CN111797533B - Nuclear power device operation parameter abnormity detection method and system - Google Patents
Nuclear power device operation parameter abnormity detection method and system Download PDFInfo
- Publication number
- CN111797533B CN111797533B CN202010654715.4A CN202010654715A CN111797533B CN 111797533 B CN111797533 B CN 111797533B CN 202010654715 A CN202010654715 A CN 202010654715A CN 111797533 B CN111797533 B CN 111797533B
- Authority
- CN
- China
- Prior art keywords
- real
- time
- operation parameters
- value
- divergence
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2111/00—Details relating to CAD techniques
- G06F2111/08—Probabilistic or stochastic CAD
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computer Hardware Design (AREA)
- Geometry (AREA)
- Testing And Monitoring For Control Systems (AREA)
Abstract
The invention relates to a method and a system for detecting the running parameter abnormity of a nuclear power device, wherein the method comprises the following steps: acquiring real-time operation parameters of a nuclear power plant; calculating a covariance matrix of the real-time operation parameters; calculating a characteristic value set and a characteristic vector set of the covariance matrix; reducing the dimension of the feature vector set by adopting a principal component analysis method according to the feature value set to obtain a dimension-reduced feature vector set; selecting effective real-time operation parameters from the real-time operation parameters according to the dimensionality reduction feature vector set to obtain the effective real-time operation parameters; acquiring a reference operation parameter; calculating real-time divergence values of the effective real-time operating parameters and the reference operating parameters; acquiring a divergence value threshold; the divergence value threshold is determined from a reference operating parameter; and carrying out abnormity detection on the real-time operation parameters according to the real-time divergence value and the divergence value threshold value. The method and the system can detect the abnormality of the real-time operation parameters and improve the accuracy of the abnormality detection.
Description
Technical Field
The invention relates to the technical field of nuclear power device fault diagnosis, in particular to a method and a system for detecting the abnormal operation parameters of a nuclear power device.
Background
Nuclear power plants are complex in construction, and have a radioactive hazard, which places extremely high demands on safety. Meanwhile, the nuclear power system works continuously for a long time, and faults are easy to occur, if equipment fails and cannot be detected and found in time, serious radioactive release consequences can be caused, and public safety and environmental conditions are damaged.
At present, most of the abnormal detection technologies for nuclear power systems and key equipment adopt traditional threshold analysis and manual experience for judgment. However, these conventional techniques do not fully accommodate the reliability requirements of complex systems and critical equipment. With the continuous development of the artificial intelligence technology, the abnormal state of the operation parameters of the nuclear power device is timely detected by adopting efficient and accurate statistical analysis and the artificial intelligence technology, so that the occurrence of major damage faults and even serious accidents can be avoided, the operation guarantee capability of a nuclear power system and key equipment can be effectively improved, the potential safety hazard is reduced, and the independent guarantee is realized. The existing method for detecting the abnormity of the operation parameters of the nuclear power device is mainly based on data analysis and a model.
The method based on data analysis is characterized by big data, and a data analysis model is constructed through a large amount of training of historical data, and although the modeling process is relatively simple, the cost is high, and the accuracy is low. In the model-based method, because a system and equipment in the nuclear power plant have strong coupling and nonlinearity, an accurate and reliable mathematical model is difficult to establish, and the accuracy of a result is difficult to ensure by the model-based detection method; meanwhile, the dependence relationship between the mathematical model and the analysis object is strong, and the universality is poor, so that the modeling workload aiming at different objects is overlarge.
Disclosure of Invention
The invention aims to provide a method and a system for detecting the abnormity of the running parameters of a nuclear power device, which can detect the abnormity of the running parameters in real time and improve the accuracy of abnormity detection.
In order to achieve the purpose, the invention provides the following scheme:
a nuclear power plant operation parameter anomaly detection method comprises the following steps:
acquiring real-time operation parameters of a nuclear power plant;
calculating a covariance matrix of the real-time operating parameters;
calculating a feature value set and a feature vector set of the covariance matrix; the set of eigenvalues includes each eigenvalue of the covariance matrix; the feature vector set comprises a feature vector corresponding to each feature value;
reducing the dimension of the feature vector set by adopting a principal component analysis method according to the feature value set to obtain a dimension-reduced feature vector set;
according to the dimensionality reduction feature vector set, effective real-time operation parameters are selected from the real-time operation parameters to obtain effective real-time operation parameters;
acquiring a reference operating parameter; the reference operation parameter is an operation parameter consistent with the real-time operation parameter working condition in historical normal operation parameters;
calculating real-time divergence values of the effective real-time operating parameters and the reference operating parameters;
acquiring a divergence value threshold; the divergence value threshold is determined from the reference operating parameter;
and carrying out abnormity detection on the real-time operation parameters according to the real-time divergence value and the divergence value threshold.
Optionally, the calculating the covariance matrix of the real-time operation parameter specifically includes:
normalizing the real-time operation parameters by adopting a dispersion normalization method to obtain normalized operation parameters;
mapping the normalized operation parameters by adopting a radial basis kernel function to obtain a kernel matrix;
centralizing the kernel matrix to obtain a centralized kernel matrix;
a covariance matrix of the centered kernel matrix is calculated.
Optionally, the reducing dimensions of the feature vector set by using a principal component analysis method according to the feature value set to obtain a dimension-reduced feature vector set specifically includes:
according to the formula
Calculating the accumulated contribution rate of the principal component variance; wherein PCN is the principal component variance cumulative contribution rate, λiIs the ith characteristic value, i is 1,2,3l, lambdajJ is the j-th eigenvalue, j is 1,2,3n,l＜n；
sorting the eigenvalues in the eigenvalue set according to the size to obtain a sorted eigenvalue set;
according to the accumulated contribution rate of the pivot variance, selecting the first l eigenvalues from the sorting eigenvalue set as a dimension reduction eigenvalue set;
and selecting a feature vector corresponding to each feature value in the dimensionality reduction feature value set from the feature vector set as a dimensionality reduction feature vector set.
Optionally, the calculating a real-time variance value of the effective real-time operating parameter and the reference operating parameter specifically includes:
performing sliding time window processing on the effective real-time operation parameters to obtain a real-time operation data matrix;
performing sliding time window processing on the reference operation parameters to obtain a reference operation data matrix;
and determining a real-time divergence value according to the real-time operation data matrix and the reference operation data matrix.
Optionally, the performing, according to the real-time divergence value and the divergence threshold value, abnormality detection on the real-time operation parameter specifically includes:
judging whether the real-time divergence value is larger than the divergence value threshold value or not to obtain a judgment result;
if the judgment result shows that the real-time divergence value is larger than the divergence value threshold value, determining that the real-time operation parameter is abnormal;
and if the judgment result shows that the real-time divergence value is smaller than or equal to the divergence value threshold value, returning to the step of acquiring the real-time operation parameters of the nuclear power device.
A nuclear power plant operating parameter anomaly detection system, comprising:
the real-time operation parameter acquisition module is used for acquiring real-time operation parameters of the nuclear power device;
the covariance matrix calculation module is used for calculating a covariance matrix of the real-time operation parameters;
the eigenvalue and eigenvector calculation module is used for calculating an eigenvalue set and an eigenvector set of the covariance matrix; the set of eigenvalues comprises each eigenvalue of the covariance matrix; the feature vector set comprises a feature vector corresponding to each feature value;
the dimensionality reduction module is used for reducing dimensionality of the feature vector set by adopting a principal component analysis method according to the feature value set to obtain a dimensionality reduction feature vector set;
the effective real-time operation parameter selection module is used for selecting effective real-time operation parameters from the real-time operation parameters according to the dimensionality reduction feature vector set to obtain effective real-time operation parameters;
the reference operation parameter acquisition module is used for acquiring reference operation parameters; the reference operation parameter is an operation parameter consistent with the real-time operation parameter working condition in historical normal operation parameters;
the real-time divergence value calculation module is used for calculating real-time divergence values of the effective real-time operation parameters and the reference operation parameters;
the divergence value threshold value acquisition module is used for acquiring a divergence value threshold value; the divergence value threshold is determined from the reference operating parameter;
and the anomaly detection module is used for carrying out anomaly detection on the real-time operation parameters according to the real-time divergence value and the divergence value threshold.
Optionally, the covariance matrix calculation module specifically includes:
the normalization unit is used for normalizing the real-time operation parameters by adopting a dispersion normalization method to obtain normalized operation parameters;
the mapping unit is used for mapping the normalized operation parameters by adopting a radial basis kernel function to obtain a kernel matrix;
the centralization unit is used for centralizing the nuclear matrix to obtain a centralized nuclear matrix;
and the covariance matrix calculation unit is used for calculating the covariance matrix of the centralized kernel matrix.
Optionally, the dimension reduction module specifically includes:
a principal component variance cumulative contribution rate calculation unit for calculating a principal component variance cumulative contribution rate according to a formula
Calculating the accumulated contribution rate of the principal component variance; wherein PCN is the principal component variance cumulative contribution rate, λiIs the ith characteristic value, i is 1,2,3l, lambdajJ is 1,2,3n, l < n;
the sorting unit is used for sorting the eigenvalues in the eigenvalue set according to the sizes to obtain a sorted eigenvalue set;
a dimension reduction eigenvalue set determining unit, configured to select, according to the pivot variance cumulative contribution rate, the first l eigenvalues from the sorting eigenvalue set as a dimension reduction eigenvalue set;
and the dimensionality reduction feature vector set determining unit is used for selecting a feature vector corresponding to each feature value in the dimensionality reduction feature value set from the feature vector set to serve as the dimensionality reduction feature vector set.
Optionally, the real-time variance value calculation module specifically includes:
the first sliding time window processing unit is used for performing sliding time window processing on the effective real-time operation parameters to obtain a real-time operation data matrix;
the second sliding time window processing unit is used for performing sliding time window processing on the reference operation parameters to obtain a reference operation data matrix;
and the real-time divergence value determining unit is used for determining the real-time divergence value according to the real-time operation data matrix and the reference operation data matrix.
Optionally, the abnormality detecting module specifically includes:
the judging unit is used for judging whether the real-time divergence value is larger than the divergence value threshold value or not to obtain a judging result;
an abnormal parameter determining unit, configured to determine that the real-time operation parameter is abnormal if the determination result indicates that the real-time divergence value is greater than the divergence value threshold;
and the return unit is used for returning to the real-time operation parameter acquisition module if the judgment result shows that the real-time divergence value is smaller than or equal to the divergence value threshold value.
According to the specific embodiment provided by the invention, the invention discloses the following technical effects:
the invention provides a nuclear power device operation parameter anomaly detection method and system, which adopt a principal component analysis method to reduce the dimension of a feature vector set, select effective real-time operation parameters from real-time operation parameters according to the dimension-reduced feature vector set, remove interference information to obtain the effective real-time operation parameters, calculate a real-time divergence value according to the effective real-time operation parameters and reference operation parameters, and compare the real-time divergence value with a divergence value threshold value, thereby carrying out anomaly detection on the real-time operation parameters and improving the accuracy of anomaly detection.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings needed to be used in the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings without inventive exercise.
Fig. 1 is a flowchart of a method for detecting an abnormal operating parameter of a nuclear power plant according to an embodiment of the present invention;
fig. 2 is a schematic structural diagram of a nuclear power plant operation parameter abnormality detection system according to an embodiment of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The invention aims to provide a method and a system for detecting the abnormity of the running parameters of a nuclear power device, which can detect the abnormity of the running parameters in real time and improve the accuracy of abnormity detection.
In order to make the aforementioned objects, features and advantages of the present invention comprehensible, embodiments accompanied with figures are described in further detail below.
Fig. 1 is a flowchart of a method for detecting an abnormal operating parameter of a nuclear power plant according to an embodiment of the present invention, and as shown in fig. 1, the method for detecting an abnormal operating parameter of a nuclear power plant according to the present invention includes:
and S101, acquiring real-time operation parameters of the nuclear power device.
And S102, calculating a covariance matrix of the real-time operation parameters.
S102 specifically comprises the following steps:
step 201, normalizing the real-time operation parameters by using a dispersion normalization method to obtain normalized operation parameters.
Specifically, all data values of the real-time operating parameters are mapped between [0,1] by using a dispersion normalization method, and the conversion function is: and x is (x-min)/(max-min), wherein max is the maximum value of the real-time operation parameters, min is the minimum value of the real-time operation parameters, x is the real-time operation parameters, and x is the normalized operation parameters. By adopting the method, the influences of inconsistent dimension and overlarge and undersize data on the detection result can be avoided.
Step 202, mapping the normalized operation parameters by using a radial basis kernel function to obtain a kernel matrix. Specifically, the normalized operation parameters are mapped to a high-dimensional space through a kernel function, and the embodiment of the invention adopts a radial basis kernel function, and obtains mapped data phi (x), namely a kernel matrix K, after nonlinear mapping of the radial basis kernel function.
And 203, centralizing the kernel matrix to obtain a centralized kernel matrix.
The centering method of the kernel matrix K can be expressed as:
wherein the content of the first and second substances,
for centering the kernel matrix, K is the kernel matrix, InN is the number of rows or columns (equal number of rows and columns) of the identity matrix.
Step 204, calculating a covariance matrix of the centralized kernel matrix.
S103, calculating a characteristic value set and a characteristic vector set of the covariance matrix; the set of eigenvalues comprises each eigenvalue of the covariance matrix; the feature vector set comprises a feature vector corresponding to each feature value.
And S104, reducing the dimensions of the feature vector set by adopting a principal component analysis method according to the feature value set to obtain a dimension-reduced feature vector set.
S104 specifically comprises the following steps:
step 401, according to the formula
Calculating the accumulated contribution rate of the principal component variance; wherein PCN is the principal component variance cumulative contribution rate, λiI is 1,2,3.. l, λjN is the jth eigenvalue, j ═ 1,2,3.
And 402, sorting the eigenvalues in the eigenvalue set according to the size to obtain a sorted eigenvalue set. Sorted eigenvalues λ1≥λ2≥λ3≥...≥λnThe feature vector corresponding to each feature value is v1, v2n。
Step 403, according to the pivot variance cumulative contribution rate, selecting the first l eigenvalues from the sorting eigenvalue set as a dimension reduction eigenvalue set.
In order to achieve the purpose of reducing the dimension, the number of kernel principal elements, i.e. the number of feature vectors, needs to be determined according to the size of the feature values. If the number of the core principal elements is excessive, the proportion of the redundant information is still high, and finally the interference information is still excessive; however, if the number of selected kernel principal elements is too small, a lot of important information may be lost. In the embodiment of the invention, the number l of kernel principal elements in a high-dimensional space is determined by adopting the accumulated contribution rate of the principal element variance.
Order to
Thus determining the specific value of l, and then selecting the first l characteristic values, namely lambda, from the sorted characteristic value set1,λ2,λ3,...,λlWherein λ is1≥λ2≥λ3≥...≥λl。
Step 404, selecting a feature vector corresponding to each feature value in the dimension-reduced feature value set from the feature vector set as a dimension-reduced feature vector set. Specifically, will lambda1,λ2,λ3,...,λlThe corresponding feature vector is used as a dimension reduction feature vector set, namely v1,v2,...,vl。
And S105, selecting effective real-time operation parameters from the real-time operation parameters according to the dimensionality reduction feature vector set to obtain the effective real-time operation parameters.
After the dimension reduction feature vector set is determined, the nonlinear kernel principal component feature vector is obtained by calculating the principal component feature vector projection of the mapped data in the linear space:
obtaining kernel principal component feature vector t of real-time operation parameterkThen according to tkAnd selecting effective useful information from the real-time operation parameters, namely selecting the effective real-time operation parameters.
S106, acquiring reference operation parameters; the reference operation parameter is an operation parameter consistent with the real-time operation parameter working condition in historical normal operation parameters.
S107, calculating real-time divergence values of the effective real-time operation parameters and the reference operation parameters.
S107 specifically comprises the following steps:
and 701, performing sliding time window processing on the effective real-time operation parameters to obtain a real-time operation data matrix.
And 702, performing sliding time window processing on the reference operation parameters to obtain a reference operation data matrix.
And 703, determining a real-time divergence value according to the real-time operation data matrix and the reference operation data matrix.
Specifically, the effective real-time running parameters and the reference running parameters are respectively processed by continuous Sliding time windows, the time interval of the Sliding time windows is 1s, and the length of the Sliding time windows is Sliding _ length. Thus, the input data at each time is not an isolated operating parameter at a time, but a combination of data over a period of time.
Setting a fixed time length Total _ length, obtaining a real-time operation data matrix (with the dimension of Sliding _ length multiplied by l) after effective real-time operation parameters are subjected to Sliding time window processing within the time, obtaining a reference operation data matrix (with the dimension of Sliding _ length multiplied by l) after reference operation parameters are subjected to Sliding time window processing, and calculating Jensen-Shannon dispersion values of the two data matrices, wherein the calculation formula is as follows:
wherein, P1For the real-time operation of the probability density distribution of the data matrix, P2For reference to the probability density distribution of the operational data matrix, the calculation formula of the KL divergence is as follows:
wherein, P1(xi) For the real-time operation of the probability density distribution, P, of the ith data in the data matrix2(xi) Is the probability density distribution of the ith data in the reference operational data matrix.
S108, acquiring a divergence value threshold; the divergence value threshold is determined from the reference operating parameter.
And S109, carrying out abnormity detection on the real-time operation parameters according to the real-time divergence value and the divergence threshold value.
S109 specifically comprises:
step 901, judging whether the real-time divergence value is larger than the divergence value threshold value, and obtaining a judgment result.
And 902, if the judgment result shows that the real-time divergence value is larger than the divergence value threshold value, determining that the real-time operation parameter is abnormal, indicating that the system has a fault and needs to be diagnosed or isolated in time.
Step 903, if the judgment result shows that the real-time divergence value is smaller than or equal to the divergence value threshold, returning to the step 101, and continuously detecting the real-time operation parameters.
The divergence value threshold determination process in S108 specifically includes:
the method comprises the steps of collecting and storing historical normal operation data of the nuclear power device under different working conditions, and storing the data in a database in a classified manner (classifying the data according to different working conditions) to obtain data sets of different types. And then, processing S101-S107 on each data set to obtain a divergence value under each working condition, and setting a divergence threshold value of each working condition according to the divergence value under each working condition, so that the interference of random noise on abnormal monitoring can be avoided. The historical normal operation data includes pressure of a pressure stabilizer in a reactor coolant system, temperature of a fluctuation pipe, flow of a primary side outlet of a steam generator, temperature of an inlet and an outlet of a reactor core, water level of a secondary side of the steam generator, water supply temperature and water supply flow, steam yield and steam temperature, upper filling flow, lower discharging flow and water level of a volume control box of a chemical volume system and the like.
The present invention also provides a nuclear power plant operation parameter abnormality detection system, as shown in fig. 2, the nuclear power plant operation parameter abnormality detection system includes:
and the real-time operation parameter acquisition module 1 is used for acquiring real-time operation parameters of the nuclear power device.
And the covariance matrix calculation module 2 is used for calculating a covariance matrix of the real-time operation parameters.
An eigenvalue and eigenvector calculation module 3, configured to calculate an eigenvalue set and an eigenvector set of the covariance matrix; the set of eigenvalues comprises each eigenvalue of the covariance matrix; the feature vector set comprises a feature vector corresponding to each feature value.
And the dimensionality reduction module 4 is used for reducing dimensionality of the feature vector set by adopting a principal component analysis method according to the feature value set to obtain a dimensionality reduction feature vector set.
And the effective real-time operation parameter selection module 5 is used for selecting effective real-time operation parameters from the real-time operation parameters according to the dimensionality reduction feature vector set to obtain the effective real-time operation parameters.
A reference operation parameter obtaining module 6, configured to obtain a reference operation parameter; the reference operation parameter is an operation parameter which is consistent with the real-time operation parameter working condition in the historical normal operation parameters.
And the real-time dispersion value calculating module 7 is used for calculating real-time dispersion values of the effective real-time operation parameters and the reference operation parameters.
A divergence value threshold acquisition module 8, configured to acquire a divergence value threshold; the divergence value threshold is determined from the reference operating parameter.
And the anomaly detection module 9 is used for carrying out anomaly detection on the real-time operation parameters according to the real-time divergence value and the divergence threshold value.
Preferably, the covariance matrix calculation module 2 specifically includes:
and the normalization unit is used for normalizing the real-time operation parameters by adopting a dispersion normalization method to obtain normalized operation parameters.
And the mapping unit is used for mapping the normalized operation parameters by adopting a radial basis kernel function to obtain a kernel matrix.
And the centralization unit is used for centralizing the nuclear matrix to obtain a centralized nuclear matrix.
And the covariance matrix calculation unit is used for calculating the covariance matrix of the centralized kernel matrix.
Preferably, the dimension reduction module 4 specifically includes:
a principal component variance cumulative contribution rate calculation unit for calculating a principal component variance cumulative contribution rate according to a formula
Calculating the accumulated contribution rate of the principal component variance; wherein PCN is the principal component variance cumulative contribution rate, λiIs the ith characteristic value, i is 1,2,3l, lambdajJ is 1,2,3n, l < n.
And the sorting unit is used for sorting the eigenvalues in the eigenvalue set according to the sizes to obtain a sorted eigenvalue set.
And the dimension reduction characteristic value set determining unit is used for selecting the first l characteristic values from the sorting characteristic value set as the dimension reduction characteristic value set according to the pivot variance cumulative contribution rate.
And the dimensionality reduction feature vector set determining unit is used for selecting a feature vector corresponding to each feature value in the dimensionality reduction feature value set from the feature vector set to serve as the dimensionality reduction feature vector set.
Preferably, the real-time variance value calculation module 7 specifically includes:
and the first sliding time window processing unit is used for performing sliding time window processing on the effective real-time operation parameters to obtain a real-time operation data matrix.
And the second sliding time window processing unit is used for performing sliding time window processing on the reference operation parameters to obtain a reference operation data matrix.
And the real-time divergence value determining unit is used for determining the real-time divergence value according to the real-time operation data matrix and the reference operation data matrix.
Preferably, the abnormality detecting module 9 specifically includes:
and the judging unit is used for judging whether the real-time divergence value is larger than the divergence value threshold value or not to obtain a judgment result.
And the abnormal parameter determining unit is used for determining that the real-time operation parameter is abnormal if the judgment result shows that the real-time divergence value is larger than the divergence value threshold value.
And the returning unit is used for returning to the real-time operation parameter acquisition module 1 if the judgment result shows that the real-time divergence value is smaller than or equal to the divergence value threshold value.
The invention has the advantages that:
1. the method comprises the steps of extracting the characteristics of the operation parameters by combining the principle of kernel principal component analysis, firstly mapping the operation parameters to a high-dimensional approximate linear space through a kernel function, and avoiding the adverse effect of the nonlinear relation between data on the data characteristic extraction; by centralizing the kernel matrix, the data can be subjected to principal component analysis in a high-dimensional space; finally, the eigenvalue and the eigenvector can be obtained through eigenvalue decomposition, and an important basis is provided for the subsequent operation parameter anomaly detection.
2. The number of the core pivot elements is reasonably determined by the accumulated contribution rate of the core pivot elements. If the core principal metadata is artificially set, the value setting is too large, the proportion of redundant information is still high, and finally, the interference information is too much; but if its value is set too small, a lot of important information may be lost. And the number of the corresponding core principal elements when the cumulative contribution rate of the core principal elements exceeds 95 percent is selected, so that the maximization of the feature extraction and data dimension reduction effect can be ensured.
3. The phase space reconstruction is carried out on the operation data by adopting the sliding time window, and an isolated data point can be converted into a section of continuous time sequence data, so that the probability density distribution condition of the section of continuous time sequence data can be calculated, and the subsequent similarity measurement is convenient; meanwhile, through the continuous sliding time window, the continuity of data is guaranteed, and then the fact that an analysis result is given out in time after a fault occurs is guaranteed.
4. Noise interference level in historical normal operation data is calculated through a Jensen-Shannon divergence formula, a divergence value threshold value can be reasonably determined through calculation, and the problems of misdiagnosis and missed diagnosis caused by improper threshold value setting are solved.
5. Dimension reduction results are obtained for the real-time operation parameters and the reference operation parameters under the same mapping matrix, and the problem that low-dimensional data is not comparable due to inconsistent mapping relations is avoided.
6. The real-time operation parameters are processed by sliding time windows, so that the real-time operation parameters and the reference operation parameters can be converted under the same phase space reconstruction scale, and meanwhile, the real-time operation data can be analyzed and monitored continuously by processing the continuous sliding time windows (the sliding interval of the sliding time windows is 1 s).
7. The real-time divergence values of the data matrix of the real-time operation parameters and the data matrix of the reference operation parameters under the sliding time window are calculated, the similarity and the difference between the two data matrices can be measured according to the probability density distribution of the data, and the inaccuracy and the deviation caused by only numerically observing the data are avoided.
8. And comparing the real-time divergence value with a divergence value threshold, if the real-time divergence value exceeds the divergence value threshold, indicating that the real-time operation parameters are abnormal, and giving an alarm. The timeliness of the abnormity monitoring structure can be guaranteed, meanwhile, through comparison of the divergence value threshold value and the real-time divergence value, an abnormity detection result is visual and vivid, and operating personnel can be well supported to judge abnormity.
The embodiments in the present description are described in a progressive manner, each embodiment focuses on differences from other embodiments, and the same and similar parts among the embodiments are referred to each other. For the system disclosed by the embodiment, the description is relatively simple because the system corresponds to the method disclosed by the embodiment, and the relevant points can be referred to the method part for description.
The principles and embodiments of the present invention have been described herein using specific examples, which are provided only to help understand the method and the core concept of the present invention; meanwhile, for a person skilled in the art, according to the idea of the present invention, the specific embodiments and the application range may be changed. In view of the above, the present disclosure should not be construed as limiting the invention.
Claims (6)
1. A nuclear power plant operation parameter anomaly detection method is characterized by comprising the following steps:
acquiring real-time operation parameters of a nuclear power plant;
calculating a covariance matrix of the real-time operating parameters;
calculating a feature value set and a feature vector set of the covariance matrix; the set of eigenvalues comprises each eigenvalue of the covariance matrix; the feature vector set comprises a feature vector corresponding to each feature value;
reducing the dimension of the feature vector set by adopting a principal component analysis method according to the feature value set to obtain a dimension-reduced feature vector set;
according to the dimension reduction characteristic vector set, effective real-time operation parameters are selected from the real-time operation parameters to obtain effective real-time operation parameters;
acquiring a reference operation parameter; the reference operation parameter is an operation parameter consistent with the real-time operation parameter working condition in historical normal operation parameters;
calculating real-time divergence values of the effective real-time operating parameters and the reference operating parameters;
acquiring a divergence value threshold; the divergence value threshold is determined from the reference operating parameter;
performing anomaly detection on the real-time operation parameters according to the real-time divergence value and the divergence value threshold value
The calculating the covariance matrix of the real-time operation parameters specifically includes:
normalizing the real-time operation parameters by adopting a dispersion normalization method to obtain normalized operation parameters;
mapping the normalized operation parameters by adopting a radial basis kernel function to obtain a kernel matrix;
centralizing the kernel matrix to obtain a centralized kernel matrix;
calculating a covariance matrix of the centralized kernel matrix;
reducing the dimensions of the feature vector set by adopting a principal component analysis method according to the feature value set to obtain a dimension-reduced feature vector set, which specifically comprises the following steps:
according to the formula
Calculating the accumulated contribution rate of the principal component variance; wherein PCN is the principal component variance cumulative contribution rate, λiI is 1,2,3.. l, λjN is the jth characteristic value, j is 1,2,3.. n, and l is less than n;
sorting the eigenvalues in the eigenvalue set according to the size to obtain a sorted eigenvalue set;
according to the accumulated contribution rate of the pivot variance, selecting the first l eigenvalues from the sorting eigenvalue set as a dimension reduction eigenvalue set;
and selecting a feature vector corresponding to each feature value in the dimensionality reduction feature value set from the feature vector set as a dimensionality reduction feature vector set.
2. The method for detecting an abnormality in an operating parameter of a nuclear power plant according to claim 1, wherein the calculating a real-time variance value of the effective real-time operating parameter and the reference operating parameter specifically includes:
performing sliding time window processing on the effective real-time operation parameters to obtain a real-time operation data matrix;
performing sliding time window processing on the reference operation parameters to obtain a reference operation data matrix;
and determining a real-time divergence value according to the real-time operation data matrix and the reference operation data matrix.
3. The method for detecting the abnormality of the operating parameters of the nuclear power plant according to claim 1, wherein the detecting the abnormality of the real-time operating parameters according to the real-time divergence value and the divergence threshold value specifically includes:
judging whether the real-time divergence value is larger than the divergence value threshold value or not to obtain a judgment result;
if the judgment result shows that the real-time divergence value is larger than the divergence value threshold value, determining that the real-time operation parameter is abnormal;
and if the judgment result shows that the real-time divergence value is smaller than or equal to the divergence value threshold value, returning to the step of acquiring the real-time operation parameters of the nuclear power device.
4. A nuclear power plant operating parameter anomaly detection system, comprising:
the real-time operation parameter acquisition module is used for acquiring real-time operation parameters of the nuclear power device;
the covariance matrix calculation module is used for calculating a covariance matrix of the real-time operation parameters;
the eigenvalue and eigenvector calculation module is used for calculating an eigenvalue set and an eigenvector set of the covariance matrix; the set of eigenvalues comprises each eigenvalue of the covariance matrix; the feature vector set comprises a feature vector corresponding to each feature value;
the dimensionality reduction module is used for reducing dimensionality of the feature vector set by adopting a principal component analysis method according to the feature value set to obtain a dimensionality reduction feature vector set;
the effective real-time operation parameter selection module is used for selecting effective real-time operation parameters from the real-time operation parameters according to the dimensionality reduction feature vector set to obtain effective real-time operation parameters;
the reference operation parameter acquisition module is used for acquiring reference operation parameters; the reference operation parameter is an operation parameter consistent with the real-time operation parameter working condition in historical normal operation parameters;
the real-time divergence value calculation module is used for calculating real-time divergence values of the effective real-time operation parameters and the reference operation parameters;
the divergence value threshold value acquisition module is used for acquiring a divergence value threshold value; the divergence value threshold is determined from the reference operating parameter;
the anomaly detection module is used for carrying out anomaly detection on the real-time operation parameters according to the real-time divergence value and the divergence threshold value;
the covariance matrix calculation module specifically includes:
the normalization unit is used for normalizing the real-time operation parameters by adopting a dispersion normalization method to obtain normalized operation parameters;
the mapping unit is used for mapping the normalized operation parameters by adopting a radial basis kernel function to obtain a kernel matrix;
the centralization unit is used for centralizing the nuclear matrix to obtain a centralized nuclear matrix;
a covariance matrix calculation unit for calculating a covariance matrix of the centralized kernel matrix;
the dimension reduction module specifically comprises:
a principal component variance cumulative contribution rate calculation unit for calculating a principal component variance cumulative contribution rate according to a formula
Calculating the accumulated contribution rate of the principal component variance; wherein PCN is the principal component variance cumulative contribution rate, λiI is 1,2,3.. l, λjN is the jth characteristic value, j is 1,2,3.. n, and l is less than n;
the sorting unit is used for sorting the eigenvalues in the eigenvalue set according to the sizes to obtain a sorted eigenvalue set;
a dimension reduction eigenvalue set determining unit, configured to select, according to the pivot variance cumulative contribution rate, the first l eigenvalues from the sorting eigenvalue set as a dimension reduction eigenvalue set;
and the dimensionality reduction feature vector set determining unit is used for selecting a feature vector corresponding to each feature value in the dimensionality reduction feature value set from the feature vector set to serve as the dimensionality reduction feature vector set.
5. The nuclear power plant operating parameter anomaly detection system according to claim 4, wherein the real-time divergence value calculation module specifically includes:
the first sliding time window processing unit is used for performing sliding time window processing on the effective real-time operation parameters to obtain a real-time operation data matrix;
the second sliding time window processing unit is used for performing sliding time window processing on the reference operation parameters to obtain a reference operation data matrix;
and the real-time divergence value determining unit is used for determining the real-time divergence value according to the real-time operation data matrix and the reference operation data matrix.
6. The nuclear power plant operating parameter anomaly detection system as recited in claim 4, wherein the anomaly detection module specifically comprises:
the judging unit is used for judging whether the real-time divergence value is larger than the divergence value threshold value or not to obtain a judging result;
an abnormal parameter determining unit, configured to determine that the real-time operation parameter is abnormal if the determination result indicates that the real-time divergence value is greater than the divergence value threshold;
and the return unit is used for returning to the real-time operation parameter acquisition module if the judgment result shows that the real-time divergence value is smaller than or equal to the divergence value threshold value.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010654715.4A CN111797533B (en) | 2020-07-09 | 2020-07-09 | Nuclear power device operation parameter abnormity detection method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010654715.4A CN111797533B (en) | 2020-07-09 | 2020-07-09 | Nuclear power device operation parameter abnormity detection method and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111797533A CN111797533A (en) | 2020-10-20 |
| CN111797533B true CN111797533B (en) | 2022-05-13 |
Family
ID=72811388
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010654715.4A Active CN111797533B (en) | 2020-07-09 | 2020-07-09 | Nuclear power device operation parameter abnormity detection method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111797533B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112666918B (en) * | 2020-12-01 | 2022-06-14 | 沈阳化工大学 | Method for self-adaptive industrial process fault detection based on online compressed KECA |
| CN113642105B (en) * | 2021-08-03 | 2023-10-27 | 中国船舶重工集团公司第七一九研究所 | Multi-scale model construction method and device for ship power system and electronic equipment |
| CN116522088B (en) * | 2023-04-27 | 2024-01-26 | 哈尔滨工程大学 | Nuclear power plant operation data analysis method and system based on machine learning |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101470426A (en) * | 2007-12-27 | 2009-07-01 | 北京北方微电子基地设备工艺研究中心有限责任公司 | Fault detection method and system |
| CN105137957A (en) * | 2015-06-23 | 2015-12-09 | 黄红林 | Control method of nuclear power steam generator |
| CN107316057A (en) * | 2017-06-07 | 2017-11-03 | 哈尔滨工程大学 | Based on the nuclear power unit method for diagnosing faults being locally linear embedding into K nearest neighbor classifiers |
| CN109902564A (en) * | 2019-01-17 | 2019-06-18 | 杭州电子科技大学 | A kind of accident detection method based on the sparse autoencoder network of structural similarity |
| CN110738274A (en) * | 2019-10-26 | 2020-01-31 | 哈尔滨工程大学 | nuclear power device fault diagnosis method based on data driving |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20090306943A1 (en) * | 2006-10-05 | 2009-12-10 | North Carolina State University | Methods, systems and computer program products for reduced order model adaptive simulation of complex systems |
-
2020
- 2020-07-09 CN CN202010654715.4A patent/CN111797533B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101470426A (en) * | 2007-12-27 | 2009-07-01 | 北京北方微电子基地设备工艺研究中心有限责任公司 | Fault detection method and system |
| CN105137957A (en) * | 2015-06-23 | 2015-12-09 | 黄红林 | Control method of nuclear power steam generator |
| CN107316057A (en) * | 2017-06-07 | 2017-11-03 | 哈尔滨工程大学 | Based on the nuclear power unit method for diagnosing faults being locally linear embedding into K nearest neighbor classifiers |
| CN109902564A (en) * | 2019-01-17 | 2019-06-18 | 杭州电子科技大学 | A kind of accident detection method based on the sparse autoencoder network of structural similarity |
| CN110738274A (en) * | 2019-10-26 | 2020-01-31 | 哈尔滨工程大学 | nuclear power device fault diagnosis method based on data driving |
Non-Patent Citations (2)
| Title |
|---|
| A channelized hotelling observer for treaty-verification tasks;Christopher J. MacGahan等;《2015 IEEE Nuclear Science Symposium and Medical Imaging Conference (NSS/MIC)》;20161006;第1-1页 * |
| 基于贝叶斯的大数据异常值检测模型研究;周梁琦等;《脑知识与技术》;20200106;第207-209页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111797533A (en) | 2020-10-20 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111797533B (en) | Nuclear power device operation parameter abnormity detection method and system | |
| CN108805202B (en) | Machine learning method for electrolytic bath fault early warning and application thereof | |
| CN111768082A (en) | Power equipment state evaluation method based on big data analysis | |
| CN113255848B (en) | Water turbine cavitation sound signal identification method based on big data learning | |
| CN110738274A (en) | nuclear power device fault diagnosis method based on data driving | |
| CN108399277B (en) | Bridge damage identification method based on temperature and strain correlation | |
| CN110348150B (en) | Fault detection method based on correlation probability model | |
| CN108665119B (en) | Water supply pipe network abnormal working condition early warning method | |
| CN105259895A (en) | Method and monitoring system for detecting and separating micro fault in industrial process | |
| CN111143173A (en) | Server fault monitoring method and system based on neural network | |
| CN103103570B (en) | Based on the aluminium cell condition diagnostic method of pivot similarity measure | |
| CN110687895B (en) | Chemical process fault detection method based on self-adaptive kernel principal component analysis | |
| CN116383636A (en) | Coal mill fault early warning method based on PCA and LSTM fusion algorithm | |
| CN114559819A (en) | Electric vehicle battery safety early warning method based on signal processing | |
| CN111767657B (en) | Nuclear power system fault diagnosis method and system | |
| CN112036087A (en) | Multi-strategy fused nuclear power key equipment fault diagnosis method and system | |
| WO2023273249A1 (en) | Tsvm-model-based abnormality detection method for automatic verification system of smart electricity meter | |
| CN115112978A (en) | Fault detection method for insulating ring main unit of power system | |
| CN112036496A (en) | Nuclear power device fault diagnosis method and system | |
| CN111507374A (en) | Power grid mass data anomaly detection method based on random matrix theory | |
| CN114971308A (en) | Transformer health condition assessment method based on multi-dimensional data analysis | |
| CN111913463B (en) | State monitoring method for chemical volume control system of nuclear power plant | |
| CN115345414A (en) | Method and system for evaluating information security of oil and gas pipeline industrial control network | |
| CN114112390A (en) | Early fault diagnosis method for nonlinear complex system | |
| CN113283375A (en) | Early fault real-time early warning method for mine hydraulic support system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |