Method and device for monitoring on-line state of actuating mechanism

Abstract

The invention provides an execution mechanism on-line state monitoring method and device, and relates to the technical field of fault monitoring, and the execution mechanism on-line state monitoring method comprises the steps: sending a model request to a central unit, and obtaining an initial local model from the central unit according to the model request; extracting local private historical sample data from the control unit for model training to obtain a first local model; the central unit performs federated learning on the initial parameters of the plurality of first local models to obtain aggregation parameters; performing correction training on the first local model according to the aggregation parameter, and determining a target local model; and receiving the real-time data of the execution mechanism acquired by the control unit, and processing the real-time data according to the target local model to obtain the real-time state of the execution mechanism, thereby solving the technical problem of low accuracy of online state monitoring of the execution mechanism caused by rare local operation data in the prior art.

Description

Technical Field

[0001] The present application relates to the technical field of fault monitoring, and in particular to an online status monitoring method and device for an actuator. Background Art

[0002] The current global economy is showing a trend of digital development. The establishment of equipment asset management and operation and maintenance optimization service products based on industrial cloud platforms aims to improve the reliability and availability of equipment through online monitoring of equipment operating status, model-based health status analysis, and predictive maintenance, reduce passive maintenance, reduce operation and maintenance costs, and improve the availability of power plants.

[0003] Pump and valve actuators are widely used equipment in power plants and other process industrial systems. The performance and health of pump and valve actuators are important factors in ensuring production operation and safety and preventing unplanned downtime and maintenance. Although some operating units have deployed condition monitoring systems for key actuators such as control regulating valves and drive pumps, there are still many pump actuators and valve actuators, especially switch-type pump and valve equipment with switching processes or blocking isolation functions. Due to low input-output and fewer measuring points, online condition monitoring has not yet been achieved. Most of them adopt manual periodic testing and manual periodic inspection maintenance methods, which are inefficient and the risk of actuator failure is difficult to effectively avoid. Summary of the invention

[0004] The present application aims to solve one of the technical problems in the related art at least to some extent.

[0005] To this end, the first objective of the present application is to propose an online status monitoring method for an actuator to achieve accurate online status monitoring.

[0006] The second objective of the present application is to propose another method for online status monitoring of an actuator.

[0007] The third objective of the present application is to provide an online status monitoring device for an actuator.

[0008] The fourth objective of the present application is to provide another actuator online status monitoring device.

[0009] A fifth objective of the present application is to provide an electronic device.

[0010] A sixth object of the present application is to provide a computer-readable storage medium.

[0011] A seventh object of the present application is to provide a computer program product.

[0012] To achieve the above-mentioned purpose, the first embodiment of the present application proposes an online state monitoring method for an actuator, comprising:

[0013] Sending a model request to a central unit, and obtaining an initial local model from the central unit according to the model request;

[0014] Extract historical sample data from the control unit to perform model training on the initial local model to obtain a corresponding first local model;

[0015] Uploading the initial parameters of the first local model to the central unit for federated learning to obtain aggregated parameters;

[0016] receiving the aggregation parameters sent by the central unit, and training the first local model according to the aggregation parameters to determine a target local model;

[0017] The real-time data of the actuator collected by the control unit is received, and the real-time data is processed according to the target local model to obtain the real-time status of the actuator.

[0018] To achieve the above-mentioned purpose, the second embodiment of the present application proposes another method for online status monitoring of an actuator, comprising:

[0019] Receiving a model request sent by one or more computing units, and sending an initial local model to a corresponding computing unit according to the model request;

[0020] Receiving initial parameters of a first local model obtained after training the initial local model and sent by one or more computing units;

[0021] Aggregating the initial parameters to obtain a first aggregation parameter, and determining an aggregation parameter according to the first aggregation parameter and the initial federation parameter;

[0022] The aggregated parameters are sent to one or more computing units to perform online status monitoring of the actuator.

[0023] To achieve the above-mentioned purpose, the third embodiment of the present application proposes an actuator online status monitoring device, comprising:

[0024] A first acquisition module, used for sending a model request to a central unit, and acquiring an initial local model from the central unit according to the model request;

[0025] A second acquisition module is used to extract historical sample data from the control unit to perform model training on the initial local model to obtain a corresponding first local model;

[0026] A third acquisition module is used to upload the initial parameters of the first local model to the central unit for federated learning to obtain aggregated parameters;

[0027] A training module, configured to receive the aggregation parameters sent by the central unit, and train the first local model according to the aggregation parameters to determine a target local model;

[0028] The monitoring module is used to receive the real-time data of the actuator collected by the control unit, and process the real-time data according to the target local model to obtain the real-time status of the actuator.

[0029] To achieve the above-mentioned purpose, the fourth aspect of the present application proposes another actuator online status monitoring device, including:

[0030] A processing module, configured to receive a model request sent by one or more computing units, and send an initial local model to a corresponding computing unit according to the model request;

[0031] A receiving module, configured to receive initial parameters of a first local model obtained after training the initial local model and sent by one or more computing units;

[0032] an aggregation module, configured to aggregate the initial parameters to obtain a first aggregation parameter, and determine an aggregation parameter according to the first aggregation parameter and an initial federation parameter;

[0033] The sending module is used to send the aggregation parameters to one or more computing units to perform online status monitoring of the actuator.

[0034] To achieve the above-mentioned purpose, a fifth aspect of the present application provides an electronic device, including: a processor, and a memory communicatively connected to the processor;

[0035] The memory stores computer-executable instructions;

[0036] The processor executes the computer-executable instructions stored in the memory to implement the method as described in the first aspect or the second aspect.

[0037] To achieve the above-mentioned purpose, the sixth aspect embodiment of the present application proposes a computer-readable storage medium, in which computer execution instructions are stored. When the computer execution instructions are executed by a processor, they are used to implement the method described in the first aspect or the second aspect embodiment.

[0038] To achieve the above-mentioned purpose, the seventh aspect of the present application proposes a computer program product, including a computer program, which implements the method described in the first aspect or the second aspect when executed by a processor.

[0039] The method and device for online status monitoring of an actuator provided in the present application obtain an initial local model from a central unit, and train the initial local model based on local historical sample data of each power plant or other process industrial system to obtain a first local model. The initial parameters of the first local model are uploaded to the central unit for federated learning to obtain aggregated parameters, which have a better learning effect and aggregate data information in multiple power plants or other process industrial systems. The aggregated parameters are fed back to the computing unit and the first local model is corrected and trained based on the aggregated parameters. Even when the number of samples is small, a target local model with better learning and training effects can be obtained. The target local model is used to perform online status monitoring to determine the real-time status of the actuator, thereby improving the accuracy of fault risk monitoring of the actuator and reducing the occurrence of accidents.

[0040] Additional aspects and advantages of the present application will be given in part in the description below, and in part will become apparent from the description below, or will be learned through the practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] The above and / or additional aspects and advantages of the present application will become apparent and easily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which:

[0042] Figure 1 A schematic diagram of a flow chart of an actuator online status monitoring method provided in an embodiment of the present application;

[0043] Figure 2 A flow chart of another method for online status monitoring of an actuator provided in an embodiment of the present application;

[0044] Figure 3 A schematic diagram of a flow chart of an actuator online status monitoring method provided in an embodiment of the present application;

[0045] Figure 4 A system block diagram is provided for realizing an online condition monitoring method of an actuator;

[0046] Figure 5 A logic block diagram of federated learning proposed in an embodiment of the present application;

[0047] Figure 6 A logic block diagram of a model update proposed in an embodiment of the present application;

[0048] Figure 7 A schematic diagram of the structure of an actuator online status monitoring device provided in an embodiment of the present application;

[0049] Figure 8 A schematic diagram of the structure of another actuator online status monitoring device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0050] Embodiments of the present application are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present application, and should not be construed as limiting the present application.

[0051] This embodiment is suitable for industrial production scenarios with a small number of samples, such as the switch-type equipment actuator scenario in a power plant. The switch-type actuator generally acts mainly during the system start-up and shutdown process, such as the auxiliary shut-off valve, relief valve, and purge valve in the fuel gas system of the gas turbine. Before the ignition of the engine startup process, the auxiliary shut-off valve is opened and the relief valve is closed at the same time to allow the fuel gas to enter the fuel gas pipeline for ignition and combustion; during the normal or shutdown process of the engine, the auxiliary shut-off valve will be closed and the relief valve will be opened at the same time to stop or block the fuel gas, and the residual fuel gas will be emptied through the relief valve; and the purge valve is used to ensure that there is no combustible gas in the fuel gas pipeline by nitrogen purge before the fuel mode switching occurs. Due to the damage of the internal driving components of the actuator and the aging of the seals, the actuator will be stuck, leaked, etc., resulting in slow action of the actuator, inability to switch, and inability to close tightly. However, compared with the control and regulation type pump and valve actuator, due to the low action frequency and less operation data information of the switch-type pump and valve actuator, it is difficult to establish a more accurate state monitoring model and realize fault warning. The following describes the online status monitoring method and device of the actuator according to the embodiment of the present application with reference to the accompanying drawings.

[0052] Figure 1 The following is a flow chart of an online state monitoring method for an actuator provided in an embodiment of the present application. Figure 1 As shown, the method is executed by a computing unit and includes the following steps:

[0053] S101, sending a model request to a central unit, and obtaining an initial local model from the central unit according to the model request.

[0054] Optionally, the central unit may be composed of a server with certain computing power and data storage functions, and data interaction and communication may be performed between the central unit and the computing unit.

[0055] In some implementations, the model request is used to request the initial local model data in the central unit, and after receiving the model request from the computing unit, the central unit returns the initial local model to the corresponding computing unit. Optionally, the initial local model can be modeled using a machine learning algorithm such as a support vector machine or a neural network.

[0056] S102, extracting historical sample data from the control unit to perform model training on the initial local model to obtain a corresponding first local model.

[0057] Optionally, the computing unit can be composed of a server with certain computing power and data storage functions, which has the ability to exchange information and communicate with the control unit, and can obtain and store real-time operating data related to the actuator status monitoring from the control unit; wherein the actuator refers to a mechanism that converts the control signal into a corresponding action, such as valve actuators including but not limited to: diaphragm actuators, piston actuators with valve positioners, butterfly valves, and angle valves.

[0058] In some implementations, each power plant or other process industrial system corresponds to a control unit. In this embodiment, taking the power plant as an example, each power plant corresponds to a control unit. The control unit is used to collect and store sample data of the on-site actuator. The sample data may include but is not limited to position feedback of the actuator, current information, fluid pressure, flow or temperature of the pipeline where the actuator is located, and may also include status information required for the operation of the power plant or other process industrial system.

[0059] In some implementations, the control unit may also collect information based on the operational control requirements of a power plant or other process industrial system and the status information obtained, and issue instructions to control the actions of the actuators; for example, during the pressure monitoring process, if the pressure does not exceed the threshold condition, the actuator is always in the closed state, and the actuator is opened when a higher pressure is detected, and the instructions for opening the actuator and the subsequent closing of the actuator are recorded to ensure that more comprehensive and accurate sample information is obtained from the control unit for model training, thereby improving the accuracy of model training.

[0060] Optionally, the computing unit obtains and stores sample data of the corresponding power plant from the control unit to obtain all historical sample data of the power plant, and performs initial local model training based on the historical sample data to obtain the optimal solution for the parameters of the target cost function, that is, the local optimal solution for the private data of the power plant, thereby obtaining the corresponding first local model. The model parameters in the first local model are the initial parameters after training.

[0061] S103, uploading the initial parameters of the first local model to the central unit for federated learning to obtain aggregated parameters.

[0062] The central unit may receive the initial parameters of the first local model sent by the computing unit, and perform federated learning on the initial parameters; it is understandable that there may be one or more computing units, which are consistent with the number of power plants.

[0063] In some implementations, the central unit may include a federal model, which may be a support vector machine or a neural network model, etc., which is used to perform federated learning on the initial parameters sent by each computing unit to obtain aggregate parameters; optionally, the aggregate parameters may be model parameters corresponding to the federal model, and the initial parameters sent by each computing unit may be fused to obtain the model parameters of the federal model, that is, the aggregate parameters, for example, the weighted sum of the initial parameters sent by each computing unit is determined as the aggregate parameter.

[0064] S104, receiving the aggregation parameters sent by the central unit, and training the first local model according to the aggregation parameters to determine the target local model.

[0065] It can be understood that after obtaining the aggregation parameters, the central unit sends the aggregation parameters to each computing unit, and the computing unit trains the first local model again according to the received aggregation parameters, so as to determine the target local model.

[0066] Optionally, the model parameters of each first local model can be updated with aggregated parameters, and the first local model with updated model parameters can be trained again based on historical sample data until the training is completed to obtain a target local model. The target local model learns the sample characteristics of the power plant corresponding to each computing unit, and determines the target local model with better learning effect without the need for common historical sample data, thereby providing a better analysis basis for real-time monitoring of subsequent actuators.

[0067] S105, receiving the real-time data of the actuator collected by the control unit, and processing the real-time data according to the target local model to obtain the real-time status of the actuator.

[0068] It can be understood that the control unit can collect real-time data of the actuators in the corresponding power plant in real time. The real-time data may include but is not limited to position feedback of the actuator, current information, fluid pressure, temperature or temperature of the pipeline where the actuator is located, and other information, as well as status information required for the operation of the power plant or other process industrial systems, and may also include action instructions for controlling the actuator.

[0069] The real-time data is input into the pre-trained target local model, and the target local model outputs the real-time status of the actuator, that is, to perform online actuator status monitoring and fault warning.

[0070] In this embodiment, the initial local model is obtained through the central unit, and the initial local model is trained according to the historical sample data in the computing unit corresponding to each power plant and the control unit itself to obtain a first local model. The initial parameters of the first local models of the same power plants are uploaded to the central unit for federated learning to obtain aggregated parameters. The aggregated parameters have learned the initial parameters of multiple power plants and have better learning effect. The aggregated parameters are fed back to the computing unit and the first local model is corrected and trained according to the aggregated parameters to obtain a target local model with better learning and training effect. The target local model performs online status monitoring to determine the real-time status of the actuator, improve the accuracy of fault risk monitoring of the actuator, and reduce accidents.

[0071] Based on the above embodiments, Figure 2 A flow chart of another method for online status monitoring of an actuator provided in an embodiment of the present application. Figure 2 As shown, the method is executed by a computing unit and includes the following steps:

[0072] S201, sending a model request to a central unit, and obtaining an initial local model from the central unit according to the model request.

[0073] In the embodiment of the present application, the implementation method of step S201 can be implemented by any of the methods in the embodiments of the present disclosure, which is not limited here and will not be repeated.

[0074] S202, extracting historical sample data from the control unit to perform model training on the initial local model to obtain a corresponding first local model.

[0075] In the embodiment of the present application, the implementation method of step S202 can be implemented by any of the methods in the embodiments of the present disclosure, which is not limited here and will not be repeated.

[0076] S203, uploading the initial parameters of the first local model to the central unit for federated learning to obtain aggregated parameters.

[0077] Optionally, the initial parameters can be federated learned by an aggregation algorithm, and the initial parameters of the federated model can be modified and optimized to obtain the aggregation parameters. The calculation of the aggregation parameters can be:

[0078] W c,i+1 =W c，i +f(W e1,i+1 , W e2,i+1 , ..., W eN,i+1 )

[0079] Among them, W c，i+1 is the aggregation parameter; W c，i is the initial parameter of the federated model; We1，i+1 The initial parameters uploaded by computing unit e1; W e2，i+1 The initial parameters uploaded by computing unit e2; W eN,i+1 is the initial parameter uploaded by the computing unit eN, where N is the number of computing units, that is, the number of corresponding power plants; f(W e1，i+1 , W e2，i+1 , ..., W eN,i+1 ) is a parameter aggregation function, which can be aggregated by weighted summation or averaging.

[0080] In the embodiments of the present disclosure, the implementation method of step S203 may be implemented in any manner in the embodiments of the present disclosure, which is not limited here and will not be described in detail.

[0081] S204, receiving the aggregation parameters sent by the central unit, replacing the model parameters of the first local model with the aggregation parameters, and obtaining a second local model.

[0082] Optionally, key rules can also be obtained, and aggregation parameters sent by the central unit can be received according to the key rules; that is, the interaction of model data between the central unit and the computing unit is realized through specific keys in the key rules, thereby achieving effective isolation of the local private data of the power plant in the computing unit and ensuring the data security of private data.

[0083] After acquiring the aggregation parameters sent by the central unit according to the key rule, the model parameters of the first local model are replaced with the aggregation parameters to obtain the second local model.

[0084] S205: Train the second local model based on the historical sample data until the training conditions are met to obtain a target local model.

[0085] A new round of model training and parameter optimization is performed on the second local model until a target local model is obtained when the training conditions are met; optionally, the training conditions include model convergence or the number of training times reaching a maximum threshold, that is, the training is determined to be finished when the model converges, or the training is determined to be finished when the number of training times reaches a preset maximum threshold, and the obtained target local model is a trained model that can be used for online status monitoring of the actuator.

[0086] In some implementations, the actual accuracy corresponding to the target local model can also be determined; in response to the actual accuracy being less than the ideal accuracy threshold, the model parameters of one or more target local models are uploaded to the central unit for federated learning to obtain new aggregation parameters; the new aggregation parameters are received, and the target local model is trained according to the new aggregation parameters to determine the pre-trained target local model.

[0087] As an example, verification data can be obtained to determine the actual accuracy of the online monitoring of the target local model, and the current actual accuracy is compared with the ideal accuracy threshold. If the actual accuracy is greater than or equal to the ideal accuracy threshold, it is determined that the monitoring effect of the current online monitoring of the target local model is good. If the actual accuracy is less than the accuracy threshold, it is determined that the monitoring effect of the current online monitoring of the target local model is average. The model parameters of the target local models corresponding to each power plant can be uploaded to the central unit again, and the central unit will perform federated learning on the model parameters and return new aggregation parameters. The computing unit will train the target local model again according to the new aggregation parameters to obtain a target local model with better training effect, so as to improve the accuracy of online status monitoring.

[0088] In some implementations, an optimization cycle can also be set. After the optimization cycle is reached, the model parameters of the target local model are uploaded to the central unit for federated learning to obtain new aggregation parameters. The target local model is trained and optimized again based on the new aggregation parameters to obtain a target local model with better training effect, so as to improve the accuracy of online status monitoring.

[0089] S206, receiving the real-time data of the actuator collected by the control unit, and processing the real-time data according to the target local model to obtain the real-time status of the actuator.

[0090] In the embodiment of the present application, the implementation method of step S206 can be implemented by any of the methods in the embodiments of the present disclosure, which is not limited here and will not be repeated.

[0091] In this embodiment, an initial local model is obtained by a central unit, and the initial local model is trained according to local historical data of each computing unit to obtain a first local model. The central unit aggregates the initial parameters of the first local model and obtains aggregated parameters in combination with the initial parameters of the federated model. After receiving the aggregated parameters sent by the central unit, the model parameters of the first local model are replaced with the aggregated parameters to obtain a second local model. The second local model is continuously trained and learned until the training conditions are met to obtain a target local model with better training effect. Model data interaction between the central unit and the computing unit is achieved through key rules, and the local private data of the power plant corresponding to the computing unit is effectively isolated to ensure the data security of the private data. The online status monitoring of the actuator is performed based on the target local model with better training effect, so as to improve the accuracy of fault risk monitoring of the actuator and reduce the occurrence of accidents.

[0092] Figure 3 The following is a flow chart of an online state monitoring method for an actuator provided in an embodiment of the present application. Figure 3 As shown, the method is performed by a central unit and includes the following steps:

[0093] S301, receiving a model request sent by one or more computing units, and sending an initial local model to a corresponding computing unit according to the model request.

[0094] Optionally, the central unit may be composed of a server with certain computing power and data storage functions, and data interaction and communication may be performed between the central unit and the computing unit.

[0095] In some implementations, the model request is used to request the initial local model data in the central unit. After receiving the model request from the computing unit, the central unit returns the initial local model to the corresponding computing unit. It can be understood that the central unit includes the initial local model, and the initial local model can be modeled using machine learning algorithms such as support vector machines or neural networks.

[0096] S302, receiving initial parameters of a first local model obtained after training the initial local model from one or more computing units.

[0097] Optionally, the computing unit can be composed of a server with certain computing power and data storage functions, which has the ability to exchange information and communicate with the control unit, and can obtain and store real-time operating data related to the actuator status monitoring from the control unit; wherein the actuator refers to a mechanism that converts the control signal into a corresponding movement, such as valve actuators including but not limited to: diaphragm actuators, piston actuators with valve positioners, butterfly valves, and angle valves.

[0098] Optionally, each power plant may correspond to a control unit, which is used to collect and store sample data of on-site actuators. The sample data may include but is not limited to position feedback of the actuator, current information, fluid pressure, flow or temperature of the pipeline where the actuator is located, and may also include status information required for the operation of the power plant or other process industrial systems.

[0099] In some implementations, the first local model is a model that is initially trained in the computing unit using local private data of the power plant. Since the amount of local private data in the power plant is small, the accuracy of the first local model is low. Therefore, each computing unit uploads the initial parameters of its corresponding first local model to the central unit, and the central unit performs federated learning to achieve more accurate online status monitoring.

[0100] S303: Aggregate the initial parameters to obtain a first aggregation parameter, and determine an aggregation parameter according to the first aggregation parameter and the initial federation parameter.

[0101] Optionally, a weighted summation may be performed on the initial parameters to obtain a first aggregation parameter; or an average value of the initial parameters may be calculated to obtain the first aggregation parameter; and a summation result of the first aggregation parameter and the initial federation parameter may be obtained as the aggregation parameter.

[0102] For example, the aggregation parameters can be expressed as:

[0103] W c，i+1 =W c，i +f(W e1，i+1 , W e2，i+1 , ..., W eN，i+1 )

[0104] Among them, W c，i+1 is the aggregation parameter; W c，i is the initial parameter of the federated model; W e1，i+1 The initial parameters uploaded by computing unit e1; W e2，i+1 The initial parameters uploaded by computing unit e2; W eN，i+1 is the initial parameter uploaded by the computing unit eN, where N is the number of computing units, that is, the number of corresponding power plants; f(W e1，i+1 , W e2，i+1 , ..., W eN，i+1 ) is a parameter aggregation function, and the first aggregation parameter is obtained by aggregation through weighted summation or average value.

[0105] S304, sending the aggregated parameters to one or more computing units to perform online status monitoring of the actuator.

[0106] Optionally, the required aggregation parameters may be sent to the computing unit according to the key rule to improve the privacy and security of data transmission.

[0107] It can be understood that the computing unit can perform model training and learning by aggregating parameters to obtain a pre-trained target local model, and perform online status monitoring of the actuator based on the target local model.

[0108] In this embodiment, by acquiring the initial parameters of the first local model sent by the computing unit, the initial parameters of each computing unit are weightedly summed to obtain the first aggregation parameter, and the sum of the first aggregation parameter and the initial federal parameter of the federal model is used as the aggregation parameter. The aggregation parameter integrates the characteristics of the first local models of multiple power plants, and the fusion parameter is fed back to the computing unit to acquire the target local model and monitor the online status of the actuator, thereby improving the accuracy of fault risk monitoring of the actuator.

[0109] Figure 4The present invention is a system block diagram that can be used to implement an online status monitoring method for an actuator, including a central cloud (central unit), an edge computing unit (computing unit) and a control system (control unit), wherein the central cloud includes an actuator status monitoring system, that is, a federated model, and each edge computing unit includes a corresponding local model after training the initial local model. The local model is trained with the private data of the corresponding power plant. The private data of the power plant can be collected and stored by the control system. It can be understood that the power plant is an exemplary scenario in this embodiment, and online status monitoring of the actuator can also be performed for other industrial system scenarios with less private data.

[0110] Figure 5 A logical block diagram of federated learning proposed in an embodiment of the present application, wherein each edge computing unit uploads the model parameters of the local model to the central cloud, and the local model is initially updated through the private data in the database corresponding to the power plant. The central cloud performs federated learning and model updates on the parameters of each sub-model, and transmits down the federated model parameters and keys. The computing unit can interact with the central cloud for data based on the keys to ensure the security of private data.

[0111] Figure 6 A logic block diagram of a model update proposed in an embodiment of the present application, the central cloud determines the initial federated model and obtains the model parameters of other edge computing units, calculates the aggregated parameters of the federated model, and sends the aggregated parameters to the edge computing units. The edge computing units update the local model based on the aggregated parameters, train the initial model with local private operating data, obtain the model at the current optimal solution, and determine that the model training is completed when the model converges or reaches a preset number of times to obtain the target local model. The model data can also be uploaded to the central cloud again during the training process that has not reached the preset number of times, and the central cloud will calculate the aggregated parameters and provide feedback to obtain a more accurate target local model and improve the model training effect.

[0112] In order to implement the above embodiment, the present application also proposes an online status monitoring device for an actuator.

[0113] Figure 7 This is a schematic diagram of the structure of an actuator online status monitoring device provided in an embodiment of the present application. Figure 7 As shown, the actuator online state monitoring device 700 includes:

[0114] The first acquisition module 701 is used to send a model request to the central unit, and acquire an initial local model from the central unit according to the model request;

[0115] A second acquisition module 702 is used to extract historical sample data from the control unit to perform model training on the initial local model to obtain a corresponding first local model;

[0116] The third acquisition module 703 is used to upload the initial parameters of the first local model to the central unit for federated learning to obtain aggregated parameters;

[0117] A training module 704 is used to receive the aggregation parameters sent by the central unit, and train the first local model according to the aggregation parameters to determine the target local model;

[0118] The monitoring module 705 is used to receive the real-time data of the actuator collected by the control unit, and process the real-time data according to the target local model to obtain the real-time status of the actuator.

[0119] Furthermore, in a possible implementation of the embodiment of the present application, the training module 704 includes:

[0120] Replacing the model parameters of the first local model with the aggregation parameters to obtain a second local model;

[0121] The second local model is trained based on the historical sample data until the training conditions are met to obtain the target local model.

[0122] Furthermore, in a possible implementation of the embodiment of the present application, the training condition includes model convergence, or the number of training times reaches a maximum threshold number of times.

[0123] Furthermore, in a possible implementation of the embodiment of the present application, the apparatus 700 further includes:

[0124] Determine the actual accuracy corresponding to the target local model;

[0125] In response to the actual accuracy being less than the desired accuracy threshold, uploading model parameters of one or more target local models to the central unit for federated learning to obtain new aggregated parameters;

[0126] A new aggregation parameter is received, and a target local model is trained according to the new aggregation parameter to determine a pre-trained target local model.

[0127] Furthermore, in a possible implementation of the embodiment of the present application, the apparatus 700 further includes:

[0128] The key rule is obtained, and the aggregation parameter sent by the central unit is received according to the key rule.

[0129] It should be noted that the above explanations of the embodiment of the method for online status monitoring of an actuator are also applicable to the online status monitoring device for an actuator of this embodiment, and will not be repeated here.

[0130] In this embodiment, an initial local model is obtained by a central unit, and the initial local model is trained according to local historical data of each computing unit to obtain a first local model. The central unit aggregates the initial parameters of the first local model and obtains aggregated parameters in combination with the initial parameters of the federated model. After receiving the aggregated parameters sent by the central unit, the model parameters of the first local model are replaced with the aggregated parameters to obtain a second local model. The second local model is continuously trained and learned until the training conditions are met to obtain a target local model with better training effect. Model data interaction between the central unit and the computing unit is achieved through key rules, and the local private data of the power plant corresponding to the computing unit is effectively isolated to ensure the data security of the private data. The online status monitoring of the actuator is performed based on the target local model with better training effect, so as to improve the accuracy of fault risk monitoring of the actuator and reduce the occurrence of accidents.

[0131] Figure 8 This is a schematic diagram of the structure of another actuator online status monitoring device provided in an embodiment of the present application. Figure 8 As shown, the actuator online state monitoring device 800 includes:

[0132] The processing module 801 is used to receive a model request sent by one or more computing units, and send an initial local model to a corresponding computing unit according to the model request;

[0133] A receiving module 802 is used to receive initial parameters of a first local model obtained after training the initial local model and sent by one or more computing units;

[0134] Aggregation module 803, used to aggregate the initial parameters to obtain a first aggregation parameter, and determine the aggregation parameter according to the first aggregation parameter and the initial federation parameter;

[0135] The sending module 804 is used to send the aggregated parameters to one or more computing units to perform online status monitoring of the actuator.

[0136] Furthermore, in a possible implementation of the embodiment of the present application, the apparatus 800 includes:

[0137] Performing weighted summation on the initial parameters to obtain a first aggregation parameter;

[0138] Or, calculating the average value of the initial parameters to obtain the first aggregation parameter;

[0139] A sum of the first aggregation parameter and the initial federation parameter is obtained as the aggregation parameter.

[0140] In this embodiment, by acquiring the initial parameters of the first local model sent by the computing unit, the initial parameters of each computing unit are weightedly summed to obtain the first aggregation parameter, and the sum of the first aggregation parameter and the initial federal parameter of the federal model is used as the aggregation parameter. The aggregation parameter integrates the characteristics of the first local models of multiple power plants, and the fusion parameter is fed back to the computing unit to acquire the target local model and monitor the online status of the actuator, thereby improving the accuracy of fault risk monitoring of the actuator.

[0141] In order to implement the above embodiments, the present application also proposes an electronic device, comprising: a processor, and a memory communicatively connected to the processor; the memory stores computer-executable instructions; the processor executes the computer-executable instructions stored in the memory to implement the method provided by the above embodiments.

[0142] In order to implement the above embodiments, the present application also proposes a computer-readable storage medium, in which computer-executable instructions are stored. When the computer-executable instructions are executed by a processor, they are used to implement the methods provided by the above embodiments.

[0143] In order to implement the above embodiments, the present application also proposes a computer program product, including a computer program, which implements the methods provided by the above embodiments when executed by a processor.

[0144] The collection, storage, use, processing, transmission, provision and disclosure of user personal information involved in this application are in compliance with relevant laws and regulations and do not violate public order and good morals.

[0145] It should be noted that personal information from users should be collected for legitimate and reasonable purposes and should not be shared or sold outside of these legitimate uses. In addition, such collection / sharing should be carried out after receiving the user's informed consent, including but not limited to notifying the user to read the user agreement / user notice and sign the agreement / authorization including authorization of relevant user information before the user uses the function. In addition, any necessary steps should be taken to protect and safeguard access to such personal information data and ensure that others who have access to personal information data comply with its privacy policy and procedures.

[0146] The present application is expected to provide an implementation scheme for users to selectively block the use or access of personal information data. That is, the present disclosure is expected to provide hardware and / or software to prevent or block access to such personal information data. Once the personal information data is no longer needed, the risk can be minimized by limiting data collection and deleting the data. In addition, when applicable, such personal information is de-identified to protect the privacy of the user.

[0147] In the description of the aforementioned embodiments, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, without contradiction.

[0148] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In the description of this application, the meaning of "plurality" is at least two, such as two, three, etc., unless otherwise clearly and specifically defined.

[0149] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, fragment or portion of code comprising one or more executable instructions for implementing the steps of a custom logical function or process, and the scope of the preferred embodiments of the present application includes alternative implementations in which functions may not be performed in the order shown or discussed, including performing functions in a substantially simultaneous manner or in the reverse order depending on the functions involved, which should be understood by technicians in the technical field to which the embodiments of the present application belong.

[0150] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in combination with these instruction execution systems, devices or apparatuses. For the purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in combination with these instruction execution systems, devices or apparatuses. More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic devices), a portable computer disk box (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), a fiber optic device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium and then editing, interpreting or processing in other suitable ways if necessary, and then stored in a computer memory.

[0151] It should be understood that the various parts of the present application can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0152] A person skilled in the art may understand that all or part of the steps in the method for implementing the above-mentioned embodiment may be completed by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, which, when executed, includes one or a combination of the steps of the method embodiment.

[0153] In addition, each functional unit in each embodiment of the present application may be integrated into a processing module, or each unit may exist physically separately, or two or more units may be integrated into one module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of a software functional module. If the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it may also be stored in a computer-readable storage medium.

[0154] The storage medium mentioned above may be a read-only memory, a magnetic disk or an optical disk, etc. Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limiting the present application. A person of ordinary skill in the art may change, modify, replace and modify the above embodiments within the scope of the present application.

Claims

1. A method for online status monitoring of an actuator, characterized in that: The process is executed by a computing unit and includes the following steps: Sending a model request to a central unit, and obtaining an initial local model from the central unit according to the model request; Extract historical sample data from the control unit to perform model training on the initial local model to obtain a corresponding first local model; Uploading the initial parameters of the first local model to the central unit for federated learning to obtain aggregated parameters; receiving the aggregation parameters sent by the central unit, and training the first local model according to the aggregation parameters to determine a target local model; The real-time data of the actuator collected by the control unit is received, and the real-time data is processed according to the target local model to obtain the real-time status of the actuator.

2. The method according to claim 1, characterized in that The training of the first local model according to the aggregation parameter to determine the target local model includes: Replacing the model parameters of the first local model with the aggregation parameters to obtain a second local model; The second local model is trained based on the historical sample data until a training condition is met to obtain a target local model.

3. The method according to claim 2, characterized in that The training conditions include model convergence, or the number of training times reaching a maximum threshold.

4. The method according to claim 3, characterized in that The method further comprises: Determining an actual accuracy corresponding to the target local model; In response to the actual accuracy being less than an ideal accuracy threshold, uploading model parameters of one or more target local models to the central unit for federated learning to obtain new aggregated parameters; A new aggregation parameter is received, and a target local model is trained according to the new aggregation parameter to determine a pre-trained target local model.

5. The method according to any one of claims 1 to 4, characterized in that The method further comprises: Acquire a key rule, and receive an aggregation parameter sent by the central unit according to the key rule.

6. A method for online status monitoring of an actuator, characterized in that: Executed by the central unit, it includes the following steps: Receiving a model request sent by one or more computing units, and sending an initial local model to a corresponding computing unit according to the model request; Receiving initial parameters of a first local model obtained after training the initial local model and sent by one or more computing units; Aggregating the initial parameters to obtain a first aggregation parameter, and determining an aggregation parameter according to the first aggregation parameter and the initial federation parameter; The aggregated parameters are sent to one or more computing units to perform online status monitoring of the actuator.

7. The method according to claim 5, characterized in that The method for obtaining the aggregation parameters includes: Performing weighted summation on the initial parameters to obtain the first aggregation parameter; Or, calculating the average value of the initial parameters to obtain the first aggregation parameter; A sum of the first aggregation parameter and the initial federation parameter is obtained as the aggregation parameter.

8. An actuator online status monitoring device, characterized in that: include: A first acquisition module, used for sending a model request to a central unit, and acquiring an initial local model from the central unit according to the model request; A second acquisition module is used to extract historical sample data from the control unit to perform model training on the initial local model to obtain a corresponding first local model; A third acquisition module is used to upload the initial parameters of the first local model to the central unit for federated learning to obtain aggregated parameters; A training module, configured to receive the aggregation parameters sent by the central unit, and train the first local model according to the aggregation parameters to determine a target local model; The monitoring module is used to receive the real-time data of the actuator collected by the control unit, and process the real-time data according to the target local model to obtain the real-time status of the actuator.

9. An actuator online status monitoring device, characterized in that: include: A processing module, configured to receive a model request sent by one or more computing units, and send an initial local model to a corresponding computing unit according to the model request; A receiving module, configured to receive initial parameters of a first local model obtained after training the initial local model and sent by one or more computing units; an aggregation module, configured to aggregate the initial parameters to obtain a first aggregation parameter, and determine an aggregation parameter according to the first aggregation parameter and an initial federation parameter; The sending module is used to send the aggregation parameters to one or more computing units to perform online status monitoring of the actuator.

10. An electronic device, characterized in that: include: A processor, and a memory communicatively connected to the processor; The memory stores computer-executable instructions; The processor executes the computer-executable instructions stored in the memory to implement the method according to any one of claims 1-5 or 6-7.

Images