法国专利FR3063369A1 ISOLATION MANAGEMENT DEVICE AND INSULATION MANAGEMENT METHOD

专利PDF首页>>法国专利

专利附录

专利说明

权利要求

类似技术

同族专利

引用文献

法律状态

优先权

专利摘要:
An isolation management device (1) comprising: a database (2) configured to store information relating to a plant constructed with a plurality of components (26), the information including a relationship between the a plurality of components (26); a receiver configured to receive a designation of the target area information defining a target area on the plant; an analyzer configured to analyze a plurality of respective state profiles of the plurality of components (26) in connection with a change of state of at least one of the plurality of components (26) in the target area on the basis of the information stored in the database; deep learning circuitry (9) configured to extract at least one specific profile from the plurality of profiles analyzed by the analyzer as an extraction profile; a plan generator (17) configured to produce a work plane based on the extraction profile; and an output interface (18) configured to output the work plane produced by the plan generator.
公开号:FR3063369A1
申请号:FR1851684
申请日:2018-02-27
公开日:2018-08-31
发明作者:Kei TAKAKURA；Susumu Naito；Hidehiko Kuroda；Hiroki Shiba
申请人:Toshiba Corp；Toshiba Energy Systems and Solutions Corp；
IPC主号:

专利说明:

Holder (s): KABUSHIKI KAISHA TOSHIBA, TOSHIBA ENERGY SYSTEMS & SOLUTIONS CORPORATION.
Extension request (s)
Agent (s): CABINET FEDIT LORIOT.
p4) ISOLATION MANAGEMENT DEVICE AND INSULATION MANAGEMENT METHOD.
FR 3,063,369 - A1 (57) an isolation management device (1) comprising: a database (2) configured so as to store information which relates to a factory built with a plurality of components (26 ), the information comprising a relationship between the plurality of components (26); a receiver configured to receive a designation of the target area information defining a target area on the factory; an analyzer configured to analyze a plurality of respective state profiles of the plurality of components (26) in connection with a change of state of at least one of the plurality of components (26) in the target area , based on the information stored in the database; deep learning circuits (9) configured to extract at least one specific profile from the plurality of profiles analyzed by the analyzer as an extraction profile; a plan generator (17) configured to produce a work plan based on the extraction profile; and an output interface (18) configured to deliver the work plan produced by the plan generator.

ISOLATION MANAGEMENT DEVICE AND INSULATION MANAGEMENT METHOD
In the field of the invention, embodiments described above relate, in general, to an isolation management technology intended to ensure the management of the isolation work in order to temporarily isolate a target device. in a factory during an event in the factory, such as construction, maintenance control and / or repair.
In the background, conventionally, before isolation work in a factory such as a power plant, a specialized engineer refers to a developed connection diagram representative of a connection relation of respective components and creates a work plan while considering the influence of isolation work on other components. In order to reduce the task involved in such isolation work, a process for automating the work schedule to ensure the inspection of each bus in the factory has been proposed. In addition, a method of extracting the target drawing from design documents has been proposed. In addition, a method of preventing erroneous operation at the time of performing the isolation work has been proposed.
[Patent document 1] Publication of Japanese patent application not examined ΝΉ6-46528 [Patent document 2] Publication of Japanese patent application unexamined No. 2011-96029 [Patent document 3] Publication of Japanese patent application no
-2 examined N ° 2008-181283
In a factory, a large number of components, such as different types of devices, are installed as a whole. Thus, in the case of the creation of an isolation work plan taking all the components into consideration, a large number of calculations is required. For example, when there are 100 devices in the target area and each of these 100 devices has two ACTIVE / INACTIVE states, there are 2 to the power of 100 status profiles (ΙχΙΟ ^ θ or more). For this reason, it is not efficient to calculate and obtain the set of state profiles and there is a problem that it is not possible to create a work plan effectively.
In view of the problem described above, embodiments of the present invention have the objective of creating an isolation management technology which makes it possible to efficiently produce a work plan which is most suitable for the work of isolation.
In order to solve the problems mentioned above, an isolation management device is created comprising: a database configured so as to store information which relates to a factory built with a plurality of components, the information comprising a relationship between the plurality of components; a receiver configured to receive target area information defining a target area on the factory; an analyzer configured to analyze a plurality of respective state profiles of the plurality of components in connection with a state change of at least one of the plurality of components in the target area, based on the information stored in the database; deep learning circuits configured to extract at least one specific profile from the plurality of profiles analyzed by the analyzer as an extraction profile; a plan generator configured to produce a work plan based on the extraction profile; and an output interface configured to deliver the work plan produced by the plan generator.
The isolation management device may also include a verifier configured to verify the profile of respective states on the
-3 components outside the target area in connection with the change of state of each component in the target area according to the work plan.
Preferably, the deep learning circuits comprise an intermediate layer comprising a multilayer neural network and are configured so as to acquire a quantity characteristic of each of the plurality of profiles; and the deep learning circuits are further configured to extract the extraction profile according to the characteristic magnitude of each of the plurality of profiles.
Preferably, the deep learning circuits include a training data generator configured to produce training data configured to build the multilayer neural network.
Preferably, the database is configured so as to store information on at least one previous work plan; and the training data generator is configured to produce the training data on the basis of the previous work plan stored in the database.
Preferably, the plurality of components comprises a first predetermined type of component and a second type of component linked to the first type of component;
the training data generator is configured so as to produce first matrix data, in which a state of the first type of component analyzed by the analyzer is treated as an input quantity, and second matrix data, in which a state of the second type of component analyzed by the analyzer is treated as an output quantity; and the deep learning circuits are configured to cause learning by the multilayer neural network of the training data which includes the first matrix data and the second matrix data.
Preferably, the deep learning circuits are configured so as to define a reward in relation to the information stored in
-4the database, extract a plurality of specific profiles from the plurality of profiles analyzed by the analyzer, in the form of a plurality of extraction profiles, and extract a profile having a value of the most reward high among the plurality of extraction profiles.
Preferably, the deep learning circuits are configured so as to extract an operational procedure from the isolation work based on the extraction profile; and the plan generator is configured to generate the work plan based on the operational procedure extracted by the deep learning circuits.
Preferably, the analyzer is configured to perform at least one of an analog circuit analysis, a logic circuit analysis and a path search analysis.
The present invention further relates to an isolation management method comprising: memorizing information, which relates to a factory built with a plurality of components and defines the relationship between the plurality of components in a database of data ; receiving target area information defining a target area on the factory; analysis of a plurality of respective state profiles of the plurality of components in connection with a change of state of at least one of the plurality of components in the target area, based on the information stored in the database of data ; extracting a specific profile from the plurality of profiles analyzed by the analyzer, as an extraction profile; generation of a work plan on the basis of the extraction profile; and producing the work plan.
A brief description of the accompanying drawings is given below, among which:
FIG. 1 is a functional diagram representing a device for managing the insulation of a first embodiment;
Figure 2 is a simplified diagram showing a multilayer neural network;
FIG. 3 is a configuration diagram showing a state of an energy distribution device before the isolation work;
FIG. 4 is a configuration diagram representing a state of an energy distribution device during the isolation work;
Figure 5 is an algorithm showing the first part of isolation management processing;
FIG. 6 is an algorithm representing the second part of the processing for managing the isolation consecutive to that of FIG. 5;
FIG. 7 is an algorithm representing the third part of the isolation management processing consecutive to those of FIGS. 5 or 6;
FIG. 8 is an algorithm representing the final part of the processing for managing the isolation subsequent to that of FIG. 7.
A detailed description will now be made below.
In an embodiment of the present invention, an isolation management device comprises:
a database configured to store information relating to a factory built with a plurality of components, the information comprising a relationship between the plurality of components;
a receiver configured to receive target area information defining a target area on the factory;
an analyzer configured to analyze a plurality of respective state profiles of the plurality of components in connection with a change of state of at least one of the plurality of components in the target area, based on the information stored in the database;
deep learning circuits configured to extract at least one specific profile from the plurality of profiles analyzed by the analyzer as an extraction profile;
a plan generator configured to produce a work plan based on the extraction profile; and an output interface configured to deliver the work plan produced by the plan generator.
In another embodiment of the present invention, a method of managing the isolation comprises:
-6 memorizing information, which relates to a factory built with a plurality of components and defines the relationship between the plurality of components in a database;
receiving target area information defining a target area on the factory;
analysis of a plurality of respective state profiles of the plurality of components in connection with a change of state of at least one of the plurality of components in the target area, based on the information stored in the database of data ;
extracting a specific profile from the plurality of profiles analyzed by the analyzer, as an extraction profile;
generation of a work plan on the basis of the extraction profile; and producing the work plan.
According to embodiments of the present invention there is created an isolation management technology which allows to efficiently produce a work plan which is most suitable for isolation work.
Below, embodiments will be described with reference to the accompanying drawings. First, a factory such as a power plant is configured from several components such as an energy distribution device, a control device and a control device. When an event such as a construction, maintenance check or repair of a specific device or assembly is carried out in such a factory, it is necessary to minimize the influence of the event on the safety of workers and other devices or assemblies. Thus, the target device or target assembly in the event is electrically isolated from the other devices or assemblies and stopped (powered down). Such work is called isolation.
In the case of the creation of an isolation work plan in conventional technology, a specialized engineer refers to design documents which include a simple wired connection diagram representative of a connection relationship of respective components, a diagram ECWD (elementary control wiring diagram, i.e., a type of circuit diagram developed)
-7 representative of a control relationship of respective components, an IBD diagram (nested functional diagram), and a programmed logic diagram. In view of these documents, the specialized engineer creates an isolation work plan while considering the influence of the isolation work. For example, when an engineer formulates an isolation plan for a nuclear power plant, it is necessary to study thousands to tens of thousands of associated documents. In addition, an engineer must have extensive expertise and experience and significant work is spent. In addition, an alarm informing that an anomaly occurs due to a plan error which can be attributed to insufficient examination or supervision by an engineer. For the same reason, there is also an event for which the operation of the factory stops.
In addition, there is a predetermined procedure for actual isolation work. When the isolation work does not take place exactly according to this procedure (sequence), an alarm is issued or a lockout is activated in order to trigger an event that affects the factory. Thus, for each device that requires an operation for isolation work, it is necessary for a specialized engineer to evaluate such a device for each procedure by referring to design documents and the state of the factory. This requires significant work. Although there is a method for simulating and evaluating such procedures evaluated manually for each procedure, this simulation method involves a significant computation cost.
In addition, in the case of isolation work planning, it is, for example, conceivable that a rule is provided beforehand for a jumper terminal or circuit breaker in order to greatly reduce the number of profiles simulation. However, when an isolation profile is extracted by a simulator, it is not obvious whether the extracted isolation profile is the optimum plane. The definition of the optimum term described above depends on the management principles of the administrator. For example, an isolation plan that minimizes the dose of worker exposure is assumed to be a main idea of the optimum isolation plan. Similarly, an isolation plan which
-8minimizing the number of stages of work (duration of operation) is supposed to be a main idea of the optimum isolation plan.
The reference numeral 1 in FIG. 1 is an isolation management device 1 which manages an isolation work plan and automatically produces a work plan. The isolation management device 1 is equipped with an integrated database 2 which stores (a) factory design documents, (b) operation information (i.e. processing data), (c) personnel planning information, (d) environment information, (e) construction information, (f) incident information and (g) a work plan isolation created in the past. Factory design documents include, for example, a factory construction diagram, layout diagram, P&ID diagram, ECWD diagram, IBD diagram, simple link diagram, and programmed logic diagram. The operation information is, for example, information on an operational state of a plant control, monitoring and instrumentation equipment. The personnel planning information includes, for example, a construction and plant progress plan. The environmental information includes, for example, a dose of radiation, a temperature and a humidity at each work site in the factory. Construction information is information about feasibility such as obstacles at the work site, objects interfering at the work site, and work at a location with a high altitude. Incident information is information about previous incident events, each of which has its associated information, such as date, duration, location, device name, set name, and build.
The different types of information element described above are associated with each other on the integrated database 2. In other words, data representative of different types of information element are structured. In addition, the integrated database 2 can be built on a data server arranged in the factory or can be built on a server arranged on an installation external to the factory. In addition or alternatively, the base of
Integrated data 2 can be built on a server distributed over a network. In addition, these different types of information elements are entered beforehand on the integrated database 2.
The isolation management device 1 includes a factory simulator 3 which simulates a change of influence on other devices or other assemblies in the case of the isolation of a predetermined device or a predetermined set. The factory simulator 3 comprises an analysis section (that is to say, an analyzer or any other type of circuit) 4, a verification section (that is to say, a verifier or any other types of circuits) 5, and a data storage section (i.e., database, buffer, memory or all other types of circuits) 81 which stores different data. Analysis section 4 is used to simulate the plant in the case of production of an isolation work plan. Verification section 5 is used to simulate different changes occurring in the factory when the isolation work is performed according to the isolation work plan produced.
In addition, the analysis section 4 comprises analog circuit analysis circuits 6 configured so as to analyze an analog circuit, logic circuit analysis circuits 7 configured so as to analyze a logic circuit and the circuits of path search analysis 8 configured to perform path search analysis, for example, based on graph theory. It is also possible to install an arbitrary (logical) analysis process in the analysis section 4 in addition to the three analysis circuits 6, 7 and 8 described above. When changing a state of a device or assembly associated with a target area (i.e., a target site or a target part) of the isolation work, the analysis section 4 analysis of the respective state change profiles occurring on other devices or assemblies on the basis of the information stored in the integrated database 2. The verification section 5 also has the same configuration as the analysis section 4, and checks the work plan produced on the basis of the information stored in the integrated database 2.
-10The isolation management device 1 comprises deep learning circuits (for example, a deep learning unit or a deep learning model) 9 which performs a processing associated with the production of a work plan isolation based on the data stored in the integrated database 2 and the analysis result of the factory simulator 3. The deep learning circuits 9 include a multilayer neural network 10. The factory simulator 3 is a computer that simulates the behavior of the factory. The deep learning circuit 9 is a computer equipped with artificial intelligence which implements machine learning.
The deep learning circuits 9 include a training data generation section (i.e., circuits) 11 configured to produce learning data which is necessary in order to build the multilayer neural network 10 which has completed learning. The training data generation section 11 includes circuits for generating second matrix data 12 and circuits for generating second matrix data 13. The circuits for generating first matrix data 12 produce first matrix data in which the state of the first type I device (component) analyzed by the analysis section 4 is treated as its input quantity X. The circuits for generating second matrix data 13 produce the second matrix data in which the state of the second type of device (component) analyzed by the analysis section 4 is treated as its output quantity Y.
The deep learning circuits 9 further include a reward definition section (i.e., circuits) 14 configured to define respective rewards for different types of information items stored in the integrated database 2, a reinforcement learning section (i.e., circuits) 15 configured to extract the model maximizing the value of the isolation plan based on the rewards, and a operational procedure extraction (i.e., circuits) section 16 configured to extract the operational procedure (execution order) from the isolation job.
The factory simulator 3 and the in-depth learning circuits 9 can be mounted on individual devices or installed in a computer or a server on an installation associated with fusine. In addition, or as a variant, the factory simulator 3 and the in-depth learning circuits 9 can be installed on a server distributed outside the installation associated with the factory.
The isolation management device 1 comprises a plan generator 17 configured so as to produce a work plan on the basis of a predetermined model extracted by the in-depth learning circuits 9 and further comprises an interface d user 18 used by an administrator of the isolation management device L
The user interface 18 is constituted, for example, by a personal computer or a terminal on a tablet in an installation associated with a factory. In addition, the user interface 18 includes a receiving section (i.e., a receiver or an input interface) 19 and an output section (i.e., an interface output) 20. The receiving section 19 receives the designation of a location (or area) in which a target device (component) to be subjected to isolation work in a factory exists, as target area information . The output section 20 delivers the work plan produced. To harm this, the reception section 19 includes input devices such as a keyboard and a mouse with which the administrator performs an input job. In addition, the output section 20 includes components for forming a destination of a work plan such as a display device, a printing device and a data storage device.
In addition, the isolation management device 1 comprises a main control unit 100 which unitarily controls the integrated database 2, the factory simulator 3, the deep learning circuits 9, the plan generator 17, and the user interface 18. In addition, the deep learning circuits 9 include a data storage section (that is to say, database, buffer, memory or all other types of circuits ) 82 which stores different data.
FIG. 2 represents a first case of the multilayer neural network
-12 10. In this multilayer neural network 10, units are arranged in multiple layers and are connected to each other. Each unit receives multiple inputs U and calculates an output Z. The output Z of each unit is expressed in the form of an output of an activation function F of the total input U. The activation function F presents a weight and an offset. The neural network 10 comprises an input layer 21, an output layer 22 and at least one intermediate layer 23.
In the present embodiment, the neural network 10 comprising the intermediate layer 23 comprising six layers 24 is used. Each layer 24 of the intermediate layer 23 is composed of 300 units. By causing the multilayer neural network 10 to learn the training data beforehand, it is possible to automatically extract a characteristic quantity on the profile of a change in state of the circuit or of the assembly. The multilayer neural network 10 can define an arbitrary number of intermediate layers, an arbitrary number of units, an arbitrary learning rate, an arbitrary learning number and an arbitrary activation function on the user interface 18 .
The neural network 10 is a mathematical model which expresses characteristics of a brain function by computer simulation. For example, an artificial neuron (node) that has formed a network by synaptic link changes the synaptic coupling force by learning and presents (that is, constitutes) a model that has acquired problem-solving ability. It may be noted that the neural network 10 of the present embodiment acquires the capacity for problem solving by deep learning.
Next, a description of the processes for producing an isolation work plan according to this embodiment will be given. In the present embodiment, a description will be given of the work of remodeling the energy distribution assembly 25 which constitutes a part of the energy supply device in the factory.
FIG. 3 is a diagram representing the state of the assembly of
-13 energy distribution 25 before isolation work. FIG. 4 is a configuration diagram showing the state of the power distribution assembly 25 during the isolation work. For reasons of ease of understanding, the circuits of the power distribution assembly 25 are simplified in FIGS. 3 and 4.
As shown in FIGS. 3 and 4, the power distribution assembly 25 includes several circuit breakers 26 to 34, several disconnectors 35 to 45, several transformers 46 to 52, and several energy distribution panels 53 to 60. The power distribution assembly 25 is constructed using these components. Circuit breakers 26 to 34 and disconnectors 35 to 45 constitute the first type of components and power distribution panels 53 to 60 connected to the first type of components constitute the second type of components. In addition, several buses 61 to 63 are present and the electrical energy is supplied to the respective devices of the factory from these buses 61 to 63 via the energy distribution panels 53 to 60.
The upper side of the sheet of each of Figures 3 and 4 shows components which are on the upstream side and close to the power source. The bottom side of the sheet in each of Figures 3 and 4 shows components that are on the downstream side and away from the power source. In the present embodiment, a case of isolation of the energy distribution panel 53 from the energy distribution assembly 25 is shown in order to repair one of the predetermined energy distribution panels 53. Among all the circuit breakers 26 to 34 and the disconnectors 35 to 45 in FIGS. 3 and 4, those marked by X are open (that is to say, in an isolated state or an out of service state) and the rest (i.e., those not marked with X) are closed (i.e., in a conductive state or an active state).
In the present embodiment, the energy distribution panels 53 to 55 are connected respectively to the three buses 61 to 63. The energy distribution panels 53 to 55 are connected to the buses 61 to 63 via the circuit breakers 26 to 28 and transformers 46 and 47. Electrical energy is supplied to power distribution panels 56 to 60 on the side
-14 plus downstream via the energy distribution panels 53 to 55. The energy distribution panels 53 to 55 on the upstream side are connected to the energy distribution panels 56 to 60 on the downstream side by via circuit breakers 29 to 34, disconnectors 35 to 39 and transformers 48, 49, 51, and 52. In addition, the power distribution panels 56 to 60 on the downstream side are connected one to the other via the disconnectors 40 to 44.
Each of the circuit breakers 26 to 34 and the disconnectors 35 to 45 has two states: Active and Inactive. In addition, each of the power distribution panels 53 to 60 has two states: operation and shutdown. In the present embodiment, there are several state profiles when the state of each of these components is changed. Among these state profiles, the state profile indicative of the optimum state for isolation is specified. In the following description, the first of the power distribution panels 53 to be insulated is appropriately designated as the power distribution panel 53 of the target area T in the present embodiment.
As shown in FIG. 3, before the isolation work, the electrical energy is delivered from the predetermined bus 61 to the energy distribution panel 53 of the target area T. In addition, the electrical energy is delivered to the energy distribution panels 56 and 57 on the downstream side via the energy distribution panel 53. As for other energy distribution panels, the energy distribution panel 54 is stopped and the circuit breakers 27, 33 and the disconnector 38 which are connected to this energy distribution panel 54 are open. Another energy distribution panel 55 is in operation, but the circuit breaker 34 and the disconnector 39 on the downstream side of this energy distribution panel 55 are open. In other words, the electrical energy is delivered to the five energy distribution panels 56 to 60 on the downstream side via the energy distribution panel 53 on the target area T.
For example, in the case of isolating the power distribution panel 53 from the target area T, all the circuit breakers 26 and 29 to 32 connected directly to the power distribution panel 53 are open (the circuit breaker
Circuit 15 is shown in the open state in FIG. 3) and the disconnectors 35 and 36 on the downstream side of open circuit breakers 29 to 32 are open. In this case, the supply of electrical energy from the bus 61 is stopped for the energy distribution panel 53 of the target area T and all the energy distribution panels 56 to 60 on the downstream side. In other words, when the respective states of the circuit breakers 26, 29 to 32 and the disconnectors 35 and 36 are changed relative to the target area T, the states of the power distribution panels 56 to 60 respectively at the level other locations change.
Here, it is assumed that there is an operational rule such that the particular power distribution panel 56 on the downstream side retains the energized state. Based on this operational rule, when isolation of the power distribution panel 53 from the target location T is implemented, the particular power distribution panel 56 is placed in a power failure state and thus an anomaly alarm is issued. As previously described, it is required to specify the status profile of the electrical power supply to the particular power distribution panel 56 through another power supply path of a in such a way that the profile of the change of state on each component does not become a profile in which a fault alarm is issued.
For example, an electrical power supply path from bus 63 is secured as an alternate power supply path as shown in Figure 4. Electrical energy is supplied to the distribution panel. energy 60 on the downstream side by closing the circuit breaker 34 and the disconnector 39 which are connected to the energy distribution panel 55 corresponding to this bus 63. In this way, the electrical energy is supplied to the distribution panel particular energy 56 from the power distribution panel 60. The state shown in Figure 4 is the specific profile representative of the optimum state in which isolation is completed.
Incidentally, the isolation work involves an operational procedure (sequence) of predetermined devices. For example, when there is a particular energy distribution panel 56, the isolation work is implemented
-16 after fixing another energy supply path for this energy distribution panel 56. In addition, after closing the predetermined circuit breaker 34 and disconnector 39, the other circuit breakers 26 to 32 and disconnector 35 and 36 are open. In addition, when the circuit breakers 30 and 31 and the disconnectors 35 and 36 are connected to each other, the circuit breakers 30 and 31 are open, and after that, the respective disconnectors 35 and 36 corresponding to the circuit breakers circuit 30 and 31 are open.
In the present embodiment, the profile of the change of state in each optimum component for the isolation is extracted automatically using the factory simulator 3 and the deep learning circuits 9. First, it will be given a description of a case in which there is no model of the multilayer neural network 10 which has completed the learning necessary for in-depth learning.
As shown in FIG. 1, during the production of a work plan, the isolation management device 1 receives first target zone information defining the target isolation zone T. After that, an administrator performs an input operation to specify the power distribution panel 53 of the target area T using the user interface 18. Upon receipt of this input operation, the device management system 1 acquires data such as design documents associated with the device (s) and assembly, to which the power distribution board 53 of the target area T is connected, from the base of integrated data 2.
In addition, the isolation management device 1 builds lists of the link information, device information and attribute information contained in the design documents, and incorporates the lists into the analysis section 4 of the simulator. 3. In addition, the isolation management device 1 incorporates the process information and the status information of the devices stored in the integrated database 2 (for example, information indicating whether the circuit breakers 26 to 34 respectively are open or closed) of the analysis section 4.
Here, the analysis section 4 performs a simulation based on the lists of device information, attribute information, link information and status information using the analog circuit analysis circuits 6, the logic circuit analysis circuits 7 and / or the path search analysis circuits 8. It can be noted that two or more of these analysis functions 6, 7, 8 can be associated as a function of the circuit target or target set. For example, it is possible to associate the logic circuit analysis circuits 7 and the path search analysis function 8 in the case of a targeting simulation which is composed of an IBD scheme and a device diagram based on a simple link diagram. In this way, it is possible to simulate the behavior of each component of the plant and the influence on each component of the plant when performing isolation work.
In addition, the analysis section 4 delivers the state of each component (device), for example, the conduction state of the power distribution panel 53 of the target area T in the case of the distinct change of the respective states. of all the circuit breakers 26 to 34 and all of the disconnectors 35 to 45. There are many change profiles on the respective states of these components. These change profiles are transmitted to the training data generation section 11 of the deep learning circuits 9.
In addition, the training data generation section 11 processes the attributes or states of the circuit breakers 26 to 34 and the disconnectors 35 to 45 (the first type of components) as input variable X and constructs lists. attributes or states of the power distribution panels 53 to 60 (the second type of components) as an output quantity Y. It can be noted that the attributes or states of the first type of components and of the second type of components are delivered from analysis section 4.
The first matrix data generation function 12 of the training data generation section 11 expresses the state (that is, open state or blocked state) of each of the circuit breakers 26 to 34 and disconnectors 35 to 45 in the form of 0 or 1, and thus produces the first matrix data of the input variable X which are state data
-18respectives of these components 26 to 34 and 35 to 45.
The second matrix data generation function 13 of the training data generation section 11 assigns 0 or 1 to the state (i.e., conductive state or non-conductive state) of each of the panels power distribution 53 to 60 when each of the circuit breakers 26 to 34 and disconnector 35 to 45 is in a predetermined state. In other words, the function for generating second matrix data 13 expresses the state of each of the power distribution panels 53 to 60 in the form of 0 or 1, and thus produces the second matrix data of the output quantity Y which are data of the respective states of these components 26 to 34 and 35 to 45 in terms of conduction. In the present embodiment, discrete values 0 and 1 are supplied as an output quantity. However, by appropriately defining functions and parameters such as the activation function in the output layer, it is possible to classify them into multiple classes other than 0 and 1 and it is also possible to output continuous values.
The isolation management device 1 causes the multilayer neural network 10 to learn this listed matrix data as training data. The deep learning circuits 9 build the neural network 10 which has completed the learning, in such a way that the correct response rate of the output result becomes high. For example, the deep learning circuits 9 build the neural network 10 which has completed learning, in such a way that the uncertainty between the output result and the response (expected output) in the case of l verification data input becomes weak.
Next, a description of a procedure for producing an isolation work plan using the multilayer neural network 10 which has completed learning will be given. First, the designation of the power distribution panel 53 of the target area T is received as target area information using the user interface 18. In the present embodiment, an instruction to deactivate the power distribution panel 53 of the installation location T is entered as information of
-19 target area.
In addition, the status information of the power distribution panel 53 of the target area T and the status information of the circuit breakers 26 to 34 and the disconnectors 35 to 45 are delivered from the integrated database. 2 to the deep learning circuits 9. The circuit breakers 26 to 34 and the disconnectors 35 to 45 are connected as devices to the power distribution panel 53 and are components of this assembly. The deep learning circuits 9 use the neural network 10, which has been constructed on the basis of the input quantity X and has completed learning, so as to extract such a profile of association of the states of the circuit breakers of circuit 26 to 34 and disconnectors 35 to 45 in which the power distribution panel 53 of the target area T is deactivated.
In the present embodiment, active / inactive association profiles of circuit breakers 26 to 34 and disconnectors 35 to 45 relating to the power distribution panel 53 of the target area T are entered as quantity d input X on the neural network 10 which has completed learning. The deep learning circuits 9 extract such a profile of ACTIVE / INACTIVE associations from circuit breakers 26 to 3.4 and disconnectors 35 to 45 in which the power distribution panel 53 of the target location T is deactivated, from all states of power distribution panels 53 to 60.
When there is no operational procedure (i.e., when the operator on site can start from any operation) regarding the actual operation of circuit breakers 26 to 34 and disconnectors 35 to 45, it is possible to produce the isolation work plan on the basis of the profile extracted from the ACTIVE / INACTIVE association.
On the contrary, when there is a specific operational procedure (that is, when the operator on the site must start from a specific operation), the in-depth learning circuits 9 introduce the ACTIVE association profile / INACTIVE extract (that is to say, a specific profile) and the rules and logic of the operational procedure in the extraction section of
-20operative procedure 16. The operational procedure extraction section 16 extracts the ACTIVE / INACTIVE operational procedure from circuit breakers 26 to 34 and disconnectors 35 to 45 which corresponds to the rules and logic and delivers the extracted operational procedure. The rules and logic of the operational procedure can be entered on the user interface 18 or be stored beforehand in the integrated database 2.
The operational procedure extraction section 16 introduces respective ACTIVE / INACTIVE association profiles of circuit breakers 26 to 34 and disconnectors 35 to 45, which can be removed during the processing of the isolation work, as input quantity X in the neural network 10 which has completed learning. The operational procedure extraction section 16 delivers respective state profiles of the power distribution panels 53 to 60 as an output quantity Y. In this processing, the operational procedure extraction section 16 limits the quantity input X and the output quantity Y based on the rules or logic of the operational procedure entered, and then finally extracts (list) the operational procedure in which the energy distribution panel 53 of the target area T is brought into the target state.
In addition, it is assumed that several proposed plans (choice) exist in the extracted profiles (list) and the operational procedure. Thus, using arbitrary information such as environment information in the factory, the optimum plan proposed is extracted from the different plans proposed using the reinforcement learning section 15. The reinforcement learning section 15 uses reinforcement learning which is a type of machine learning. In reinforcement learning, an agent, which is a consistent body of learning, such as a software agent, learns to maximize value in a given environment.
When a state S _t at time t of the environment is given, the agent perceives such a state S _t of the environment and selects an action (or a set of actions) A | at time t. With such an action A ^, the agent obtains a numerical reward q + j and the state of the environment transits from state to state S ^ + p With
-21 reinforcement learning, the agent selects a set of actions in order to maximize a value of the total reward obtained (or whose achievement is hoped for) during such a set of actions. Such a total reward obtained (or whose achievement is hoped for) in the course of a set of actions is called a value and such a value is formulated as a value function Q (s, a), in which s represents a state of the environment and a represents an action possibly taken or selected. In the present embodiment, deep reinforcement learning which expresses the value function by the multilayer neural network 10 is used.
The extracted profile and the extracted operational procedure are entered in the reinforcement learning section 15. In addition, the arbitrary information comprising the environment information stored in the integrated database 2 is entered in the reinforcement learning section. 15. For example, radiation dose, temperature, humidity, position (coordinates) information for each zone in the power plant and / or the operator's travel distance is entered. In addition, these pieces of information are defined by rewards. For example, when the environment of the area in which the energy distribution panel 53 of the target area T is arranged is indicated with a dose of radiation 1 pSv / h, a temperature of 25 ° C, a humidity of 30 % and a movement distance of 10 m, the rewards corresponding to these four parameter values are defined respectively as -1 point, -1 point, -6 points and -6 points.
To define these rewards, an arbitrary function or conversion formula defined by the administrator can be used. For example, the environmental information is defined as a reward for each zone in which each component is arranged, such as the zone in which the circuit breakers 30 and 31 are arranged and the zone in which the disconnectors 35 and 36 are arranged .
The input variable X is defined as the transition from the work area associated with the ACTIVE / INACTIVE operation of the circuit breakers 26 to 34 and the disconnectors 35 to 45, which transition is at least one of the elements
-22information associated with the reward s, the entry profile, and the operational procedure. A value function is expressed using the multilayer neural network 10. By using such a value function, the plane which has the highest value among the different proposed planes is determined.
Based on the determined proposed plan, the plan generator 17 produces a work plan. This work plan can be a document made up of sentences and figures that can be identified by an operator or work support data. The work plan produced by the plan generator 17 entered the verification section 5 of the factory simulator 3 before being possibly delivered.
The verification section 5 verifies the influence on the factory in the case of the execution of the isolation work according to the work plan. For example, in the principle of evaluation based on the simulator, the verification is carried out on the basis of physical models such as circuit diagrams and overall diagrams. In addition, it is checked whether a problem, such as a fault alarm and an error on the insulation work occurs or not, in the case of the execution of the isolation work according to the work plan . In this way, it is possible to check whether the work plan on the basis of the specific profile extracted by the deep learning circuits 9 is appropriate or not, before actually carrying out the isolation work. When there is no problem on the work plan as a result of this verification, this work plan is delivered by the output section 20 of the user interface 18.
In the present embodiment, as described above, it is possible to automatically produce an isolation work plan by associating the factory simulator 3 and the deep learning circuits 9 which comprise the network of multilayer neurons 10. In addition, by comparison with the case in which an isolation work plan is produced by the simulator alone, the computation cost can be eliminated. Furthermore, by using the reinforcement learning section 15, it is possible to automatically create the isolation work plan by which the isolation work can be carried out more effectively.
In the present embodiment, a characteristic quantity of change profiles is acquired by the multilayer neural network 10 and a specific profile is extracted on the basis of the characteristic quantity. Thus, the processing efficiency in order to extract a specific profile from several change profiles can be improved.
In addition, it is possible to shorten an extraction time for a specific profile from several change profiles by causing the multilayer neural network 10, which has completed learning to extract the specific profile.
Furthermore, the training data generation section 11 can produce a work plan which follows the isolation work carried out in the past, by producing learning data on the basis of the previous work plans stored in the database. Integrated Data 2. As a result, the reliability of the work plan can be improved.
Furthermore, the deep learning circuits 9 can produce training data which correspond to the types of respective components constituting the factory, by causing the multilayer neural network 10 to learn the training data which comprise the first training data. matrix and the second matrix data. Thus, it is possible to build the multilayer neural network 10 suitable for isolation work in the factory.
The reinforcement learning section 15 can extract the most suitable profile for the isolation work by extracting the proposed plan with the highest value on the basis of the reward from several respective proposed plans which are produced from of several specific profiles. Incidentally, the reinforcement learning section 15 includes an in-depth reinforcement learning function 15A as an option of the reinforcement learning, and this in-depth reinforcement learning function 15A uses a neural network.
In addition, the operational procedure extraction section 16 can extract the most suitable operational procedure for the isolation work, extracting the operational procedure from the isolation work based on the profiles.
-24 specific extracts.
The isolation management device 1 of the present embodiment comprises hardware resources such as a CPU unit (central processing unit), a ROM memory (read only memory), a RAM memory (random access memory) and an HDD disk. (hard disks), and is configured as a computer in which information processing by software is ensured with the use of hardware resources by causing the CPU to execute different programs. In addition, the isolation management method of the present embodiment is obtained by causing the computer to execute the various programs.
Next, a description of the processing executed by the isolation management device 1 will be given with reference to the algorithms of FIGS. 5 to 8.
As shown in FIG. 5, in step S11 corresponding to the path RI in FIG. 1, the integrated database 2 first stores various information comprising plant design documents, ordering information, personnel planning information, environment information, construction information, incident information and previous work plans.
In the following step S12 corresponding to the paths R2 and R3 in FIG. 1, the reception section 19 of the user interface 18 receives target area information defining the target area T of the isolation work based on the operation entered by the administrator. For example, the designation of the power distribution panel 53 of the target area T is received as target area information.
In the following step SI3 corresponding to the paths R6 and R11 in FIG. 1, the main control unit 100 of the insulation management device 1 causes the data storage section 81 of the simulator to be acquired. factory 3 and the data storage section 82 of the deep learning circuits 9, of information on the energy distribution panel 53 of the target area T from the integrated database 2. More specifically, the data storage sections 81 and 82 acquire information relating to the power distribution panel
-2553 (component) of the target zone T specified on the user interface 18 and also constitute information on the circuit breakers 26 to 34 and the disconnectors 35 to 45 near the power distribution panel 53. By For example, the data storage sections 81 and 82 acquire the ACTIVE / INACTIVE state or the open / closed state of each of the power distribution panels as well as circuit breakers 26 to 34 and disconnectors 35 to 45.
In the following step S14 corresponding to the path R4 in FIG. 1, the main control unit 100 determines whether or not there is a neural network 10 which has completed learning with respect to the target area specified by the user interface 18. When there is no such neural network 10 which has completed learning, the processing proceeds to step S20 which will be described later. On the contrary, when there is a neural network 10 which has completed learning, the processing goes to step S15.
In step S15 corresponding to the path R6 in FIG. 1, the main control unit 100 defines the component (s) and state of the target area T in the deep learning circuits 9 on the basis of the information acquired from the integrated database 2. For example, the main control unit 100 sets the power distribution panel 53 to the OFF state.
In the following step S16, the main control unit 100 produces a list of association profiles of the states of the respective components associated with the target area T on the basis of the information stored in the integrated database 2. For example, the main control unit 100 produces a list of associations representative of the respective ON / OFF states of the circuit breakers 26 to 34 and the disconnectors 35 to 45 which are connected directly or indirectly to the power distribution panel 53 of the zone target T.
In the following step SI7 corresponding to the path R7 in FIG. 1, the main control unit 100 delivers the produced list of association profiles of the respective states of the components relating to the target area T to the neural network 10, which has completed learning and belongs to learning circuits
-26 deepened 9.
In the following step S18, the neural network 10 acquires the state of each of the components of the target area T (that is to say, the components related to the target area T), and acquires the results of analysis such as the influence on other components (i.e., components unrelated to the target area T) and whether an alarm is issued.
In the next step S19 corresponding to the path R20 in FIG. 1, the main control unit 100 extracts a specific state profile of the respective components by deep learning of the neural network 10 and causes the memorization, by the section data storage 82, of the extracted profile. More specifically, the main control unit 100 extracts such an association profile from the respective states of the circuit breakers 26 to 34 and the disconnectors 35 to 45 when the power distribution panel 53 of the target area T is caused to be disabled. After that, processing proceeds to step S30 in Figure 7 which will be described later.
Step S20 in FIG. 6 is the processing to be executed immediately after step S14 when there is no neural network 10 which has completed learning in step S14. In step S20 corresponding to the path R8 in FIG. 1, the training data generation section 11 lists different pieces of information contained in the information acquired from the integrated database 2 or acquires the information that has already been listed. It can be noted that the verb lister mentioned above signifies the processing of acquired data or the execution of a conversion, in the present embodiment.
In the following step S21 corresponding to the path R9 in FIG. 1, the analysis section 4 of the factory simulator 3 acquires the list of different pieces of information.
In the following step S22 corresponding to the path R21 in FIG. 1, the analysis section 4 produces a simulation model of the energy distribution device 25 of the factory on the basis of the data stored in the storage section of data 81.
In the next step S23, the main control unit 100 determines whether to use the deep learning. When the quantity of calculations (i.e. the target determination value) in order to ensure the extraction of a specific profile suitable for isolation work is less than a predetermined threshold value, it that is, when the processing can be performed by a Round-Robin simulation, the main control unit 100 determines not to use the deep learning and advances the processing to step S28 which will be described later. On the contrary, when the quantity of calculations (that is, target value of determination) in order to ensure the extraction of a specific profile suitable for the isolation work is greater than or equal to the threshold value predetermined, i.e., when processing with the use of deep learning is required, the main control unit 100 determines the use of deep learning and advances processing to the step S24.
In step S24 corresponding to the path RIO in FIG. 1, the analysis section 4 of the factory simulator 3 produces data representative of the state of each component and transmits the data produced to the data generation section d learning 11. For example, the analysis section 4 produces data representative of the conduction state of the power distribution panel 53 of the target area T in the case of change of the respective states of all the circuit breakers 26 to 34 and disconnectors 35 to 45.
In the next step S25, the training data generation section 11 of the deep learning circuits 9 produces learning data. For example, the training data generation section 11 produces the first matrix data representative of the respective states of the circuit breakers 26 to 34 and the disconnectors 35 to 45, and further produces the second matrix data representative of the respective states of the power distribution panels 53 to 60.
In the following step S26 corresponding to the path R5 in FIG. 1, the main control unit 100 causes the execution by the neural network
Multilayer 10 of the deep learning circuits 9 of a learning in which the matrix data are treated as training data.
In the following step S27, the deep learning circuits 9 build the neural network 10 which has completed the learning and return the processing to step S15 in FIG. 5.
Step S28 in Figure 6 is the processing to be performed immediately after step S23 when it is determined not to use the deep learning. In step S28 corresponding to the path RI 1 in FIG. 1, the factory simulator 3 positions the components and state of the target area T in the simulation model of the analysis section 4.
In the following step S29, the Round-Robin simulation is implemented and a specific profile suitable for the isolation work is extracted, and then the processing proceeds to step S30 in FIG. 7.
In step S30 of Fig. 7, the main control unit 100 determines whether a specific operational procedure (i.e., a specific operating profile which has been extracted and has been stored in the storage section 81) is necessary for the actual operation of circuit breakers 26 to 34 and disconnectors 35 to 45 or. not. When the specific operational procedure is not necessary, the processing proceeds to step S34 which will be described later. On the contrary, when the specific operational procedure is necessary, the processing proceeds to step S31.
In step S31 corresponding to the paths RI 2 and RI 3 in FIG. 1, the main control unit 100 enters the specific profile memorized in the data storage sections 81 and 82 in the operational procedure extraction section 16 deep learning circuits 9.
In the following step S32 corresponding to the paths R12 and R13 in FIG. 1, the main control unit 100 delivers the rules and logic of the operational procedure relating to the actual operation of the circuit breakers 26 to 34 and the disconnectors 35 to 45 in the operational procedure extraction section 16 of the deep learning circuits 9.
In the next step S33, the procedure extraction section
-29operational 16 specifies and acquires the operational procedure which corresponds to the rules and logic.
In step S34, the main control unit 100 causes the deep learning circuits 9 to produce several plans proposed as a choice on the basis of the specific profile and the operational procedure.
In the next step S35 corresponding to the path RI 5 in FIG. 1, the main control unit 100 delivers the different plans proposed as a choice to the reinforcement learning section 15 of the deep learning circuits 9.
In the next step S36 corresponding to the path RI 5 in FIG. 1, the main control unit 100 enters arbitrary information in the reinforcement learning section 15, which arbitrary information relates to the factory and includes the information of environment acquired from the integrated database 2.
In the following step S37 corresponding to the path RI4 in FIG. 1, the main control unit 100 causes the reward definition section 14 of the deep learning circuits 9 to define a reward with respect to the arbitrary information entered on the factory, and then advances processing to step S38 in Fig. 8. The reward having been defined by the reward definition section 14 is entered into the reinforcement learning section 15, which corresponds to the path R23 on the figure 1. The information on the operational procedure is also entered in the reinforcement learning section 15, which corresponds to the path R24 in figure 1.
In step S38 of Figure 8, the main control unit 100 determines whether the deepening reinforcement learning should be used in order to ensure the extraction of the optimum plane from the different proposed planes or not. When the quantity of calculations (ie, a target value of determination) in order to ensure the extraction of the optimum plan proposed is. below the predetermined threshold, the main control unit 100 determines that the deepening reinforcement learning is useless, then defines a value function by methods such as the Monte Carlo method or the learning of value Q at
The step S40 and then advances the processing to the step S41.
On the contrary, when the quantity of calculations (that is to say, a target value of determination) in order to ensure the extraction of the optimum proposed plan is greater than or equal to the value at the predetermined threshold, i.e. -to say, when it is necessary to carry out the extraction treatment of the optimum proposed plan using the reinforcement learning, the main control unit 100 determines the use of the deepening reinforcement learning, then brings the neural network multilayer 10 to express a value function in step S39, and then advances the processing to step S41.
In step S41 corresponding to the path R16 in FIG. 1, the main control unit 100 causes the reinforcement learning section 15 of the deep learning circuits 9 to specify a value calculated by the value function for each of the different plans offered (that is to say, choices) and delivers information on the specified value to the plan generator 17.
In the following step S42 corresponding to the path R17 in FIG. 1, the plan generator 17 produces the work plan on the basis of the specified proposed plan which has the highest value, and delivers the work plan produced in the section 5 of the factory simulator 3.
In the following step S43 corresponding to the path R22 in FIG. 1, the verification section 5 executes a processing for verifying the influence on the factory in the case of the execution of the isolation work according to the plan of work, based on the data stored in the data storage section 81.
In the next step S44 corresponding to the path R18 in FIG. 1, the verification section 5 determines whether the work plan is appropriate or not. When it is determined that the work plan is suitable, the processing proceeds to step S45 in which this work plan is delivered by the output section 20 of the user interface 18 via the generator. plane 17 as indicated by the path R19 in FIG. 1, and then all of the treatment is completed. On the contrary, when it is determined that the work plan is not suitable, the output section 20 of the user interface 18 performs a
-31 notification that the work plan is not appropriate, and then all processing is completed.
In the present embodiment, determining a first value (i.e., a target value) using a reference value (i.e., a threshold value) can be a determining whether the target value is greater than or equal to the reference value or not.
In addition, or alternatively, the determination of the target value using the reference value can be a determination of whether the target value exceeds the reference value or not.
Additionally, or alternatively, determining the target value using the reference value can be a determination of whether the target value is less than or equal to the reference value or not.
In addition, or alternatively, determining the first value using the reference value can be a determination of whether the target value is less than the reference value or not.
In addition, or alternatively, the reference value is not necessarily fixed and the reference value can be changed. Thus, a predetermined range of values can be used rather than the reference value, and the determination of the target value can be a determination of whether the target value is within the predetermined range or not.
Although a mode in which each step is executed in series is represented on the algorithms of the present embodiment, the order of execution of the respective steps is not necessarily fixed and the order of execution of a part steps can be changed. In addition, some steps can be performed in parallel with another step.
The isolation management device 1 of this embodiment comprises a storage device such as a ROM memory (read only memory) and a RAM memory (random access memory), an external storage device such as an HDD disk ( hard disk) and an SDD (semiconductor disk), a display device such as a display screen, an input device such as a mouse and keyboard, a communication interface, and a
-32 control device which includes a processing unit with a high level of integration such as a specific use electronic chip, an FPGA network (door network programmable on site), a GPU unit (graphics processing unit) and a unit CPU (central processing unit). The isolation management device 1 can be implemented by a hardware configuration using a normal computer.
It can be noted that each program executed on the isolation management device 1 of this embodiment is proposed by being incorporated beforehand in a memory, such as a ROM memory. In addition, or as a variant, each program can be offered by being stored in the form of a file having a format which can be installed or executed on a non-volatile storage medium which can be read by computer, such as a CDROM disc, CD-R disc, memory card, DVD disc and flexible disc (FD).
In addition, each program executed in the isolation management device 1 can be stored on a computer connected to a network such as Internet and can be offered for download via a network. In addition, the isolation management device 1 can also be configured by interconnection and association of separate modules, which independently present the respective functions of the components, via a network or a specific line.
Although a remodeling work of the energy distribution device 25 constituting a part of the plant energy supply device is described by way of example in the present embodiment, the present invention can be applied in the purpose of producing an isolation work plan other than for the power distribution device.
Note that the deep learning circuits 9 can extract the profile with the smallest change occurring at other locations in the form of a specific profile. In this way, it is possible to extract the profile which has the slightest influence on other components (i.e. components unrelated to the target area T) and is
-33the most suitable for isolation work.
According to the embodiments described above, it is possible to efficiently produce a work plan more suitable for isolation work by integrating (a) an analyzer configured so as to analyze patterns of change of state occurring on components at other locations in the event of a change in the state of a component relating to a defined target area and (b) deep learning circuits configured so as to extract a specific profile from several profiles of the change of state analyzed by the analyzer on the basis of in-depth learning.
Although certain embodiments have been described, these embodiments have been presented only by way of example and are not intended to limit the scope of the invention. Obviously, the new methods and devices described here can be implemented in a variety of other forms; in addition, various omissions, substitutions and modifications to the form of the methods and devices described here can be carried out without departing from the spirit of the invention.

权利要求:
Claims (10)
[1" id="c-fr-0001]
1. Isolation management device (1) comprising:
a database (2) configured to store information relating to a factory built with a plurality of components (26), the information comprising a relationship between the plurality of components (26);
a receiver configured to receive target area information defining a target area on the factory;
an analyzer configured to analyze a plurality of respective state profiles of the plurality of components (26) in connection with a change of state of at least one of the plurality of components (26) in the target area , based on the information stored in the database (2);
deep learning circuits (9) configured to extract at least one specific profile from the plurality of profiles analyzed by the analyzer as an extraction profile;
a plan generator (17) configured to produce a work plan based on the extraction profile; and an output interface (18) configured to deliver the work plan produced by the plan generator.
[2" id="c-fr-0002]
2. Isolation management device (1) according to claim 1, further comprising a verifier configured so as to verify the profile of respective states on the components (26) outside of the target area in link with the change of state of each component in the target area according to the work plan.
[3" id="c-fr-0003]
3. Isolation management device (1) according to claim 1 or
2, in which the deep learning circuits (9) comprise an intermediate layer comprising a multilayer neural network and are configured so as to acquire a quantity characteristic of each of the
-35 plurality of profiles; and the deep learning circuits (9) are further configured to extract the extraction profile according to the characteristic magnitude of each of the plurality of profiles.
[4" id="c-fr-0004]
4. The isolation management device (1) according to claim 3, wherein the deep learning circuits (9) include a training data generator configured to produce training data configured to construct the multilayer neural network.
[5" id="c-fr-0005]
5. An isolation management device (1) according to claim 4, wherein the database (2) is configured so as to store information on at least one previous work plan; and the training data generator is configured to produce training data on the basis of the previous work plan stored in the database.
[6" id="c-fr-0006]
6. An isolation management device (1) according to claim 4 or 5, wherein the plurality of components (26) comprises a first predetermined type of component and a second type of component linked to the first type of component;
the training data generator is configured so as to produce first matrix data, in which a state of the first type of component analyzed by the analyzer is treated as an input quantity, and second matrix data, in which a state of the second type of component analyzed by the analyzer is treated as an output quantity; and the deep learning circuits (9) are configured so as to cause the learning, which comprises the first matrix data and the second matrix data, to be learned by the multilayer neural network.
[7" id="c-fr-0007]
7. Isolation management device (1) according to any one of claims 3 to 6,
-In which the deep learning circuits (9) are configured so as to:
define a reward in relation to the information stored in the database, extract a plurality of specific profiles from the plurality of profiles analyzed by the analyzer, in the form of a plurality of extraction profiles, and extract a profile having the highest reward value among the plurality of extraction profiles.
[8" id="c-fr-0008]
8. Isolation management device (1) according to any one of claims 1 to 7, in which the deep learning circuits (9) are configured so as to extract an operational procedure from the isolation work on the base of the extraction profile; and the plan generator (17) is configured to generate the work plan based on the operational procedure extracted by the deep learning circuits.
[9" id="c-fr-0009]
9. An insulation management device (1) according to any one of claims 1 to 8, in which the analyzer is configured so as to execute at least one of an analog circuit analysis, of a logic circuit analysis and path search analysis.
[10" id="c-fr-0010]
10. Isolation management method comprising:
storing information, which relates to a factory built with a plurality of components (26) and defines the relationship between the plurality of components (26) in a database;
receiving target area information defining a target area on the factory;
analyzing a plurality of respective state profiles of the plurality of components (26) in connection with a change of state of at least one of the plurality of components (26) in the target area, based information stored in the database;
Extraction of a specific profile from the plurality of profiles analyzed by the analyzer, as an extraction profile;
generation of a work plan on the basis of the extraction profile; and producing the work plan.
1/8 i
U *
2/8

类似技术:

公开号 | 公开日 | 专利标题

US10057144B2|2018-08-21|Remote system data collection and analysis framework

Groth et al.2019|A hybrid algorithm for developing third generation HRA methods using simulator data, causal models, and cognitive science

Boyko et al.2017|Concept implementation of decision support software for the risk management of complex technical system

Yang et al.2016|Online application of a risk management system for risk assessment and monitoring at NPPs

FR3063369A1|2018-08-31|ISOLATION MANAGEMENT DEVICE AND INSULATION MANAGEMENT METHOD

Levitin et al.2010|Approximation algorithm for evaluating time-to-failure distribution of k-out-of-n system with shared standby elements

Zhu et al.2014|A system-of-systems framework for performance assessment in complex construction projects

US20170248944A1|2017-08-31|A method for managing subsystems of a process plant using a distributed control system

Di Girolamo et al.2020|Operational intelligence for distributed computing systems for exascale science

Trubiani et al.2019|Plus: Performance learning for uncertainty of software

Singh et al.2019|Impact analysis of change in component reliabilities in safety‐critical systems

Alshawish et al.2019|Risk mitigation in electric power systems: Where to start?

Staron et al.2014|Industrial self-healing measurement systems

US10545465B2|2020-01-28|System and method for selecting controllable parameters for equipment operation safety

Kumar et al.2017|Probabilistic analysis of performance measures of redundant systems under Weibull failure and repair laws

Bouchart et al.1998|Is rational decision making appropriate for management of irrigation reservoirs?

Alturkistani et al.2017|An analytical model for reliability evaluation of cloud service provisioning systems

Khatri et al.2011|Using Artificial Neural-Networks in Stochastic Differential Equations Based Software Reliability Growth Modeling.

Jahanshahi et al.2020|Predicting the number of reported bugs in a software repository

Comes2013|Robust emergency management strategies: Supporting interdependent decisions

US10997513B1|2021-05-04|Dynamic, risk informed decision support systems and methods

Teitsma et al.2020|Jiskefet, a bookkeeping application for ALICE

Baumann et al.2017|Simulation driven development—Validation of requirements in the early design stages of complex systems—The example of the German toll system

Parnell2009|Evaluation of risks in complex problems

Elnasr et al.2014|Towards a Framework Empowering Software Testing Process for Better Quality

同族专利:

公开号 | 公开日

CN108508852A|2018-09-07|

KR20200074936A|2020-06-25|

RU2678146C1|2019-01-23|

GB201802549D0|2018-04-04|

CA2996576A1|2018-08-27|

JP2018142060A|2018-09-13|

KR20180099469A|2018-09-05|

JP6789848B2|2020-11-25|

GB2561073B|2020-10-14|

CN108508852B|2021-08-03|

US20180246478A1|2018-08-30|

GB2561073A|2018-10-03|

引用文献:

公开号 | 申请日 | 公开日 | 申请人 | 专利标题

JP2008181283A|2007-01-24|2008-08-07|Toshiba Corp|Erroneous isolation prevention device|

JP2011096029A|2009-10-30|2011-05-12|Hitachi Ltd|Method and system for supporting maintenance operation plan|

CN1019723B|1985-06-11|1992-12-30|三菱电机株式会社|Electric interlocking loop for switching device|

JP3058462B2|1991-03-01|2000-07-04|中部電力株式会社|Power supply route inference device for bus inspection|

JP3231403B2|1992-07-22|2001-11-19|株式会社東芝|Automatic isolation device for busbars|

KR100380308B1|2000-12-12|2003-04-16|한국전기연구원|Switching on/off control system for synchronous circuit breaker using neural network|

JP4059014B2|2001-06-19|2008-03-12|富士電機システムズ株式会社|Optimal plant operation method and optimal plant design method|

JP3619998B2|2002-01-04|2005-02-16|富士通株式会社|Maintenance management system, method and program|

CN1648922A|2005-03-03|2005-08-03|南京科远控制工程有限公司|Enterprise managing integrated managing information control method|

JP5268400B2|2008-03-18|2013-08-21|株式会社東芝|Process management system|

DE102012023424B4|2012-11-29|2019-08-14|Kostal Industrie Elektrik Gmbh|Energy distribution system with a control device|

US10539934B2|2014-03-17|2020-01-21|General Electric Technology Gmbh|Outage and switch management for a power grid system|

KR20150118456A|2014-04-14|2015-10-22|엘에스산전 주식회사|Method for partial discharge diagnosis of monitoring apparatus|

CN104156766A|2014-07-28|2014-11-19|山东山大世纪科技有限公司|Application of intelligent memory and self-learning system aiming at high-voltage switch divide-shut brake time|

CN104538222B|2014-12-27|2016-09-28|中国西电电气股份有限公司|High-voltage switch gear phase-controlled device based on artificial neural network and method|

US20160241031A1|2015-02-18|2016-08-18|Nec Laboratories America, Inc.|Dynamic probability-based power outage management system|

US10243397B2|2015-02-20|2019-03-26|Lifeline IP Holdings, LLC|Data center power distribution|

US10164431B2|2015-03-17|2018-12-25|General Electric Technology Gmbh|Outage management and prediction for a power grid system|

CN104951905A|2015-07-16|2015-09-30|中国神华能源股份有限公司|Equipment dynamic ledger management system|

CN105321039B|2015-09-24|2022-03-01|国家电网公司|Online monitoring data management system and method for isolating switch|

CN105787557B|2016-02-23|2019-04-19|北京工业大学|A kind of deep-neural-network construction design method of computer intelligence identification|CN110998585A|2017-06-22|2020-04-10|株式会社半导体能源研究所|Layout design system and layout design method|

CN110221926B|2019-05-27|2021-11-16|中国电建集团华东勘测设计研究院有限公司|Isolation calculation management method for high arch dam pouring progress simulation|

法律状态:
2020-01-28| PLFP| Fee payment|Year of fee payment: 3 |

2021-01-15| PLFP| Fee payment|Year of fee payment: 4 |

2021-09-24| PLSC| Publication of the preliminary search report|Effective date: 20210924 |

2022-01-27| PLFP| Fee payment|Year of fee payment: 5 |

优先权:

申请号 | 申请日 | 专利标题

JP2017034494|2017-02-27|

JP2017034494A|JP6789848B2|2017-02-27|2017-02-27|Isolation management system and isolation management method|

[返回顶部]