Water supply system data filling method and water supply network control system
The generative adversarial network model fills the missing data in the nuclear power plant water supply system, which solves the data loss problem and improves the accuracy of data and the security of operation control.
Patent Information
- Application Number
- CN202411571947.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-06
- Publication Date
- 2025-03-18
AI Technical Summary
Nuclear power plant water supply system may lead to data loss or damage during operation, leading to potential risks of operating control systems. The existing data filling methods are insufficient when processing large-scale complex data and fail to meet the needs of nuclear power plant operation control.
Generative adversarial network model is adopted, including generative network, discriminative network and auxiliary prediction network, and target fill data is generated through training models to control the operating status of the water supply system in real time.
The missing data filling of the nuclear power plant water supply system has been realized, the accuracy of the target filling data has been improved, and the safety of the nuclear power plant operation control system has been enhanced.
Smart Images
Figure CN119646411A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the technical field of operation control of nuclear power plants, and more specifically, to a method for filling water supply system data, a water supply network control system, a computer device, a computer-readable storage medium, and a computer program product. Background Art
[0002] The safety of nuclear power plant operation control system is of vital importance. In the operation control system of nuclear power plant, the water supply system is an important component, which provides cooling water and boiler feed water for the generator set to ensure the stable operation of the nuclear power plant. However, in the actual operation process, the data of the water supply system of the nuclear power plant may be lost or damaged due to various reasons, which will bring potential risks to the operation control system of the nuclear power plant. Therefore, it is necessary to provide an effective data filling method to deal with these situations.
[0003] To address the problem of missing data, currently common data filling methods include mean interpolation, regression analysis, k-nearest neighbor interpolation, etc. These methods are usually based on historical data or local information, estimate missing values in a certain way, and improve data quality to a certain extent.
[0004] However, the common data filling methods currently have shortcomings when processing large-scale complex data. At the same time, the common data filling methods currently mainly start from the perspective of signal analysis and do not consider the operation control requirements of nuclear power plants. Summary of the invention
[0005] In order to solve the problem of data loss in the water supply system of a nuclear power plant, the present invention provides a water supply system data filling method, a water supply network control system, a computer device, a computer-readable storage medium and a computer program product, which will realize the filling of missing data in the water supply system of a nuclear power plant and improve the accuracy of the target filled data.
[0006] To achieve the above object, according to a first aspect of the present invention, a method for filling water supply system data is provided, the method comprising:
[0007] Obtain historical operating status data of the water supply system;
[0008] A generative adversarial network model for missing data filling is established according to a generative network, a discriminative network and an auxiliary prediction network. The input of the generative network includes attribute missing data and a missing position vector, and the output is the filling data. The missing position vector includes missing information of the attribute missing data. The input of the discriminative network includes the filling data and the complete data, and the output is the probability that the filling data is the complete data. The input of the auxiliary prediction network includes the filling data and the complete data, and the output is the category of the filling data.
[0009] Based on historical operating status data, train a generative adversarial network model for missing data filling;
[0010] According to the generative adversarial network model with trained missing data filling, target filling data is generated, and the target filling data is used to control the operating status of the water supply system in real time.
[0011] Furthermore, historical operating status data of the water supply system is obtained, including obtaining historical operating status data of the water supply system, the historical operating status data including water supply flow, steam flow, steam pressure and corresponding collection time; the operating status data is dimensionlessly standardized to obtain a time series data set.
[0012] Furthermore, the missing position vector includes the category of the attribute missing data and the collection time corresponding to the missing part. According to the historical operation status data, a generative adversarial network model for filling in missing data is trained, including obtaining the attribute missing data in the time series data set and the missing position vector and complete data corresponding to the attribute missing data; inputting the attribute missing data and the corresponding missing position vector into the generative network to obtain the filled data; inputting the filled data and the complete data into the discriminant network to obtain the probability that the filled data is complete data; inputting the filled data and the complete data into the auxiliary prediction network to obtain the category of the filled data; and training the generative network, the discriminant network and the auxiliary prediction network in sequence in the direction of minimizing the difference between the probability that the filled data is complete data and 1, and in the direction of minimizing the difference between the category of the filled data and the category of the missing position vector until the training stop condition is reached.
[0013] Furthermore, the generative network, the discriminative network and the auxiliary prediction network are trained in sequence until a training stop condition is reached, including adversarial training of the discriminative network and the generative network, so that the filled data has the same distribution as the complete data.
[0014] Furthermore, the generation network, judgment network and auxiliary prediction network all adopt multi-layer fully connected neural networks.
[0015] Furthermore, the missing position vector is a discrete variable or a continuous variable.
[0016] According to a second aspect of the present invention, there is also provided a water supply network control system, which implements any of the above methods, including:
[0017] A controller, used for acquiring target filling data and generating control instructions according to the target filling data;
[0018] The controlled object is used to perform water supply operations according to control instructions.
[0019] According to a third aspect of the present invention, there is also provided a computer device, which includes a memory, a processor and a computer program stored in the memory, and the processor executes the computer program to implement the steps of any one of the above methods.
[0020] According to a fourth aspect of the present invention, there is also provided a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, the steps of any of the above methods are implemented.
[0021] According to a fifth aspect of the present invention, there is also provided a computer program product, comprising a computer program, which implements the steps of any of the above methods when executed by a processor.
[0022] In general, the above technical solutions conceived by the present invention can achieve the following beneficial effects compared with the prior art:
[0023] (1) The present invention provides a method for filling data in a water supply system. First, the historical operating status data of the water supply system is obtained, and a generative adversarial network model for filling missing data is established based on a generation network, a discriminant network and an auxiliary prediction network. Then, the generative adversarial network model for filling missing data is trained based on the historical operating status data. Finally, target filling data is generated based on the trained generative adversarial network model for filling missing data. The target filling data is used to control the operating status of the water supply system in real time, and can realize the filling of missing data of the water supply system of a nuclear power plant, thereby achieving the purpose of improving the accuracy of the target filling data.
[0024] (2) The water supply network control system provided by the present invention has more accurate control instructions because the control instructions are generated based on the target filling data generated by the water supply system data filling method; thus, the controlled object is controlled according to the control instructions, and the control effect is better, thereby improving the safety of the entire nuclear power plant operation control system. BRIEF DESCRIPTION OF THE DRAWINGS
[0025] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0026] Figure 1 A flow chart of a method for filling water supply system data provided in an embodiment of the present application;
[0027] Figure 2 An application environment diagram of a water supply system data filling method provided in an embodiment of the present application;
[0028] Figure 3 A flow chart of the training process of a generative adversarial network model for missing data filling provided in an embodiment of the present application;
[0029] Figure 4 This is a schematic diagram of the structure of the existing conditional generative adversarial network;
[0030] Figure 5 A schematic diagram of the structure of a generation network provided in an embodiment of the present application;
[0031] Figure 6 A schematic diagram of the structure of the discrimination network provided in the embodiment of the present application;
[0032] Figure 7 A schematic diagram of the structure of the auxiliary prediction network provided in the embodiment of the present application;
[0033] Figure 8 A working principle diagram of a water supply network control system provided in an embodiment of the present application;
[0034] Fig. 9 A schematic diagram of the internal structure of a computer device provided in an embodiment of the present application. DETAILED DESCRIPTION
[0035] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.
[0036] The terms "first", "second", "third", etc. in the specification and claims of this application and the above-mentioned drawings are used to distinguish different objects, rather than to describe a specific order. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not limited to the listed steps or units, but optionally includes steps or units that are not listed, or optionally includes other steps or units inherent to these processes, methods, products or devices.
[0037] like Figure 1As shown, a method for filling water supply system data is provided, which can be executed by a terminal or by a server communicating with the terminal through a network. The terminal can be, but is not limited to, various personal computers, laptops, smart phones, tablet computers, etc. The server can be an independent server, or a server cluster composed of multiple servers can be used. Taking the method applied to a terminal as an example, the method includes the following steps:
[0038] Step 101, obtaining historical operating status data of the water supply system.
[0039] The historical operation status data include feed water flow, steam flow, steam pressure, and the corresponding collection time of feed water flow, steam flow, and steam pressure.
[0040] Exemplarily, the terminal obtains historical operating status data of the water supply system, performs dimensionless normalization processing on the operating status data, and obtains a time series data set.
[0041] Step 102, based on the generation network, the discriminant network and the auxiliary prediction network, a generative adversarial network model for filling missing data is established, the input end of the generation network includes attribute missing data and missing position vector, the output end is the filling data, and the missing position vector includes missing information of the attribute missing data; the input end of the discriminant network includes the filling data and complete data, and the output end is the probability that the filling data is complete data; the input end of the auxiliary prediction network includes the filling data and complete data, and the output end is the category of the filling data.
[0042] The complete data is the attribute non-missing data corresponding to the attribute missing data. The missing position vector includes the missing information of the attribute missing data, and the missing information includes the category of the attribute missing data and the collection time corresponding to the missing part.
[0043] In one embodiment, the generation network, the judgment network and the auxiliary prediction network all use multi-layer fully connected neural networks. Preferably, LSTM (Long Short-Term Memory) is used.
[0044] Step 103: training a generative adversarial network model for missing data filling based on historical operating status data.
[0045] Step 104, generating target filling data according to the trained generative adversarial network model for missing data filling, and the target filling data is used to control the operating status of the water supply system in real time.
[0046] The target filling data is the filling data at the next moment of the current moment. The filling data at the next moment of the current moment is predicted by the generative adversarial network model with trained missing data filling, so as to avoid the influence of reduced control efficiency of the water supply system due to the missing operation status data at the next moment.
[0047] The above-mentioned water supply system data filling method first obtains the historical operating status data of the water supply system, and establishes a generative adversarial network model for missing data filling based on the generation network, the discriminant network and the auxiliary prediction network, and then trains the generative adversarial network model for missing data filling based on the historical operating status data; finally, based on the trained generative adversarial network model for missing data filling, generates target filling data, and the target filling data is used to control the operating status of the water supply system in real time, so as to realize the filling of missing data of the water supply system of the nuclear power plant.
[0048] Figure 2 : is an application environment diagram of a water supply system data filling method provided by an embodiment of the present application. Figure 2 As shown, the water supply system includes a terminal and a controlled object (such as a water pump, a water tank, and a boiler regulating valve). The terminal controls multiple water supply nodes by controlling a network bus, and multiple water supply nodes (water supply nodes 1-4) can control the controlled object.
[0049] In one embodiment, a method for filling water supply system data includes the following steps:
[0050] Step 1: Obtain the operating status data monitored by the sensor network in the nuclear power plant feed water system during a historical period. The operating status data includes feed water flow, steam flow and steam pressure and the corresponding collection time.
[0051] Step 2: Perform dimensionless processing and standardization on the operating status data to obtain a time series data set.
[0052] Step 3, such as Figure 3As shown in the figure, a generative adversarial network model (MIGAN for short) for missing data filling is created with reference to the conditional generative adversarial network (full name: Conditional Generative Adversarial Nets, CGAN for short). The MIGAN includes three sub-networks: a generative network (also called a generator, G for short), a discriminator network (also called a discriminator, D for short) and an auxiliary prediction network (P for short). These three networks are trained in sequence during the network training process. The alternating training of the generative network and the discriminator network ensures that the distribution of the filled data is consistent with that of the complete data, and the gradient returned by the auxiliary prediction network also participates in the re-update of the parameters of the generative network, so that the filled data has a better prediction effect in the prediction problem.
[0053] Preferably, the generation network, the judgment network and the auxiliary prediction network are all composed of a multi-layer fully connected neural network, such as LSTM (Long Short-Term Memory).
[0054] Figure 4 This is the structure diagram of the existing conditional generative adversarial network. Figure 4 As shown in the figure, the conditional generative adversarial network is a conditional version of the generative adversarial network (full name: Generative adversarial networks, referred to as GAN). Unlike GAN, the input of the generator and the judge includes conditional data. For the generator, its input has two parts: one is the conditional data, and the other is random noise. The conditional data can be a category label, and the random noise can be a noise distribution sample. By modifying the corresponding data, a variable of a specified category can be generated.
[0055] For the generative adversarial network model MIGAN for missing data filling provided in this embodiment, since the input data is a vector in a time series data set (with time attributes), the input of the generator is a conditional vector, that is, a category vector. By taking the missing information of the missing data (that is, the missing position vector) as the conditional vector, the generator outputs the filling data, and by modifying the corresponding category vector, the network generates the filling data of the corresponding category. However, unlike the conditional generative adversarial network CGAN, the conditional vector is not added to the input of the discriminator.
[0056] There are three types of category vectors: feedwater flow vector, steam flow vector, and steam pressure vector. By modifying the corresponding category vector, you can generate a variable of a specified category.
[0057] Figure 5Structural diagram of the generating network in this embodiment. In the training process of the generative adversarial network model MIGAN for missing data filling, the gradient sources for updating the parameters of the generating network have two parts. One is the judgment of the discriminator on the sampling of the generating network's generating results. Minimization means that the filling data (i.e., the generating results of the generating network) is mistaken as real data by the discriminator as much as possible; the second is the evaluation of the filling data by the auxiliary prediction network. Minimization means that the generating results of the generating network can be predicted as the correct category by the auxiliary prediction network as much as possible. For the filling data of missing data, it is usually necessary to have the same distribution as the complete samples in terms of data distribution. Here, this requirement is achieved through adversarial training between the discriminator and the generator.
[0058] Figure 6 It is a schematic diagram of the structure of the discriminant network. The discriminant network D is used to distinguish the filled data generated by the generative network from the complete data without missing values. It performs a binary classification on the data in the filled data set G and the complete data set E to distinguish the source of the data. The output is the probability that the input comes from the complete data.
[0059] In contrast to the training of the generative network, which is to make the discriminative network believe that the padded data comes from the complete data, the training purpose of the discriminator is to find the difference between the complete data and the real data. This makes the discriminator output as close to 1 as possible when the input is the complete data, and as close to 0 as possible when the input is the padded data obtained by the generative network, so as to distinguish their sources.
[0060] Figure 7 It is a structural diagram of the auxiliary prediction network. The auxiliary prediction network P is used to determine the label category of the complete data and the filled data obtained by the generating network. It is a multi-classification problem: if the label is a continuous variable, it is a regression problem. The difference between these two prediction problems lies in the inconsistency in the use of the loss function. The purpose of introducing the auxiliary prediction network P is to guide the generation of the filled data of the generating network and improve the prediction accuracy of the filled data. In addition, the MIGAN proposed in this embodiment can classify or regress all data while completing data filling, so that it can effectively combine the complete data with the missing data for high-precision prediction. Compared with the traditional filling algorithm, it is often necessary to separately model the filled data and the complete data after filling the missing data (this is because the filled data and the complete data are different in distribution, and predicting together will affect the prediction accuracy), which has certain advantages.
[0061] The generative adversarial network model MIGAN for missing data filling provided in this embodiment converts missing data into complete data through a generative model and keeps the distribution consistent with that of complete data, and makes the filling result more closely related to the category label by adding an auxiliary prediction model. In terms of the training process, it is proposed to perform training based on the batch gradient descent method, so that incomplete data sets with high dimensions and high missing rates can be filled.
[0062] Figure 8 is a working principle diagram of a water supply network control system. In one embodiment, Figure 8 As shown, a water supply network control system is also provided, which includes a controller and a controlled object. The controller is used to obtain target filling data generated by a terminal, generate control instructions according to the target filling data, and send the control instructions to the controlled object (such as a water pump, a water tank, a boiler regulating valve). The controlled object performs water supply operations according to the control instructions.
[0063] Combined with operation control, the generated data is input into the control system, and the control system generates control instructions to control the controlled object. At the same time, since the control instructions are generated by the generated data input into the control system, the quality of the generated control instructions is determined by the control strategy and control parameters on the one hand, and on the other hand, it is also determined by the generated data. Usually, the control strategy is fixed, then the more realistic the generated data and the better the control parameters, the better the generated control instructions and the better the control effect on the controlled object. Therefore, the present invention can be combined with the controlled object to realize data filling and control parameter training at the same time to achieve better control effect.
[0064] In addition, by using multiple different types of sensors at the same time and fusing their information, the sensor network can obtain more accurate and comprehensive environmental information, providing a more reliable basis for subsequent control. By selecting different types of sensors (such as position, velocity or acceleration sensors) for combination, specific needs can be met, thereby reducing the cost of sensor hardware. By adjusting the selection and combination of sensors in real time, the best sensing solution can be dynamically adjusted according to actual environmental changes, thereby improving the adaptability and flexibility of the system.
[0065] In addition, sensor information fusion can integrate multi-sensor data into a unified reference framework, simplifying the design process of the control system and making the control system easier to implement and maintain. Sensor information fusion can improve the robustness of the system by processing the differences between sensors; in the event of failure of some sensors, other sensors can continue to work, thus maintaining the stability of the system.
[0066] The present application also provides a computer device, whose internal structure diagram can be as follows: Fig. 9As shown. The computer device includes a processor, a memory, an input / output interface, a communication interface, a display unit and an input device. The processor, the memory and the input / output interface are connected via a system bus, and the communication interface, the display unit and the input device are connected to the system bus via the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be implemented through WIFI, a mobile cellular network, NFC (near field communication) or other technologies. When the computer program is executed by the processor, a method for filling water supply system data is implemented. The display unit of the computer device is used to form a visually visible image, and can be a display screen, a projection device or a virtual reality imaging device. The input device of the computer device can be a touch layer covered on the display screen, or a button, trackball or touchpad set on the computer device casing, or an external keyboard, touchpad or mouse, etc.
[0067] Those skilled in the art will understand that Fig. 9 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.
[0068] like Fig. 9 As shown, the present application also provides a computer device, which includes a memory, a processor, and a computer program stored in the memory, and the processor executes the computer program to implement the steps in the above-mentioned method embodiments.
[0069] The present application 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 steps in the above-mentioned method embodiments are implemented. The computer-readable storage medium may include, but is not limited to, any type of disk, including a floppy disk, an optical disk, a DVD, a CD-ROM, a micro drive, and a magneto-optical disk, a ROM, a RAM, an EPROM, an EEPROM, a DRAM, a VRAM, a flash memory device, a magnetic card or an optical card, a nanosystem (including a molecular memory IC), or any type of medium or device suitable for storing instructions and / or data.
[0070] The present application also provides a computer program product, including a computer program, which implements the steps in the above-mentioned method embodiments when executed by a processor.
[0071] It should be noted that, for the aforementioned method embodiments, for the sake of simplicity, they are all expressed as a series of action combinations, but those skilled in the art should be aware that the present application is not limited by the described order of actions, because according to the present application, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the present application.
[0072] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.
[0073] The above is only an exemplary embodiment of the present disclosure, and the scope of the present disclosure cannot be limited thereto. That is, any equivalent changes and modifications made according to the teachings of the present disclosure are still within the scope of the present disclosure. After considering the specification and practicing the disclosure here, those skilled in the art will easily think of the implementation scheme of the present disclosure. This application is intended to cover any modification, use or adaptation of the present disclosure, which follows the general principles of the present disclosure and includes common knowledge or customary technical means in the technical field not recorded in the present disclosure. The description and examples are only regarded as exemplary, and the scope and spirit of the present disclosure are defined by the claims.
[0074] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.
[0075] It will be easily understood by those skilled in the art that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.
Claims
1. A method for filling water supply system data, characterized in that: include: Obtain historical operating status data of the water supply system; A generative adversarial network model for missing data filling is established according to a generative network, a discriminative network and an auxiliary prediction network, wherein the input end of the generative network includes attribute missing data and a missing position vector, and the output end is the filling data, and the missing position vector includes missing information of the attribute missing data; the input end of the discriminative network includes the filling data and the complete data, and the output end is the probability that the filling data is the complete data; the input end of the auxiliary prediction network includes the filling data and the complete data, and the output end is the category of the filling data; Training a generative adversarial network model for missing data filling based on the historical operating status data; According to the generative adversarial network model for missing data filling that has been trained, target filling data is generated, and the target filling data is used to control the operating status of the water supply system in real time.
2. The method according to claim 1, characterized in that The obtaining of historical operation status data of the water supply system includes: Acquire historical operation status data of the water supply system, wherein the historical operation status data includes water supply flow, steam flow, steam pressure and corresponding collection time; The operating status data is dimensionlessly normalized to obtain a time series data set.
3. The method according to claim 2, characterized in that The missing position vector includes the category of attribute missing data and the collection time corresponding to the missing part, and the training of the generative adversarial network model for missing data filling based on the historical operating status data includes: Obtaining attribute missing data in the time series data set and missing position vectors and complete data corresponding to the attribute missing data; Inputting the attribute missing data and the corresponding missing position vector into the generation network to obtain the filling data; Inputting the filled data and the complete data into a discriminant network to obtain a probability that the filled data is the complete data; Inputting the filled data and the complete data into an auxiliary prediction network to obtain a category of the filled data; The generation network, the discriminant network, and the auxiliary prediction network are trained sequentially in the direction of minimizing the difference between the probability that the filled data is complete data and 1, and minimizing the difference between the category of the filled data and the category of the missing position vector, until the training stop condition is reached.
4. The method according to claim 3, characterized in that The generation network, the discrimination network, and the auxiliary prediction network are trained in sequence until a training stop condition is reached, including: The discriminative network and the generative network are trained adversarially so that the padded data follows the same distribution as the complete data.
5. The method according to claim 1, characterized in that The generation network, judgment network and auxiliary prediction network all use multi-layer fully connected neural networks.
6. The method according to claim 1, characterized in that The missing position vector is a discrete variable or a continuous variable.
7. A water supply network control system, characterized in that: Implementing the water supply system data filling method according to any one of claims 1 to 6 comprises: A controller, used to obtain target filling data and generate a control instruction according to the target filling data; The controlled object is used to perform water supply operation according to the control instruction.
8. A computer device, characterized in that: The invention comprises a memory, a processor and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the method according to any one of claims 1 to 6.
9. A computer-readable storage medium, characterized in that: A computer program is stored thereon, and when the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.
10. A computer program product, characterized in that The invention comprises a computer program, which implements the steps of the method according to any one of claims 1 to 6 when being executed by a processor.