Multi-model integrated nuclear power station DCS equipment fault prediction method and system

Abstract

The invention discloses a multi-model integrated nuclear power station DCS equipment fault prediction method and system, and relates to the field of nuclear power DCS equipment, and the method comprises the steps: matching the working condition data of each sampling point with the input parameters of each prediction model, and generating a multi-model input matrix; loading the historical working condition data to the multi-model input matrix, and training all prediction models by adopting a linear regression model and a regularization method based on the historical result data and the loaded multi-model input matrix to obtain a multi-model coefficient matrix; comparing influence relations among different prediction models, and determining a multi-model relation matrix; and inputting the target prediction time, the updated multi-model input matrix, the updated multi-model coefficient matrix and the updated multi-model relation matrix into a gating circulation unit to obtain a group of evaluation parameters of the target nuclear power DCS equipment at the target prediction time, and comparing the evaluation parameters with fault thresholds corresponding to the evaluation parameters to determine a fault type. According to the invention, the accuracy of fault prediction of the DCS equipment of the nuclear power station is improved.

Description

Technical Field

[0001] The present application relates to the technical field of nuclear power DCS equipment, and in particular to a multi-model integrated nuclear power plant DCS equipment fault prediction method and system. Background Art

[0002] Distributed Control System (DCS) is one of the important systems in nuclear power plants. It can realize the control, monitoring and data collection of various electronic equipment in nuclear power plants. However, due to various reasons, DCS equipment may fail, which will seriously affect the normal operation of nuclear power plants. Therefore, it is very important to regularly predict the faults of DCS equipment in nuclear power plants.

[0003] In the existing fault prediction technology of nuclear power plant DCS equipment, a single prediction model is generally used to realize the prediction of fault types in a specific dimension. For example, the cabinet temperature prediction model is used to realize the prediction of temperature changes, and the cabinet humidity prediction model is used to realize the prediction of humidity changes. Although these single prediction models realize the prediction of specific parameters of nuclear power plant DCS equipment, the influence relationship between the input and output of each model is not considered in the actual prediction process.

[0004] For the circuit board aging prediction model, changes in its input parameters such as temperature, humidity, vibration and radiation will not only directly affect the final output of the model, but will also indirectly affect the prediction of cabinet current and cabinet voltage. If these influence relationships between models are not considered, the final prediction of the failure result of the nuclear power plant DCS equipment will also be inaccurate. Summary of the invention

[0005] The purpose of this application is to provide a multi-model integrated nuclear power plant DCS equipment fault prediction method and system, which improves the accuracy of nuclear power plant DCS equipment fault prediction.

[0006] To achieve the above objectives, this application provides the following solutions:

[0007] In a first aspect, the present application provides a multi-model integrated nuclear power plant DCS equipment fault prediction method, the multi-model integrated nuclear power plant DCS equipment fault prediction method comprising:

[0008] Import multiple prediction models and complete the basic configuration of each prediction model; one prediction model is used to predict an evaluation parameter; the evaluation parameter is a parameter for determining whether a target nuclear power DCS device has a fault;

[0009] Determine the operating data of each sampling point in the target nuclear power DCS equipment; the operating data includes one or more operating parameters; the operating parameters are parameters of the target nuclear power DCS equipment during service;

[0010] Based on the binding relationship between each prediction model and each sampling point, the operating data of each sampling point is matched with the input parameters of each prediction model, and a multi-model input matrix is ​​generated; each row element of the multi-model input matrix corresponds to an input parameter of a prediction model;

[0011] Acquire historical operating condition data and historical result data; the historical operating condition data includes operating condition data of each sampling point at different historical moments; the historical result data includes evaluation parameters that should be predicted by each prediction model at different historical moments;

[0012] The historical operating condition data is loaded into the multi-model input matrix, and based on the historical result data and the loaded multi-model input matrix, all prediction models are trained using a linear regression model and a regularization method to obtain a multi-model coefficient matrix; the multi-model coefficient matrix is ​​composed of coefficients corresponding to the input parameters of each prediction model;

[0013] Compare the influence relationships between different prediction models and determine the multi-model relationship matrix; the influence relationship between different prediction models means that the output result of a prediction model at the current moment can be used as the input parameter of another or more prediction models at the next moment;

[0014] Acquire current operating condition data, and update the multi-model input matrix based on the current operating condition data; the current operating condition data includes the operating condition data of each sampling point at the current moment;

[0015] Determine a target prediction time, and input the target prediction time, the updated multi-model input matrix, the multi-model coefficient matrix, and the multi-model relationship matrix into a gated recurrent unit in a deep learning algorithm to obtain a set of evaluation parameters of a target nuclear power DCS device at the target prediction time;

[0016] A set of evaluation parameters of the target nuclear power DCS equipment at the target prediction time is compared with the fault threshold corresponding to each evaluation parameter to determine the fault type.

[0017] In a second aspect, the present application also provides a computer system, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the multi-model integrated nuclear power plant DCS equipment fault prediction method described in the first aspect.

[0018] According to the specific embodiments provided in this application, this application discloses the following technical effects:

[0019] In order to improve the accuracy of fault prediction of DCS equipment in nuclear power plants, this application no longer predicts evaluation parameters from a single dimension, but predicts all evaluation parameters as a whole from a multi-dimensional perspective, and finally obtains a set of evaluation parameters to determine the fault type of DCS equipment in nuclear power plants. First, this application constructs a multi-model input matrix based on the input parameters of all prediction models, and trains all prediction models through historical operating data and historical result data to obtain a multi-model coefficient matrix that characterizes the optimal coefficients corresponding to each input parameter. Secondly, this application considers the influence relationship between different prediction models, determines the multi-model relationship matrix, and ensures that each model can start from a multi-dimensional perspective when predicting evaluation parameters. Therefore, this application improves the accuracy of fault prediction of DCS equipment in nuclear power plants from a multi-dimensional perspective by determining the multi-model coefficient matrix, the multi-model relationship matrix and the target prediction time. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, 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 labor.

[0021] Figure 1 A flowchart of a multi-model integrated nuclear power plant DCS equipment fault prediction method provided in an embodiment of the present application;

[0022] Figure 2 A logic diagram of multi-model integrated fault prediction provided in an embodiment of the present application;

[0023] Figure 3 This is a diagram of the internal structure of a computer system provided in an embodiment of the present application. DETAILED DESCRIPTION

[0024] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0025] The purpose of this application is to provide a multi-model integrated nuclear power plant DCS equipment fault prediction method and system, which improves the accuracy of nuclear power plant DCS equipment fault prediction.

[0026] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0027] Example 1

[0028] This embodiment provides a multi-model integrated fault prediction method for nuclear power plant DCS equipment. Since the existing nuclear power plant DCS equipment is composed of hundreds of control cabinets, each control cabinet contains multiple control boxes, and each control box contains multiple industrial boards, the following will take a single control cabinet as the target to introduce how a single control cabinet can achieve multi-model integrated fault prediction (the fault prediction method of each control cabinet is the same). Figure 1 and Figure 2 As shown, the multi-model integrated nuclear power plant DCS equipment fault prediction method includes:

[0029] Step S1: Import multiple prediction models and complete the basic configuration of each prediction model.

[0030] In this embodiment, the failure of a single control cabinet needs to be comprehensively judged based on multiple evaluation parameters. The multiple evaluation parameters are cabinet temperature, cabinet humidity, cabinet vibration amplitude, air pollution index, radiation concentration, cabinet current, cabinet voltage, circuit board aging rate and load change rate. One prediction model predicts one evaluation parameter.

[0031] Specifically, the imported prediction models are respectively a cabinet temperature prediction model, a cabinet humidity prediction model, a cabinet vibration amplitude prediction model, an air pollution index prediction model, a radiation concentration prediction model, a cabinet current prediction model, a cabinet voltage prediction model, a circuit board aging rate prediction model, and a load change rate prediction model. The above prediction models are imported into the industrial boards in the control box one by one, and the operating parameters, system resource availability, timeout threshold of each prediction model are set respectively, as well as the binding relationship between each prediction model and each sampling point.

[0032] Step S2: Determine the operating data of each sampling point in the target nuclear power DCS equipment.

[0033] In this embodiment, the working condition data of each sampling point includes one or more working condition parameters, and the working condition parameters are parameters of the target nuclear power DCS equipment during service. For example: it is known that sampling point 1 in the target nuclear power DCS equipment can collect the working condition parameter of "cabinet external vibration source" in real time, sampling point 2 can collect the two working condition parameters of "cabinet external temperature" and "cabinet internal temperature" in real time, and sampling point 3 can collect the working condition parameter of "cabinet ventilation volume" in real time. If the input parameters of the cabinet temperature prediction model include "cabinet external temperature, cabinet internal temperature, cabinet ventilation volume", then the cabinet temperature prediction model needs to be bound to sampling points 2 and 3; if the input parameters of the cabinet vibration amplitude prediction model include "cabinet external vibration source, cabinet ventilation volume", then the cabinet vibration amplitude prediction model needs to be bound to sampling points 1 and 3.

[0034] Step S3: Based on the binding relationship between each prediction model and each sampling point, the operating data of each sampling point is matched with the input parameters of each prediction model, and a multi-model input matrix is ​​generated.

[0035] In this embodiment, the number of rows of the multi-model input matrix corresponds to the number of prediction models, and the number of columns of the multi-model input matrix corresponds to the number of input parameters of the optimal prediction model (the prediction model with the most input parameters). For prediction models that do not meet the number of input parameters of the optimal prediction model, the positions of the spare input parameters in the multi-model input matrix are replaced by "0".

[0036] Specifically, the cabinet temperature prediction model R 1 The input parameters include: cabinet external temperature, cabinet internal temperature, air conditioner set temperature, air conditioner cooling efficiency, single control box load, cabinet ventilation volume, cabinet internal humidity, number of servers in the cabinet and server power in the cabinet.

[0037] Cabinet humidity prediction model R 2 The input parameters include: cabinet external humidity, cabinet internal humidity, air conditioning humidity control, air conditioning humidity adjustment efficiency and cabinet ventilation volume.

[0038] Cabinet vibration amplitude prediction model R 3 The input parameters include: equipment operating parameters, external vibration source, cabinet internal temperature, cabinet internal humidity and cabinet ventilation volume.

[0039] Air Pollution Index Prediction Model R 4 The input parameters include: cabinet external air quality, cabinet internal pollution sources, air conditioning filtration efficiency, cabinet internal temperature, cabinet internal humidity and cabinet ventilation volume.

[0040] Radiation concentration prediction model R 5 The input parameters include: radiation monitoring data, cabinet internal temperature, cabinet internal humidity and radiation purification rate;

[0041] Cabinet current prediction model R 6 The input parameters include: historical current data, load changes, power supply stability and cabinet temperature;

[0042] Cabinet voltage prediction model R 7 The input parameters include: historical voltage data, load changes, power supply stability and cabinet temperature;

[0043] Circuit board aging rate prediction model R 8 The input parameters include: cabinet temperature, cabinet humidity, current load, voltage load, material properties, radiation concentration, air pollution index, circuit board usage time and historical maintenance records;

[0044] Load change rate prediction model R 9 The input parameters include: CPU usage, MEM usage, Disk usage, Net usage, cabinet temperature, cabinet humidity, current load, voltage load and circuit board aging rate.

[0045] As a preferred implementation, the multi-model input matrix is ​​specifically:

[0046]

[0047] Where, T ext is the cabinet external temperature, T heat is the internal temperature of the cabinet, T ac Set the temperature for the air conditioner, E ac is the air conditioning refrigeration efficiency, L dev For a single control box load, V vent is the ventilation volume of the cabinet, H heat is the humidity inside the cabinet, N srv is the number of servers in the cabinet, P srv is the server power in the cabinet, H ext is the external humidity of the cabinet, H heat is the humidity inside the cabinet, H ac For air conditioning humidity control, E hac is the humidity control efficiency of air conditioner, P dev is the equipment operating parameter, V ext is the external vibration source, I ext is the air quality outside the cabinet, I dev is the pollution source inside the cabinet, E iac is the air conditioning filtration efficiency, E data is the radiation monitoring data, P is the radiation purification rate, A hist is the historical current data, L change is the load change, S stab is the power supply stability, T env is the cabinet temperature, V hist is the historical voltage data, H env is the cabinet humidity, A env is the current load, V env is the voltage load, M env is the material property, E env is the radiation concentration, I env is the air pollution index, M is the circuit board usage time, R is the historical maintenance record, and U is the CPU is the CPU usage, U MEM is the MEM usage rate, U Disk is the disk usage, U Net is the Net utilization rate, C env is the aging rate of the circuit board.

[0048] Step S4: Obtain historical operating condition data and historical result data.

[0049] In this embodiment, the historical operating condition data includes the operating condition data of each sampling point at different historical moments, the historical result data includes the evaluation parameters that each prediction model should predict at different historical moments, the historical operating condition data is used as the input of the prediction model, and the historical result data is used as the output of the prediction model.

[0050] Step S5: Load the historical operating condition data into the multi-model input matrix, and based on the historical result data and the loaded multi-model input matrix, use the linear regression model and regularization method to train all prediction models to obtain the multi-model coefficient matrix.

[0051] In this embodiment, the multi-model coefficient matrix is ​​composed of coefficients corresponding to the input parameters of each prediction model. The process of determining the multi-model coefficient matrix is ​​the process of training each prediction model. The multi-model coefficient matrix finally obtained by training is:

[0052]

[0053] In the formula, {α 1 ,α 2 ,α 3 ,α 4 ,α 5 ,α 6 ,α 7 ,α 8 ,α 9} is the coefficient corresponding to each input parameter in the cabinet temperature prediction model, {β 1 ,β 2 ,β 3 ,β 4 ,β 5} is the coefficient corresponding to each input parameter in the cabinet humidity prediction model, {γ 1 ,γ 2 ,γ 3 ,γ 4 ,γ 5} is the coefficient corresponding to each input parameter in the cabinet vibration amplitude prediction model, {δ 1 ,δ 2 ,δ 3 ,δ 4 ,δ 5 ,δ 6} is the coefficient corresponding to each input parameter in the air pollution index prediction model, {ε 1 ,ε 2 ,ε 3 ,ε 4} is the coefficient corresponding to each input parameter in the radiation concentration prediction model, {σ 1,σ 2 ,σ 3 ,σ 4} is the coefficient corresponding to each input parameter in the cabinet current prediction model, {θ 1 ,θ 2 ,θ 3 ,θ 4} is the coefficient corresponding to each input parameter in the cabinet voltage prediction model, is the coefficient corresponding to each input parameter in the circuit board aging rate prediction model, {μ 1 ,μ 2 ,μ 3 ,μ 4 ,μ 5 ,μ 6 ,μ 7 ,μ 8 ,μ 9} is the coefficient corresponding to each input parameter in the load change rate prediction model.

[0054] Furthermore, the trained cabinet temperature prediction model R 1 The cabinet temperature r 1 satisfy:

[0055]

[0056] The trained cabinet humidity prediction model R 2 The cabinet temperature r 2 satisfy:

[0057] r 2 =β 1 ×H ext +β 2 ×H heat +β 3 ×H ac +β 4 ×E hac +β 5 ×V vent ;

[0058] The cabinet vibration amplitude prediction model R after training 3 The cabinet temperature r 3 satisfy:

[0059] r 3 =γ 1 ×P dev +γ 2 ×V ext +γ 3 ×T heat +γ 4 ×H heat +γ 5 ×V vent ;

[0060] Air pollution index prediction model R after training 4 The cabinet temperature r 4 satisfy:

[0061] r 4 =δ 1 ×I ext +δ 2 ×I dev +δ 3 ×E iac +δ 4 ×T heat +δ 5 ×H heat +δ 6 ×V vent ;

[0062] The radiation concentration prediction model R after training 5 The cabinet temperature r 5 satisfy:

[0063] r 5 =ε 1 ×E data +ε 2 ×T heat +ε 3 ×H heat +ε 4 ×P;

[0064] The trained cabinet current prediction model R 6 The cabinet temperature r 6 satisfy:

[0065] r 6 =σ 1 ×A hist +σ 2 ×L change +σ 3 ×S stab +σ 4 ×T env ;

[0066] The trained cabinet voltage prediction model R 7 The cabinet temperature r 7 satisfy:

[0067] r 7 =θ 1 ×V hist +θ 2 ×L change +θ 3 ×S stab +θ 4 ×T env ;

[0068] The trained circuit board aging rate prediction model R 8 The cabinet temperature r 8 satisfy:

[0069]

[0070] Load change rate prediction model R after training 9 The cabinet temperature r 9 satisfy:

[0071]

[0072] Step S6: Compare the influence relationships between different prediction models and determine the multi-model relationship matrix.

[0073] In this embodiment, the influence relationship between different prediction models means that the output result of a prediction model at the current moment can be used as an input parameter of another or more prediction models at the next moment. The multi-model relationship matrix is ​​specifically:

[0074]

[0075] In the formula, the first row of elements represents R 9 The output result at the current moment is R 1 The next moment L dev Use, R 2 The output result at the current moment is R 1 The next moment H heat Use, R 1 The other input parameters remain unchanged.

[0076] The second row of elements represents R 1 The output result at the current moment is R 2 The next moment H ext Use, R 2 The other input parameters remain unchanged.

[0077] The third row of elements represents R 1 The output result at the current moment is R 3 T at the next moment heat Use, R 2 The output result at the current moment is R 3 The next moment H heat Use, R 3 The other input parameters remain unchanged.

[0078] The fourth row of elements represents R 1 The output result at the current moment is R 4 The next moment H heat Use, R 2The output result at the current moment is R 4 The next moment V vent Use, R 4 The other input parameters remain unchanged.

[0079] The fifth row element represents R 1 The output result at the current moment is R 5 T at the next moment heat Use, R 2 The output result at the current moment is R 5 The next moment H heat Use, R 5 The other input parameters remain unchanged.

[0080] The sixth row element represents R 1 The output result at the current moment is R 6 T at the next moment env Use, R 6 The other input parameters remain unchanged.

[0081] The seventh row element represents R 1 The output result at the current moment is R 7 T at the next moment env Use, R 7 The other input parameters remain unchanged.

[0082] The eighth row element represents R 1 The output result at the current moment is R 8 T at the next moment env Use, R 2 The output result at the current moment is R 8 The next moment H env Use, R 3 The output result at the current moment is R 8 A at the next moment env Use, R 7 The output result at the current moment is R 8 The next moment V env Use, R 5 The output result at the current moment is R 8 The next moment E env Use, R 4 The output result at the current moment is R 8 The next moment I env Use, R 8 The other input parameters remain unchanged.

[0083] The ninth row element represents R 1 The output result at the current moment is R 9 T at the next momentenv Use, R 2 The output result at the current moment is R 9 The next moment H env Use, R 6 The output result at the current moment is R 9 A at the next moment env Use, R 7 The output result at the current moment is R 9 The next moment V env Use, R 8 The output result at the current moment is R 9 C at the next moment env Use, R 9 The other input parameters remain unchanged.

[0084] Step S7: Acquire current operating condition data, and update the multi-model input matrix based on the current operating condition data.

[0085] In this embodiment, before predicting whether the target nuclear power DCS equipment will fail at the target prediction time, it is necessary to first obtain the operating data of each sampling point at the current moment, that is, to determine the initial value of the multi-model input matrix for subsequent predictions, and perform subsequent calculations based on the initial value of the multi-model input matrix.

[0086] Step S8: Determine the target prediction time, and input the target prediction time, the updated multi-model input matrix, the multi-model coefficient matrix and the multi-model relationship matrix into the gated loop unit in the deep learning algorithm to obtain a set of evaluation parameters of the target nuclear power DCS equipment at the target prediction time.

[0087] Specifically, when updating gate z t Inside: z t =σ(CM z ·x t +RM z ·h t-1 );CM z ,RM z The coefficient of the corresponding update gate, CM z It is automatically extracted from the multi-model coefficient matrix CM by the random forest machine learning algorithm, RM z is automatically extracted from the multi-model relationship matrix RM by the random forest machine learning algorithm; σ(·) is the activation function; x t is the input feature vector of the target prediction time t, including the input parameters of all prediction models, whose initial values ​​correspond to the updated multi-model input matrix; h t-1 is the hidden state at the previous time step.

[0088] Reset gate r t Inside: r t=σ(CM r ·x t +RM r ·h t-1 );CM t ,RM t The coefficient corresponding to the reset gate, CM r It is automatically extracted from the multi-model coefficient matrix CM by the random forest machine learning algorithm, RM r It is automatically extracted from the multi-model relationship matrix RM through the random forest machine learning algorithm.

[0089] Candidate hidden states CM h ,RM h Corresponding to the candidate hidden coefficient, Represents r t and h t-1 The element-wise product of ; the current hidden state

[0090] As a preferred implementation, a set of evaluation parameters finally predicted is:

[0091]

[0092] Step S9: Compare a set of evaluation parameters of the target nuclear power DCS equipment at the target prediction time with the fault threshold corresponding to each evaluation parameter to determine the fault type.

[0093] In this embodiment, the prediction result Result is compared with the fault threshold of each model input parameter item by item:

[0094]

[0095] In the above fault threshold set Fault_Threshold, T threshold is the cabinet temperature fault threshold, H threshold is the cabinet humidity fault threshold, V threshold is the cabinet vibration amplitude fault threshold, I threshold is the air pollution index fault threshold, E threshold is the radiation concentration fault threshold, A threshold is the cabinet current fault threshold, W threshold is the cabinet voltage fault threshold, C threshold is the circuit board aging rate failure threshold, L threshold is the load change rate fault threshold. For example, when the cabinet temperature is greater than the cabinet temperature fault threshold, the fault type "the cabinet temperature is abnormal at the target prediction time t, and the abnormal temperature is ××℃" is output.

[0096] In summary, the nuclear power plant DCS equipment fault prediction method based on the above-mentioned multi-model integration can input different prediction time values ​​respectively during the system call process to complete the multi-model integrated fault prediction of the target DCS equipment at different prediction times. Its effect is more significant than that of a single model, and the mutual correction prediction method between multiple models can further improve the accuracy of the prediction results.

[0097] Example 2

[0098] This embodiment provides a computer system, which may be a server or a terminal, and its internal structure diagram may be as follows: Figure 3 As shown. The computer system includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. The processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer system is used to provide computing and control capabilities. The memory of the computer system includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer system is used to store the operating condition data of each sampling point in the target nuclear power DCS equipment. The input / output interface of the computer system is used to exchange information between the processor and an external device. The communication interface of the computer system is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a multi-model integrated nuclear power plant DCS equipment fault prediction method is implemented.

[0099] Those skilled in the art will understand that Figure 3 The structure shown in the figure is only a block diagram of a part of the structure related to the scheme of the present application, and does not constitute a limitation on the computer system to which the scheme of the present application is applied. The specific computer system may include more or fewer components than shown in the figure, or combine certain components, or have a different arrangement of components.

[0100] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0101] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM may be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0102] The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. The non-relational database may include a distributed database based on blockchain, etc., but is not limited thereto. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but is not limited thereto.

[0103] All actions to obtain signals, information or data in this application are carried out in compliance with the relevant data protection laws and policies of the country where they are located, and with the authorization given by the owner of the corresponding device.

[0104] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0105] This article uses specific examples to illustrate the principles and implementation methods of this application. The description of the above embodiments is only used to help understand the method and core ideas of this application. At the same time, for those skilled in the art, according to the ideas of this application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A multi-model integrated nuclear power plant DCS equipment fault prediction method, characterized in that: The multi-model integrated nuclear power plant DCS equipment fault prediction method includes: Import multiple prediction models and complete the basic configuration of each prediction model; one prediction model is used to predict an evaluation parameter; the evaluation parameter is a parameter for determining whether a target nuclear power DCS device has a fault; Determine the operating data of each sampling point in the target nuclear power DCS equipment; the operating data includes one or more operating parameters; the operating parameters are parameters of the target nuclear power DCS equipment during service; Based on the binding relationship between each prediction model and each sampling point, the operating data of each sampling point is matched with the input parameters of each prediction model, and a multi-model input matrix is ​​generated; each row element of the multi-model input matrix corresponds to an input parameter of a prediction model; Acquire historical operating condition data and historical result data; the historical operating condition data includes operating condition data of each sampling point at different historical moments; the historical result data includes evaluation parameters that should be predicted by each prediction model at different historical moments; The historical operating condition data is loaded into the multi-model input matrix, and based on the historical result data and the loaded multi-model input matrix, all prediction models are trained using a linear regression model and a regularization method to obtain a multi-model coefficient matrix; the multi-model coefficient matrix is ​​composed of coefficients corresponding to the input parameters of each prediction model; Compare the influence relationships between different prediction models and determine the multi-model relationship matrix; the influence relationship between different prediction models means that the output result of a prediction model at the current moment can be used as the input parameter of another or more prediction models at the next moment; Acquire current operating condition data, and update the multi-model input matrix based on the current operating condition data; the current operating condition data includes the operating condition data of each sampling point at the current moment; Determine a target prediction time, and input the target prediction time, the updated multi-model input matrix, the multi-model coefficient matrix, and the multi-model relationship matrix into a gated recurrent unit in a deep learning algorithm to obtain a set of evaluation parameters of a target nuclear power DCS device at the target prediction time; A set of evaluation parameters of the target nuclear power DCS equipment at the target prediction time is compared with the fault threshold corresponding to each evaluation parameter to determine the fault type.

2. The multi-model integrated nuclear power plant DCS equipment fault prediction method according to claim 1 is characterized in that: The number of rows of the multi-model input matrix corresponds to the number of prediction models; the number of columns of the multi-model input matrix corresponds to the number of input parameters of the optimal prediction model; the optimal prediction model is the prediction model with the most input parameters.

3. The multi-model integrated nuclear power plant DCS equipment fault prediction method according to claim 1, characterized in that: The basic configuration of each prediction model includes at least: Set the system resource availability of each prediction model, the timeout threshold of each prediction model, and the binding relationship between each prediction model and each sampling point.

4. The multi-model integrated nuclear power plant DCS equipment fault prediction method according to claim 1, characterized in that: The multiple imported prediction models are cabinet temperature prediction model, cabinet humidity prediction model, cabinet vibration amplitude prediction model, air pollution index prediction model, radiation concentration prediction model, cabinet current prediction model, cabinet voltage prediction model, circuit board aging rate prediction model and load change rate prediction model.

5. The multi-model integrated nuclear power plant DCS equipment fault prediction method according to claim 4 is characterized in that: The input parameters of the cabinet temperature prediction model include: cabinet external temperature, cabinet internal temperature, air conditioner set temperature, air conditioner cooling efficiency, single control box load, cabinet ventilation volume, cabinet internal humidity, number of servers in the cabinet and power of servers in the cabinet; The input parameters of the cabinet humidity prediction model include: cabinet external humidity, cabinet internal humidity, air conditioning humidity control, air conditioning humidity adjustment efficiency and cabinet ventilation volume; The input parameters of the cabinet vibration amplitude prediction model include: equipment operating parameters, external vibration source, cabinet internal temperature, cabinet internal humidity and cabinet ventilation volume; The input parameters of the air pollution index prediction model include: cabinet external air quality, cabinet internal pollution source, air conditioning filtration efficiency, cabinet internal temperature, cabinet internal humidity and cabinet ventilation volume; the input parameters of the radiation concentration prediction model include: radiation monitoring data, cabinet internal temperature, cabinet internal humidity and radiation purification rate; The input parameters of the cabinet current prediction model include: historical current data, load changes, power supply stability and cabinet temperature; The input parameters of the cabinet voltage prediction model include: historical voltage data, load changes, power supply stability and cabinet temperature; The input parameters of the circuit board aging rate prediction model include: cabinet temperature, cabinet humidity, current load, voltage load, material properties, radiation concentration, air pollution index, circuit board usage time and historical maintenance records; The input parameters of the load change rate prediction model include: CPU usage, MEM usage, Disk usage, Net usage, cabinet temperature, cabinet humidity, current load, voltage load and circuit board aging rate.

6. The multi-model integrated nuclear power plant DCS equipment fault prediction method according to claim 5, characterized in that: The multi-model input matrix is: Where, T ext is the cabinet external temperature, T heat is the internal temperature of the cabinet, T ac Set the temperature for the air conditioner, E ac is the air conditioning refrigeration efficiency, L dev is the load of a single control box, V vent is the ventilation volume of the cabinet, H heat is the humidity inside the cabinet, N srv is the number of servers in the cabinet, P srv is the server power in the cabinet, H ext is the external humidity of the cabinet, H heat is the humidity inside the cabinet, H ac For air conditioning humidity control, E hac is the humidity control efficiency of air conditioner, P dev is the equipment operating parameter, V ext is the external vibration source, I ext is the air quality outside the cabinet, I dev is the pollution source inside the cabinet, E iac is the air conditioning filtration efficiency, E data is the radiation monitoring data, P is the radiation purification rate, A hist is the historical current data, L change is the load change, S stab is the power supply stability, T env is the cabinet temperature, V hist is the historical voltage data, H env is the cabinet humidity, A env is the current load, V env is the voltage load, M env is the material property, E env is the radiation concentration, I env is the air pollution index, M is the circuit board usage time, R is the historical maintenance record, and U is the CPU is the CPU usage, U MEM is the MEM usage rate, U Disk is the disk usage, U Net is the Net utilization rate, C env is the aging rate of the circuit board.

7. The multi-model integrated nuclear power plant DCS equipment fault prediction method according to claim 6, characterized in that: The multi-model coefficient matrix is: Wherein, {α1, α2, α3, α4, α5, α6, α7, α8, α9} are coefficients corresponding to the input parameters in the cabinet temperature prediction model, {β1, β2, β3, β4, β5} are coefficients corresponding to the input parameters in the cabinet humidity prediction model, {γ1, γ2, γ3, γ4, γ5} are coefficients corresponding to the input parameters in the cabinet vibration amplitude prediction model, {δ1, δ2, δ3, δ4, δ5, δ6} are coefficients corresponding to the input parameters in the air pollution index prediction model, {ε1, ε2, ε3, ε4} are coefficients corresponding to the input parameters in the radiation concentration prediction model, {σ1, σ2, σ3, σ4} are coefficients corresponding to the input parameters in the cabinet current prediction model, {θ1, θ2, θ3, θ4} are coefficients corresponding to the input parameters in the cabinet voltage prediction model, are the coefficients corresponding to the input parameters in the circuit board aging rate prediction model, and {μ1, μ2, μ3, μ4, μ5, μ6, μ7, μ8, μ9} are the coefficients corresponding to the input parameters in the load change rate prediction model.

8. The multi-model integrated nuclear power plant DCS equipment fault prediction method according to claim 6, characterized in that: The multi-model relationship matrix is: In the formula, the first row of elements represents the output result of the load change rate prediction model R9 at the current moment as the L of the cabinet temperature prediction model R1 at the next moment. dev Use the output result of the cabinet humidity prediction model R2 at the current moment as the H of R1 at the next moment. heat Use, the other input parameters of R1 remain unchanged; The second row of elements indicates that the output result of R1 at the current moment is used as the H of R2 at the next moment. ext Use, the other input parameters of R2 remain unchanged; The third row of elements indicates that the output result of R1 at the current moment is used as the T of the cabinet vibration amplitude prediction model R3 at the next moment. heat Use the output result of R2 at the current moment as the H of R3 at the next moment. heat Use, the other input parameters of R3 remain unchanged; The fourth row of elements indicates that the output result of R1 at the current moment is used as the H of the air pollution index prediction model R4 at the next moment. heat Use the output result of R2 at the current moment as the V of R4 at the next moment. vent Use, the other input parameters of R4 remain unchanged; The fifth row of elements indicates that the output result of R1 at the current moment is used as the T of the radiation concentration prediction model R5 at the next moment. heat Use the output result of R2 at the current moment as the H of R5 at the next moment. heat Use, the other input parameters of R5 remain unchanged; The sixth row of elements indicates that the output result of R1 at the current moment is used as the T of the cabinet current prediction model R6 at the next moment. env Use, the other input parameters of R6 remain unchanged; The seventh row of elements indicates that the output result of R1 at the current moment is used as the T of the cabinet voltage prediction model R7 at the next moment. env Use, the other input parameters of R7 remain unchanged; The eighth row element indicates that the output result of R1 at the current moment is used as the T of the circuit board aging rate prediction model R8 at the next moment. env Use the output result of R2 at the current moment as the H of R8 at the next moment. env Use the output result of R3 at the current moment as the A of R8 at the next moment. env Use the output result of R7 at the current moment as the V of R8 at the next moment. env Use the output result of R5 at the current moment as the E of R8 at the next moment. env Use the output result of R4 at the current moment as the I of R8 at the next moment. env Use, the other input parameters of R8 remain unchanged; The ninth row of elements indicates that the output result of R1 at the current moment is used as the T of the load change rate prediction model R9 at the next moment. env Use the output result of R2 at the current moment as the H of R9 at the next moment. env Use the output result of R6 at the current moment as the A of R9 at the next moment. env Use the output result of R7 at the current moment as the V of R9 at the next moment. env Use the output result of R8 at the current moment as the C of R9 at the next moment. env Use, the other input parameters of R9 remain unchanged.

9. The multi-model integrated nuclear power plant DCS equipment fault prediction method according to claim 1, characterized in that: A set of evaluation parameters of the target nuclear power DCS equipment at the target prediction time includes: cabinet temperature, cabinet humidity, cabinet vibration amplitude, air pollution index, radiation concentration, cabinet current, cabinet voltage, circuit board aging rate and load change rate.

10. A computer system comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the multi-model integrated nuclear power plant DCS equipment fault prediction method described in any one of claims 1 to 9.

Images