Intelligent power plant management method and system based on distributed management

Abstract

The invention provides an intelligent power plant management method and system based on distributed management, and the method comprises the steps: an initial distribution stage: a center node carries out the initial training of a first center model and a second center model based on an initial training data set, and transmits the model parameters after the initial training to each edge node; the edge node judges whether manual inspection is carried out or not based on environment parameters and equipment operation parameters which are acquired in real time; and an iteration updating stage: in the iteration updating stage, the center node obtains the inspection result of manual detection, correspondingly stores the inspection result of each monitoring position in a preset time period as an iteration training data set, and performs iteration on the basis of the iteration training data set. And the center node performs iterative training on the first center model and the second center model based on the iterative training data set, and model parameters of the first center model and the second center model after iterative training are issued to each edge node.

Description

Technical Field

[0001] The present invention relates to the technical field of smart power plants, and in particular to a smart power plant management method and system based on distributed management. Background Art

[0002] The importance of adopting a distributed architecture in smart power plants lies in its ability to significantly improve the operational efficiency of power plants, enhance the flexibility and reliability of the system, and promote the power plants to move towards intelligence and automation, so as to better adapt to the diversification and complexity of future energy development trends.

[0003] First, the distributed architecture provides a high degree of flexibility for smart power plants. Under the distributed architecture, the various system components of the power plant can be operated and managed relatively independently, which enables the power plant to quickly adjust the system configuration and optimize resource allocation according to actual needs and operating conditions. Secondly, the distributed architecture significantly enhances the reliability of smart power plants. In a distributed system, the various components exchange information and work together through a communication network. This architecture helps reduce the impact of single point failures on the entire system. In addition, the distributed architecture also supports redundant deployment and load balancing, further improving the stability and reliability of the system. This reliability is critical for power plants because any downtime or failure may have a serious impact on the stable operation of the power grid and the power supply of users.

[0004] Smart power plants with distributed architectures can analyze the situation at edge nodes to determine whether manual inspection is needed and improve operational safety. Existing distributed architectures often train models uniformly at central nodes and send the trained model parameters to edge nodes. However, the components at different locations in a power plant are different, so their normal operating standards also vary. The existing architecture makes it difficult to guarantee judgment accuracy.

[0005] In view of this, the present invention is proposed. Summary of the invention

[0006] The purpose of the present invention is to provide a smart power plant management method and system based on distributed management. This scheme sets two judgment models at each edge node. On the one hand, the judgment accuracy is ensured by increasing the model. On the other hand, the judgment accuracy is ensured through adaptive training of the corresponding edge nodes.

[0007] The present invention provides a smart power plant management method based on distributed management. The method is based on a distributed architecture and includes a central node and edge nodes corresponding to each monitoring location. The central node is provided with a first central model based on environmental parameter determination and a second central model based on equipment operation parameter determination. The steps of the method include:

[0008] An initial allocation stage, in which the central node performs initial training on the first central model and the second central model based on the initial training data set, and sends the model parameters of the first central model and the second central model after the initial training to each edge node, so that the edge node constructs the first edge model and the second edge model;

[0009] The edge node constructs a first vector and a second vector based on the environmental parameters and equipment operation parameters collected in real time, and determines whether to perform manual inspection based on the corresponding first edge model and second edge model;

[0010] An iterative update phase, in which the central node obtains the inspection results of manual detection, and stores the inspection results of each monitoring position in a preset time period as an iterative training data set. Based on the iterative training data set, the central node iteratively trains the first central model and the second central model based on the iterative training data set, and sends the model parameters of the iteratively trained first central model and the second central model to each edge node, so that the edge node updates the first edge model and the second edge model;

[0011] The edge model re-determines in real time whether to perform manual inspection based on the updated first edge model and the second edge model, and continuously performs iterative updates.

[0012] Using the above scheme, the central node of this scheme performs unified training in the initial stage, and uniformly sends the trained model parameters to each edge node. Each edge node builds an initial model based on the same model parameters. In the judgment process, since the two models are trained based on different parameters, the two models may obtain different results at the same time. The staff performs manual inspection based on the manual inspection instructions issued by any model at the same monitoring location, and gives the numerical value of the necessity of inspection after manual inspection and uploads it to the central node. Therefore, the two models can correspond to each other and construct a data set that is different from their own judgment results, providing a more accurate basis for subsequent iterative updates. In addition, this scheme can train each edge node separately to ensure processing accuracy.

[0013] In some embodiments of the present invention, in the iterative update phase, the central node obtains the inspection results of manual detection, and stores the inspection results of each monitoring location in a preset time period as an iterative training data set:

[0014] Respectively obtain instruction records of the first edge model and the second edge model in determining manual inspection;

[0015] Obtain inspection necessity parameters uploaded by the staff corresponding to each instruction record;

[0016] constructing the inspection determination parameter output by the first edge model or the second edge model when determining to perform manual inspection and the corresponding inspection necessity parameter as a piece of training data;

[0017] The training data corresponding to the instruction records of the first edge model and the second edge model are respectively constructed as iterative training data sets.

[0018] In some embodiments of the present invention, in the iterative update stage, the central node obtains the inspection results of manual detection, and stores the inspection results of each monitoring position in a preset time period as an iterative training data set, which also includes:

[0019] Respectively obtain instruction records of the first edge model and the second edge model in determining manual inspection;

[0020] Determine whether, when either the first edge model or the second edge model determines to perform manual detection, the other first edge model or the second edge model synchronously determines to perform manual detection;

[0021] If not, then for the first edge model or the second edge model that has not been determined to perform manual inspection, obtain the inspection judgment parameters output at that moment, and obtain the instruction record corresponding to the other first edge model or the second edge model that determines to perform manual inspection at that moment, and the inspection necessity parameters uploaded by the staff to construct a training data.

[0022] The above scheme is adopted. In this scheme, two judgment models are set in each edge node. On the one hand, it can increase the judgment accuracy and avoid missed detection. On the other hand, if only one judgment model is set in the edge node, then during the training process, training data can only be generated when the model judges to perform manual inspection. Part of the training data generated is the case where the model judges to perform manual inspection but the inspection necessity parameters uploaded by the staff are low. In the process of continuous iterative training, the model will continue to deviate in the direction of not judging to perform manual inspection, and eventually lead to the loss of judgment ability. However, in this scheme, two judgment models are set at the same time. When one judgment model does not judge to perform manual inspection, the other judgment model may judge to perform manual inspection. In this case, training data in the opposite direction of the deviation is generated for the judgment model to ensure that the judgment is performed and avoid the occurrence of unidirectional deviation.

[0023] In some embodiments of the present invention, in the step of the central node iteratively training the first central model and the second central model based on the iterative training data set:

[0024] Training the first central model using an iterative training data set corresponding to the first marginal model;

[0025] The second central model is trained using the iterative training data set corresponding to the second marginal model.

[0026] In some embodiments of the present invention, the steps of the iterative update phase further include:

[0027] A real-time storage unit is provided corresponding to each edge node, and the real-time storage unit stores model parameters of the first edge model or the second edge model in the edge node;

[0028] When iterative training is performed on an edge node, the model parameters in the corresponding real-time storage unit are called and applied to the first central model and the second central model, and training is performed using the collected iterative training data set.

[0029] By adopting the above scheme, a real-time storage unit is separately provided for each edge node, which can ensure the separate training of each edge node, ensure that the training data can be applied to the monitoring position corresponding to the edge node, and further ensure that each edge node can play the best judgment effect.

[0030] In some embodiments of the present invention, the steps of the iterative update phase further include:

[0031] At the beginning of each iterative update phase, the instruction records of the first edge model and the second edge model of each edge node when determining to perform manual inspection are called, and the inspection determination parameters output by the first edge model and the second edge model when determining to perform manual inspection are compared with the inspection necessity parameters uploaded by the staff;

[0032] Determine whether the inspection determination parameter recorded in each instruction matches the corresponding inspection necessity parameter, and obtain the matching degree of each edge node;

[0033] The edge nodes are sorted based on the matching degree, and each edge node is iteratively trained based on the sorting order.

[0034] In some embodiments of the present invention, in the step of determining whether the inspection determination parameter recorded in each instruction matches the corresponding inspection necessity parameter and obtaining the matching degree of each edge node:

[0035] Calculating the difference between the inspection determination parameter and the corresponding inspection necessity parameter, and determining whether there is a match based on the difference;

[0036] The ratio of the number of unmatched instruction records to the total number of instruction records is calculated as the matching degree of the edge node.

[0037] In some embodiments of the present invention, in the step of sorting the edge nodes based on the matching degree and iteratively training each edge node based on the sorting order, the edge nodes are arranged in descending order based on the matching degree, and the edge nodes arranged in descending order are iteratively trained one by one.

[0038] By adopting the above scheme, during the processing of the iterative update phase, the edge nodes are sorted by the matching degree. The matching degree is determined by the difference between the inspection judgment parameter and the corresponding inspection necessity parameter, which can represent the difference between the output result of the model and the actual situation. By sorting the edge nodes, the scheme prioritizes the edge nodes whose model judgment situation differs greatly from the actual situation, so as to ensure the accuracy of subsequent judgments.

[0039] In some embodiments of the present invention, in the step of the edge model re-determining in real time whether to perform manual inspection based on the updated first edge model and the second edge model, the first edge model and the second edge model perform a determination at every other time point:

[0040] The first edge model constructs a first vector using the real-time collected environmental parameters as input, and outputs an inspection determination parameter, compares the inspection determination parameter with a preset determination threshold, and determines whether to perform manual inspection;

[0041] The second edge model constructs a second vector using the real-time collected equipment operation parameters as input, and outputs an inspection determination parameter, and compares the inspection determination parameter with a preset determination threshold to determine whether to perform manual inspection.

[0042] On the other hand, the present invention also relates to a smart power plant management system based on distributed management, the system includes a computer device, the computer device includes a processor and a memory, the memory stores computer instructions, the processor is used to execute the computer instructions stored in the memory, when the computer instructions are executed by the processor, the system implements the steps implemented by the method.

[0043] In summary, the present invention has the following beneficial effects:

[0044] 1. The central node of this solution performs unified training in the initial stage, and uniformly sends the trained model parameters to each edge node. Each edge node builds an initial model based on the same model parameters. In the judgment process, since the two models are trained based on different parameters, the two models may produce different results at the same time. The staff conducts manual inspection based on the manual inspection instructions issued by any model at the same monitoring location, and gives the numerical value of the necessity of inspection after manual inspection and uploads it to the central node. Therefore, the two models can correspond to each other and build a data set that is different from their own judgment results, providing a more accurate basis for subsequent iterative updates. In addition, this solution can train each edge node separately to ensure processing accuracy.

[0045] 2. This solution sets two judgment models at each edge node. On the one hand, it can increase the judgment accuracy and avoid missed detection. On the other hand, if only one judgment model is set at the edge node, during the training process, training data can only be generated when the model determines to perform manual inspection. Part of the training data generated is the case where the model determines to perform manual inspection but the inspection necessity parameters uploaded by the staff are low. In the process of continuous iterative training, the model will continue to deviate in the direction of not determining to perform manual inspection, and eventually lose the judgment ability. However, this solution sets two judgment models at the same time. When one judgment model does not determine to perform manual inspection, the other judgment model may determine to perform manual inspection. Then, training data in the opposite direction of the deviation is generated for the judgment model to ensure that the judgment is performed and avoid the occurrence of unidirectional deviation.

[0046] 3. This solution sets a real-time storage unit for each edge node, which can ensure the individual training of each edge node and ensure that the training data can be applied to the monitoring position corresponding to the edge node, thereby ensuring that each edge node can play the best judgment effect;

[0047] 4. During the iterative update phase, this solution sorts the edge nodes by the degree of matching. The degree of matching is determined by the difference between the inspection judgment parameter and the corresponding inspection necessity parameter, which can represent the difference between the output result of the model and the actual situation. This solution sorts the edge nodes to prioritize the edge nodes where the difference between the model judgment and the actual situation is large, so as to ensure the accuracy of subsequent judgments. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0049] Figure 1 It is a schematic diagram of a first implementation mode of a smart power plant management method based on distributed management of the present invention;

[0050] Figure 2 A schematic diagram of a construction method of an iterative training data set in the iterative update phase of the smart power plant management method based on distributed management of the present invention;

[0051] Figure 3 A schematic diagram of another method for constructing an iterative training data set in the iterative update phase of the smart power plant management method based on distributed management of the present invention. DETAILED DESCRIPTION

[0052] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. Instead, they are merely examples of systems and methods consistent with some aspects of the present invention as detailed in the appended claims.

[0053] The terms used in the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. The singular forms "a", "the" and "the" used in the present invention and the appended claims are also intended to include plural forms unless the context clearly indicates other meanings. It should also be understood that the term "and / or" used herein refers to and includes any or all possible combinations of one or more associated listed items.

[0054] like Figure 1 As shown, the present invention provides a smart power plant management method based on distributed management. The method is based on a distributed architecture, including a central node and edge nodes corresponding to each monitoring location. The central node is provided with a first central model based on environmental parameter determination and a second central model based on equipment operation parameter determination. The steps of the method include:

[0055] In the specific implementation process, the model structures of the first central model and the second central model of the central node and the first edge model and the second edge model of each edge node are the same and can all be convolutional neural network models.

[0056] This solution is applied to actual power plants. Multiple monitoring locations are set up in the power plants, and each monitoring location is equipped with a sensor to collect environmental parameters and equipment operating parameters.

[0057] Step S100, an initial allocation stage, in which the central node performs initial training on the first central model and the second central model based on the initial training data set, and sends the model parameters of the first central model and the second central model after the initial training to each edge node, so that the edge node constructs the first edge model and the second edge model;

[0058] In the specific implementation process, in the initial allocation stage, the first central model and the second central model are trained respectively by using the environmental parameter training data set and the equipment operation parameter training data set that are pre-collected from each monitoring location. During the training process, the mean square error, cross entropy or Smooth L1 loss function can be used to calculate the loss to train the first central model and the second central model.

[0059] In some embodiments of the present invention, if the mean square error is used to calculate the loss, it is calculated based on the following formula:

[0060]

[0061] Among them, MSE represents the value of mean square error, N is the number of samples, and y i represents the true value of the i-th sample, Represents the predicted value of the i-th sample.

[0062] The model parameters sent to each edge node in the initial allocation phase are the same, the model parameters of the first edge model constructed by each edge node are the same, and the model parameters of the second edge model constructed by each edge node are also the same.

[0063] Step S200, the edge node constructs a first vector and a second vector based on the environmental parameters and equipment operation parameters collected in real time, and determines whether to perform manual inspection based on the corresponding first edge model and second edge model;

[0064] In a specific implementation process, the environmental parameters and equipment operation parameters collected in real time are encoded to obtain a first vector and a second vector. The first edge model and the second edge model take the first vector and the second vector as input respectively and output inspection determination parameters.

[0065] In the specific implementation process, the environmental parameters include environmental temperature, humidity, sound intensity, sound frequency and harmful gas concentration; the equipment operation parameters include equipment operation temperature and equipment vibration frequency;

[0066] In a specific implementation process, the environmental parameters and the equipment operation parameters are encoded into a preset length using a preset encoding method to obtain the first vector and the second vector.

[0067] Step S300, an iterative update phase, in which the central node obtains the inspection results of manual detection, and stores the inspection results of each monitoring position in a preset time period as an iterative training data set. Based on the iterative training data set, the central node iteratively trains the first central model and the second central model based on the iterative training data set, and sends the model parameters of the iteratively trained first central model and the second central model to each edge node, so that the edge node updates the first edge model and the second edge model;

[0068] In the specific implementation process, in the iterative update phase, when iterative training is performed on any edge node, the model parameters of the central node are updated to the current model parameters of the edge node, and then training is performed to ensure that the training corresponds to the edge node.

[0069] Step S400: the edge model re-determines in real time whether to perform manual inspection based on the updated first edge model and the second edge model, and continuously performs iterative updates.

[0070] In the specific implementation process, the processing frequency of the iterative update can be preset, and can also be enabled at any time based on the activation instruction of the staff.

[0071] Adopting the above scheme, this scheme, on the one hand, sets two judgment models in one edge node to ensure the safety of the corresponding monitoring location in the power plant;

[0072] In addition, the central node of this solution performs unified training in the initial stage, and uniformly sends the trained model parameters to each edge node. Each edge node builds an initial model based on the same model parameters. In the judgment process, since the two models are trained based on different parameters, the two models may obtain different results at the same time. The staff conducts manual inspection based on the manual inspection instructions issued by any model at the same monitoring location, and gives the numerical value of the necessity of inspection after manual inspection and uploads it to the central node. Therefore, the two models can correspond to each other and construct a data set that is different from their own judgment results, providing a more accurate basis for subsequent iterative updates. In addition, this solution can train each edge node separately to ensure processing accuracy.

[0073] like Figure 2 As shown, in some embodiments of the present invention, in the iterative update stage, the central node obtains the inspection results of manual detection, and stores the inspection results of each monitoring location in a preset time period as an iterative training data set:

[0074] Step S310, respectively obtaining instruction records of the first edge model and the second edge model in determining manual inspection;

[0075] During the specific implementation process, after completing a manual inspection, the staff will upload the inspection data and the inspection necessity parameters for the manual inspection evaluation to the storage center of the central node, and the central node can call the data in the storage center.

[0076] Step S321, obtaining inspection necessity parameters uploaded by the staff corresponding to each instruction record;

[0077] Step S331, constructing the inspection determination parameter output by the first edge model or the second edge model when determining to perform manual inspection and the corresponding inspection necessity parameter into a piece of training data;

[0078] Step S341, and construct the training data of the instruction records corresponding to the first edge model and the second edge model into iterative training data sets respectively.

[0079] like Figure 3 As shown, in some embodiments of the present invention, in the iterative update stage, the central node obtains the inspection results of manual detection, and stores the inspection results of each monitoring position in a preset time period as an iterative training data set, and the step also includes:

[0080] Respectively obtain instruction records of the first edge model and the second edge model in determining manual inspection;

[0081] Step S322, determining whether, when either the first edge model or the second edge model determines to perform manual detection, the other first edge model or the second edge model synchronously determines to perform manual detection;

[0082] Step S332, if not, then for the first edge model or the second edge model that has not been determined to perform manual inspection, obtain the inspection judgment parameters output at that moment, and obtain the instruction record corresponding to another first edge model or the second edge model at that moment that determines to perform manual inspection, and the inspection necessity parameters uploaded by the staff to construct a training data.

[0083] The above scheme is adopted. In this scheme, two judgment models are set in each edge node. On the one hand, it can increase the judgment accuracy and avoid missed detection. On the other hand, if only one judgment model is set in the edge node, then during the training process, training data can only be generated when the model judges to perform manual inspection. Part of the training data generated is the case where the model judges to perform manual inspection but the inspection necessity parameters uploaded by the staff are low. In the process of continuous iterative training, the model will continue to deviate in the direction of not judging to perform manual inspection, and eventually lead to the loss of judgment ability. However, in this scheme, two judgment models are set at the same time. When one judgment model does not judge to perform manual inspection, the other judgment model may judge to perform manual inspection. In this case, training data in the opposite direction of the deviation is generated for the judgment model to ensure that the judgment is performed and avoid the occurrence of unidirectional deviation.

[0084] In some embodiments of the present invention, in the step of the central node iteratively training the first central model and the second central model based on the iterative training data set:

[0085] Training the first central model using an iterative training data set corresponding to the first marginal model;

[0086] The second central model is trained using the iterative training data set corresponding to the second marginal model.

[0087] In the specific implementation process, the loss function used in the iterative update phase is the same as the loss function used in the initial allocation phase.

[0088] In some embodiments of the present invention, the steps of the iterative update phase further include:

[0089] A real-time storage unit is provided corresponding to each edge node, and the real-time storage unit stores model parameters of the first edge model or the second edge model in the edge node;

[0090] When iterative training is performed on an edge node, the model parameters in the corresponding real-time storage unit are called and applied to the first central model and the second central model, and training is performed using the collected iterative training data set.

[0091] In a specific implementation process, the model parameters of the first edge model and the second edge model of each edge node may be constructed as parameter matrices respectively and stored in corresponding real-time storage units.

[0092] By adopting the above scheme, a real-time storage unit is separately provided for each edge node, which can ensure the separate training of each edge node, ensure that the training data can be applied to the monitoring position corresponding to the edge node, and further ensure that each edge node can play the best judgment effect.

[0093] In some embodiments of the present invention, the steps of the iterative update phase further include:

[0094] At the beginning of each iterative update phase, the instruction records of the first edge model and the second edge model of each edge node when determining to perform manual inspection are called, and the inspection determination parameters output by the first edge model and the second edge model when determining to perform manual inspection are compared with the inspection necessity parameters uploaded by the staff;

[0095] Determine whether the inspection determination parameter recorded in each instruction matches the corresponding inspection necessity parameter, and obtain the matching degree of each edge node;

[0096] The edge nodes are sorted based on the matching degree, and each edge node is iteratively trained based on the sorting order.

[0097] In some embodiments of the present invention, in the step of determining whether the inspection determination parameter recorded in each instruction matches the corresponding inspection necessity parameter and obtaining the matching degree of each edge node:

[0098] Calculating the difference between the inspection determination parameter and the corresponding inspection necessity parameter, and determining whether there is a match based on the difference;

[0099] The ratio of the number of unmatched instruction records to the total number of instruction records is calculated as the matching degree of the edge node.

[0100] In the specific implementation process, in the step of calculating the difference between the inspection judgment parameter and the corresponding inspection necessity parameter and determining whether there is a match based on the difference, the absolute value of the difference is compared with a preset difference threshold. If it is greater than the difference threshold, it is determined to be mismatched.

[0101] In some embodiments of the present invention, in the step of sorting the edge nodes based on the matching degree and iteratively training each edge node based on the sorting order, the edge nodes are arranged in descending order based on the matching degree, and the edge nodes arranged in descending order are iteratively trained one by one.

[0102] By adopting the above scheme, during the processing of the iterative update phase, the edge nodes are sorted by the matching degree. The matching degree is determined by the difference between the inspection judgment parameter and the corresponding inspection necessity parameter, which can represent the difference between the output result of the model and the actual situation. By sorting the edge nodes, the scheme prioritizes the edge nodes whose model judgment situation differs greatly from the actual situation, so as to ensure the accuracy of subsequent judgments.

[0103] In some embodiments of the present invention, in the step of the edge model re-determining in real time whether to perform manual inspection based on the updated first edge model and the second edge model, the first edge model and the second edge model perform a determination at every other time point:

[0104] The first edge model constructs a first vector using the real-time collected environmental parameters as input, and outputs an inspection determination parameter, compares the inspection determination parameter with a preset determination threshold, and determines whether to perform manual inspection;

[0105] The second edge model constructs a second vector using the real-time collected equipment operation parameters as input, and outputs an inspection determination parameter, and compares the inspection determination parameter with a preset determination threshold to determine whether to perform manual inspection.

[0106] In some embodiments of the present invention, the method further comprises:

[0107] After completing the preset number of iterative updates, the model parameters of any two edge nodes are constructed into a determination group, and the difference between the model parameters of the two edge nodes in the determination group is calculated;

[0108] Determining whether two edge nodes are related edge nodes based on the difference;

[0109] If the two edge nodes are related edge nodes, when the monitoring position corresponding to one of the edge nodes is determined to be manually inspected, the monitoring position corresponding to the other related edge node is simultaneously released to the staff.

[0110] In a specific implementation process, the method of calculating the difference between the model parameters of two edge nodes in the determination group may adopt the cosine distance.

[0111] With the above scheme, after completing the preset number of iterative updates, in theory, the differences in model parameters of each edge node will become larger and larger. However, if the differences in model parameters of two edge nodes are small, it means that there is a certain correlation between the two. Therefore, in this scheme, when the monitoring position corresponding to one of the edge nodes is determined to be manually inspected, the monitoring position corresponding to the other related edge node will be synchronously released to the staff, so that the staff can conduct synchronous inspections based on the correlation to further ensure the safety of the power plant.

[0112] On the other hand, the present invention also relates to a smart power plant management system based on distributed management, the system includes a computer device, the computer device includes a processor and a memory, the memory stores computer instructions, the processor is used to execute the computer instructions stored in the memory, when the computer instructions are executed by the processor, the system implements the steps implemented by the method.

[0113] The embodiment of the present invention also provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, the aforementioned smart power plant management method based on distributed management is implemented. The computer-readable storage medium can be a tangible storage medium, such as a random access memory (RAM), a memory, a read-only memory (ROM), an electrically programmable ROM, an electrically erasable programmable ROM, a register, a floppy disk, a hard disk, a removable storage disk, a CD-ROM, or any other form of storage medium known in the technical field.

[0114] It should be understood by those skilled in the art that the exemplary components, systems and methods described in conjunction with the embodiments disclosed herein can be implemented in hardware, software or a combination of the two. Whether it is performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention. When implemented in hardware, it can be, for example, an electronic circuit, an application specific integrated circuit (ASIC), appropriate firmware, a plug-in, a function card, etc. When implemented in software, the elements of the present invention are programs or code segments used to perform the required tasks. The program or code segment can be stored in a machine-readable medium, or transmitted on a transmission medium or a communication link via a data signal carried in a carrier.

[0115] It should be clear that the present invention is not limited to the specific configuration and processing described above and shown in the figures. For the sake of simplicity, a detailed description of the known method is omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of the present invention is not limited to the specific steps described and shown, and those skilled in the art can make various changes, modifications and additions, or change the order between the steps after understanding the spirit of the present invention.

[0116] In the present invention, features described and / or illustrated for one embodiment may be used in the same or similar manner in one or more other embodiments, and / or combined with features of other embodiments or replace features of other embodiments.

[0117] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the embodiments of the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A smart power plant management method based on distributed management, characterized in that: The method is based on a distributed architecture, including a central node and edge nodes corresponding to each monitoring location. The central node is provided with a first central model based on environmental parameter determination and a second central model based on equipment operation parameter determination. The steps of the method include: An initial allocation stage, in which the central node performs initial training on the first central model and the second central model based on the initial training data set, and sends the model parameters of the first central model and the second central model after the initial training to each edge node, so that the edge node constructs the first edge model and the second edge model; The edge node constructs a first vector and a second vector based on the environmental parameters and equipment operation parameters collected in real time, and determines whether to perform manual inspection based on the corresponding first edge model and second edge model; An iterative update phase, in which the central node obtains the inspection results of manual detection, and stores the inspection results of each monitoring position in a preset time period as an iterative training data set. Based on the iterative training data set, the central node iteratively trains the first central model and the second central model based on the iterative training data set, and sends the model parameters of the iteratively trained first central model and the second central model to each edge node, so that the edge node updates the first edge model and the second edge model; The edge model re-determines in real time whether to perform manual inspection based on the updated first edge model and the second edge model, and continuously performs iterative updates.

2. The distributed management-based smart power plant management method according to claim 1 is characterized in that: In the iterative update stage, the central node obtains the inspection results of manual detection, and stores the inspection results of each monitoring location in a preset time period as an iterative training data set: Respectively obtain instruction records of the first edge model and the second edge model in determining manual inspection; Obtain inspection necessity parameters uploaded by the staff corresponding to each instruction record; constructing the inspection determination parameter output by the first edge model or the second edge model when determining to perform manual inspection and the corresponding inspection necessity parameter as a piece of training data; The training data corresponding to the instruction records of the first edge model and the second edge model are respectively constructed as iterative training data sets.

3. The distributed management-based smart power plant management method according to claim 2 is characterized in that: In the iterative update stage, the central node obtains the inspection results of manual detection, and stores the inspection results of each monitoring position in a preset time period as an iterative training data set, and the step also includes: Respectively obtain instruction records of the first edge model and the second edge model in determining manual inspection; Determine whether, when either the first edge model or the second edge model determines to perform manual detection, the other first edge model or the second edge model synchronously determines to perform manual detection; If not, then for the first edge model or the second edge model that has not been determined to perform manual inspection, obtain the inspection judgment parameters output at that moment, and obtain the instruction record corresponding to the other first edge model or the second edge model that determines to perform manual inspection at that moment, and the inspection necessity parameters uploaded by the staff to construct a training data.

4. The distributed management-based smart power plant management method according to claim 3 is characterized in that: In the step of the central node iteratively training the first central model and the second central model based on the iterative training data set: Training the first central model using an iterative training data set corresponding to the first marginal model; The second central model is trained using the iterative training data set corresponding to the second marginal model.

5. The distributed management-based smart power plant management method according to claim 1 is characterized in that: The steps of the iterative update phase also include: A real-time storage unit is provided corresponding to each edge node, and the real-time storage unit stores model parameters of the first edge model or the second edge model in the edge node; When iterative training is performed on an edge node, the model parameters in the corresponding real-time storage unit are called and applied to the first central model and the second central model, and training is performed using the collected iterative training data set.

6. The distributed management-based smart power plant management method according to claim 1 is characterized in that: The steps of the iterative update phase also include: At the beginning of each iterative update phase, the instruction records of the first edge model and the second edge model of each edge node when determining to perform manual inspection are called, and the inspection determination parameters output by the first edge model and the second edge model when determining to perform manual inspection are compared with the inspection necessity parameters uploaded by the staff; Determine whether the inspection determination parameter recorded in each instruction matches the corresponding inspection necessity parameter, and obtain the matching degree of each edge node; sorting the edge nodes based on the matching degree; Iterative training is performed on each edge node based on the sorted order.

7. The distributed management-based smart power plant management method according to claim 6 is characterized in that: In the step of determining whether the inspection determination parameter of each instruction record matches the corresponding inspection necessity parameter and obtaining the matching degree of each edge node: Calculating the difference between the inspection determination parameter and the corresponding inspection necessity parameter, and determining whether there is a match based on the difference; The ratio of the number of unmatched instruction records to the total number of instruction records is calculated as the matching degree of the edge node.

8. The distributed management-based smart power plant management method according to claim 6 is characterized in that: In the step of sorting the edge nodes based on the matching degree and iteratively training each edge node based on the sorting order, the edge nodes are sorted in descending order based on the matching degree, and the edge nodes sorted in descending order are iteratively trained one by one.

9. The smart power plant management method based on distributed management according to any one of claims 1 to 8, characterized in that: The method further comprises the steps of: After completing the preset number of iterative updates, the model parameters of any two edge nodes are constructed into a determination group, and the difference between the model parameters of the two edge nodes in the determination group is calculated; Determining whether two edge nodes are related edge nodes based on the difference; If the two edge nodes are related edge nodes, when the monitoring position corresponding to one of the edge nodes is determined to be manually inspected, the monitoring position corresponding to the other related edge node is simultaneously released to the staff.

10. A smart power plant management system based on distributed management, characterized by: The system includes a computer device, which includes a processor and a memory, wherein the memory stores computer instructions, and the processor is used to execute the computer instructions stored in the memory. When the computer instructions are executed by the processor, the system implements the steps implemented by the method described in any one of claims 1 to 9.

Images