined timing, for example, at a predetermined cycle, and transmit them to the cloud controller 100. Instead of this, for example, the control amount measurement transmission unit 211 and the disturbance value measurement transmission unit 212 may measure and transmit them after the operation unit 215 executes the operation of the operation amount (see step S24 in FIG. 12 ).

[0075] <<Variation: Distance Model>> In the above-described embodiment, the learning data 275 (see FIG. 10) of the distance model 231 is generated based on the actual disturbance measurement value-compatible manipulated variable table 270 (see FIG. 9) configured from the actual disturbance measurement value and the manipulated variable that manipulated the control target 520. The edge controller 200 may generate the actual disturbance measurement value-compatible manipulated variable table 270 and the learning data 275 based on the candidate disturbance value and the manipulated variable, and generate the distance model 231 using this learning data 275.

[0076] In the above embodiment, the distance learning unit 213 of the edge controller 200 generates the distance model 231. The distance model 231 may be generated by the cloud controller 100 and transmitted to the edge controller 200.

[0077] Other variations Although several embodiments of the present invention have been described above, these embodiments are merely illustrative and do not limit the technical scope of the present invention. The present invention can take on various other embodiments, and various modifications such as omissions and substitutions can be made without departing from the spirit of the present invention. These embodiments and their modifications are included within the scope and spirit of the invention described in this specification, etc., and are included in the invention described in the claims and their equivalents.

[0078] <Hardware configuration> The cloud controller 100 and the edge controller 200 according to the above-described embodiments are each realized by a computer 900 having a configuration as shown in FIG. 14, for example. FIG. 14 is a hardware configuration diagram showing an example of a computer 900 that realizes the functions of the cloud controller 100 and the edge controller 200 according to the above-described embodiments. The computer 900 includes a CPU 901, a ROM 902, a RAM 903, an SSD 904, an input / output interface 905 (referred to as an input / output I / F (Interface) in FIG. 14), a communication interface 906 (referred to as a communication I / F in FIG. 14), and a media interface 907 (referred to as a media I / F in FIG. 14). The computer 900 may include a hard disk drive (HDD) instead of the SSD 904, or may include a HDD in addition to the SSD 904.

[0079] The CPU 901 operates based on a program stored in the ROM 902 or the SSD 904, and performs control by the control units 110, 210. The ROM 902 stores a boot program executed by the CPU 901 when the computer 900 is started up, programs related to the hardware of the computer 900, and the like.

[0080] The CPU 901 controls an input device 910 such as a mouse or keyboard, and an output device 911 such as a display or printer, via an input / output interface 905. The CPU 901 acquires data from the input device 910 via the input / output interface 905, and outputs generated data to the output device 911.

[0081] The SSD 904 stores programs executed by the CPU 901 and data used by the programs. The communication interface 906 receives data from other devices (not shown) (for example, the cloud controller 100, the edge controller 200, etc.) via a communication network and outputs the data to the CPU 901, and also transmits data generated by the CPU 901 to other devices via the communication network.

[0082] The media interface 907 reads a program or data stored in the recording medium 912 and outputs it to the CPU 901 via the RAM 903. The CPU 901 loads the program from the recording medium 912 onto the RAM 903 via the media interface 907 and executes the loaded program. The recording medium 912 is an optical recording medium such as a DVD (Digital Versatile Disk), a magneto-optical recording medium such as an MO (Magneto Optical disk), a magnetic recording medium, a conductive memory tape medium, a semiconductor memory, or the like.

[0083] For example, when the computer 900 functions as the cloud controller 100 and the edge controller 200 according to the above-described embodiment, the CPU 901 of the computer 900 executes the programs 138, 238 (see FIGS. 2 and 7) loaded onto the RAM 903, thereby realizing the functions of the cloud controller 100 and the edge controller 200. The CPU 901 reads and executes the programs 138, 238 from the recording medium 912. Alternatively, the CPU 901 may read the programs 138, 238 from another device via a communication network, or may install the programs 138, 238 from the recording medium 912 onto the SSD 904 and execute them. [Explanation of symbols]

[0084] 10. Control System 100 Cloud Controller 111 Goal Setting Department 112 Control amount receiving unit 113 Disturbance measurement value receiver 114 Disturbance candidate value learning unit 115 Disturbance candidate value generator 116 Feedback control section 117 Feedforward control section 118 Operation amount calculation section 131 Disturbance candidate value generation model 140 Disturbance measurement value database 200 Edge Controller 211 Control amount measurement transmission unit (control amount measurement unit) 212 disturbance value measurement transmitter (disturbance value measurement unit) 213 Distance Learning Department 214 Operation amount selection unit (operation amount receiving unit) 215 Operation section 231 Distance Model 240 Disturbance measurement value database 250 Operational volume history database 520 Control Target

Claims

1. a control amount measurement unit that measures a control amount of a control object; a feedback control unit that calculates a feedback manipulated variable for the controlled object based on the controlled variable and a target value of the controlled variable; a disturbance candidate value generating unit that generates one or more disturbance candidate values, which are values ​​of disturbances that affect a controlled variable of the controlled object; a feedforward control unit that calculates, for each of the disturbance candidate values, a feedforward manipulated variable that suppresses the influence of the disturbance indicated by the disturbance candidate value; a manipulation amount calculation unit that calculates a manipulation amount for each disturbance candidate value based on the feedback manipulation amount and the feedforward manipulation amount; a disturbance value measurement unit that measures the value of the disturbance and sets it as an actual disturbance value; a manipulated variable selection unit that selects a manipulated variable corresponding to the disturbance candidate value that has the smallest distance from the actual disturbance measurement value; an operation unit that operates the selected operation amount for the controlled object; Control system.

2. The disturbance candidate value generation unit A disturbance candidate value is generated using a disturbance candidate value generation model, which is a machine learning model generated using the actual disturbance measurement value as learning data. The control system of claim 1 .

3. The disturbance candidate value generation unit Generate a disturbance candidate value so as to follow the distribution of the actual disturbance value The control system of claim 1 .

4. The disturbance candidate value generation unit Generate candidate disturbance values ​​that follow a given distribution The control system of claim 1 .

5. The operation amount selection unit calculating a distance between the actual disturbance value and the candidate disturbance value using a distance model, which is a machine learning model generated using learning data in which explanatory variables are the two actual disturbance values ​​and a response variable is the difference between manipulated variables corresponding to the actual disturbance values; The disturbance candidate value having the smallest distance from the actual disturbance measurement value is acquired; Select the manipulated variable corresponding to the disturbance candidate value. The control system of claim 1 .

6. the controlled variable measurement unit, the feedback control unit, the disturbance candidate value generation unit, the feedforward control unit, the manipulated variable calculation unit, the disturbance value measurement unit, the manipulated variable selection unit, and the manipulation unit perform repeated processing; The operation amount selection unit If the distance between the actual disturbance value and the candidate disturbance value exceeds a predetermined value, the previous manipulated variable is selected. The control system of claim 1 .

7. a feedback control unit that calculates a feedback manipulated variable for the controlled object based on a controlled variable of the controlled object and a target value of the controlled variable; a disturbance candidate value generating unit that generates one or more disturbance candidate values, which are values ​​of disturbances that affect a controlled variable of the controlled object; a feedforward control unit that calculates, for each of the disturbance candidate values, a feedforward manipulated variable that suppresses the influence of the disturbance indicated by the disturbance candidate value; a manipulation amount calculation unit that calculates a manipulation amount for each disturbance candidate value based on the feedback manipulation amount and the feedforward manipulation amount. Cloud controller.

8. a manipulated variable receiving unit that receives a manipulated variable calculated for each disturbance candidate value based on a controlled variable of a controlled object, a target value of the controlled variable, and one or more disturbance candidate values ​​that are values ​​of disturbances that affect the controlled variable of the controlled object; a disturbance value measurement unit that measures the value of the disturbance and sets it as an actual disturbance value; a manipulated variable selection unit that selects a manipulated variable corresponding to the disturbance candidate value that has the smallest distance from the actual disturbance measurement value; an operation unit that operates the selected operation amount for the controlled object; Edge controller.

Images