CN117633504A - Optical fiber sensing evaluation method and device for state of oil immersed transformer - Google Patents
Optical fiber sensing evaluation method and device for state of oil immersed transformer Download PDFInfo
- Publication number
- CN117633504A CN117633504A CN202311417399.9A CN202311417399A CN117633504A CN 117633504 A CN117633504 A CN 117633504A CN 202311417399 A CN202311417399 A CN 202311417399A CN 117633504 A CN117633504 A CN 117633504A
- Authority
- CN
- China
- Prior art keywords
- parameters
- parameter
- self
- state
- change
- 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.)
- Granted
Links
- 239000013307 optical fiber Substances 0.000 title claims abstract description 46
- 238000011156 evaluation Methods 0.000 title claims abstract description 33
- 238000000034 method Methods 0.000 claims abstract description 57
- 230000001364 causal effect Effects 0.000 claims abstract description 40
- 230000008859 change Effects 0.000 claims abstract description 33
- 230000002159 abnormal effect Effects 0.000 claims abstract description 22
- 238000004458 analytical method Methods 0.000 claims abstract description 13
- 238000012300 Sequence Analysis Methods 0.000 claims abstract description 10
- 238000012544 monitoring process Methods 0.000 claims abstract description 9
- 230000008569 process Effects 0.000 claims description 27
- 230000006870 function Effects 0.000 claims description 21
- 230000001419 dependent effect Effects 0.000 claims description 19
- 238000012731 temporal analysis Methods 0.000 claims description 7
- 238000000700 time series analysis Methods 0.000 claims description 7
- 238000005311 autocorrelation function Methods 0.000 description 16
- 230000007423 decrease Effects 0.000 description 8
- 230000001934 delay Effects 0.000 description 8
- 230000005856 abnormality Effects 0.000 description 7
- 230000000694 effects Effects 0.000 description 6
- 238000010586 diagram Methods 0.000 description 5
- 238000004364 calculation method Methods 0.000 description 4
- 230000003993 interaction Effects 0.000 description 4
- 238000010606 normalization Methods 0.000 description 4
- 238000000513 principal component analysis Methods 0.000 description 4
- 238000012886 linear function Methods 0.000 description 3
- 230000005540 biological transmission Effects 0.000 description 2
- 238000004422 calculation algorithm Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 230000007774 longterm Effects 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 230000035945 sensitivity Effects 0.000 description 2
- 238000007619 statistical method Methods 0.000 description 2
- 230000009466 transformation Effects 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000001939 inductive effect Effects 0.000 description 1
Classifications
-
- 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
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Computational Biology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Probability & Statistics with Applications (AREA)
- Testing Electric Properties And Detecting Electric Faults (AREA)
Abstract
The application provides an optical fiber sensing evaluation method and device for the state of an oil immersed transformer, and belongs to the field of data processing. The method comprises the following steps: monitoring key parameters in the transformer in real time by adopting an optical fiber sensor, and determining the variation trend of each parameter by a time sequence analysis method; determining self-change parameters and factor change parameters by adopting a causal graph according to the change trend, and setting influence weights; extracting data features from the change parameters and the change parameters, and calculating transformer state deflection according to the extracted data features and the determined influence weights; the probability of an abnormal problem is evaluated based on the state bias. The method is helpful for knowing the causal relationship among the parameters, setting the influence weight for the self-changing parameters and the parameters due to the change, and improving the analysis accuracy.
Description
Technical Field
The application relates to the field of state evaluation of oil immersed transformers, in particular to an optical fiber sensing evaluation method and device for the state of an oil immersed transformer.
Background
In the field of optical fiber sensing evaluation of the state of an oil immersed transformer, although the prior art can monitor and evaluate the state of the transformer to a certain extent, the following key problems still exist to be solved:
the data volume generated in the running process of the transformer is huge, and how to extract valuable information from massive data and determine the change trend and causal relation of parameters is a problem to be solved, so that the accuracy of transformer state evaluation is improved.
Disclosure of Invention
The purpose of the application is to overcome the problems existing in the prior art and provide an optical fiber sensing evaluation method and device for the state of an oil immersed transformer.
The application provides an optical fiber sensing evaluation method for the state of an oil immersed transformer, which comprises the following steps:
monitoring various key parameters in the transformer in real time by adopting an optical fiber sensor;
determining the variation trend of key parameters of each type by a time sequence analysis method:
according to the variation trend, determining self-variation parameters and factor variation parameters in various types of key parameters by adopting a causal graph;
setting an influence weight for the self-variation parameter, and setting the influence weight for the self-variation parameter according to the association degree of the self-variation parameter and the self-variation parameter;
extracting the data characteristics of the self-variation parameters and the factor-variation parameters;
and calculating transformer state bias of the abnormal characteristics according to the extracted data characteristics and the determined influence weights, wherein the expression is as follows:
bias=∑_iweight_i*feature_i;
wherein bias is the transformer state bias, weight_i is the influence weight of the ith parameter, and feature_i is the data characteristic of the ith parameter;
and evaluating the probability of the abnormal problem according to the state deviation.
Optionally, the change trend of each type of key parameter is determined by a time series analysis method, and the expression is as follows:
where y _ t denotes the parameter value at time t,is an autoregressive coefficient, p is the order of the model, and ε_t is the error term.
Optionally, the causal graph has the following expression:
P(Y|do(X))=∑_ZP(Y|X,Z)P(Z);
wherein Y represents a result variable, X represents a process variable, and Z represents a set of confounding variables;
identifying from the result variable and the process variable: the parameters without the parent node are independent variable parameters, and the parameters with the parent node are dependent variable parameters.
Optionally, according to the association degree between the dependent variable parameter and the self-variable parameter, the influence weight set for the dependent variable parameter is expressed as follows:
weight=f(correlation);
where weight is the independent variable influence weight, corrivation is the degree of correlation between the change parameter and the self-change parameter, and the function f is determined by fitting the data.
The application also provides an optical fiber sensing evaluation device of oil immersed transformer state, including:
the acquisition module is used for monitoring various key parameters in the transformer in real time by adopting an optical fiber sensor;
the analysis module is used for determining the variation trend of each type of key parameter through a time sequence analysis method:
the parameter module is used for determining self-variation parameters and factor variation parameters in various types of key parameters by adopting a causal graph according to the variation trend;
the weight module is used for setting influence weights for the self-change parameters and setting the influence weights for the self-change parameters according to the association degree of the self-change parameters and the self-change parameters;
the characteristic module is used for extracting the data characteristics of the self-change parameters and the change-caused parameters;
the state module is used for calculating transformer state deflection of the abnormal characteristics according to the extracted data characteristics and the determined influence weights, and the expression is as follows:
bias=∑_iweight_i*feature_i;
wherein bias is the transformer state bias, weight_i is the influence weight of the ith parameter, and feature_i is the data characteristic of the ith parameter;
and the probability module is used for evaluating the probability of the abnormal problem according to the state deviation.
Optionally, the analysis module determines the variation trend of each type of key parameter through a time sequence analysis method, and the expression is as follows:
where y _ t denotes the parameter value at time t,is an autoregressive coefficient, p is the order of the model, and ε_t is the error term.
Optionally, the causal graph has the following expression:
P(Y|do(X))=∑_ZP(Y|X,Z)P(Z);
wherein Y represents a result variable, X represents a process variable, and Z represents a set of confounding variables;
identifying from the result variable and the process variable: the parameters without the parent node are independent variable parameters, and the parameters with the parent node are dependent variable parameters.
Optionally, the weight module sets an influence weight for the dependent variable according to the association degree of the dependent variable and the self-variable, and the expression is:
weight=f(correlation);
where weight is the independent variable influence weight, corrivation is the degree of correlation between the change parameter and the self-change parameter, and the function f is determined by fitting the data.
The application also provides an optical fiber sensing evaluation device for the state of the oil immersed transformer, which comprises:
a memory;
and the processor is used for reading the computer-implemented program of the optical fiber sensing evaluation method of the oil immersed transformer state stored in the memory and executing the computer-implemented program.
The application also provides a storage medium, which stores a computer execution program and is used for being called by a processor to execute the steps of the optical fiber sensing evaluation method for the state of the oil immersed transformer.
The application has the advantages and beneficial effects that:
the application provides an optical fiber sensing evaluation method for the state of an oil immersed transformer, which comprises the following steps: monitoring various key parameters in the transformer in real time by adopting an optical fiber sensor; determining the variation trend of key parameters of each type by a time sequence analysis method: according to the variation trend, determining self-variation parameters and factor variation parameters in various types of key parameters by adopting a causal graph; setting an influence weight for the self-variation parameter, and setting the influence weight for the self-variation parameter according to the association degree of the self-variation parameter and the self-variation parameter; extracting the data characteristics of the self-variation parameters and the factor-variation parameters; and calculating transformer state bias of the abnormal characteristics according to the extracted data characteristics and the determined influence weights, wherein the expression is as follows: bias= Σjweight_i, feature_i; wherein bias is the transformer state bias, weight_i is the influence weight of the ith parameter, and feature_i is the data characteristic of the ith parameter; and evaluating the probability of the abnormal problem according to the state deviation. According to the method, the change trend of each type of key parameters is determined through a time sequence analysis method, and the self-change parameters and the factor-change parameters in the various types of key parameters are determined through a causal graph according to the change trend. This helps to understand the causal relationship between the various parameters and better understand the operating conditions of the transformer. And then, setting influence weights for the self-change parameters and the change parameters, so that the influence degree of each parameter on the state of the transformer is better reflected, and the analysis accuracy is improved.
Drawings
Fig. 1 is a schematic diagram of an optical fiber sensing evaluation flow of the state of an oil immersed transformer in the present application.
Fig. 2 is a schematic diagram of an optical fiber sensing evaluation system for the state of an oil immersed transformer in the present application.
Fig. 3 is a schematic diagram of an optical fiber sensing evaluation device for the state of an oil immersed transformer in the present application.
Detailed Description
As shown in fig. 1, a method for evaluating optical fiber sensing of an oil immersed transformer state includes the steps of:
s101, monitoring various key parameters in the transformer in real time by adopting an optical fiber sensor.
As shown in fig. 2, the oil immersed transformer is monitored in real time through the optical fiber sensor, monitored data are transmitted to the optical fiber sensing evaluation equipment for analysis, and analysis results are displayed in the display equipment.
Optical fiber sensors are used to monitor various types of critical parameters inside the transformer in real time, including but not limited to temperature, pressure, humidity, voltage, current, oil level, etc. These parameters are important criteria for evaluating the state of the transformer. It should be noted that, according to the purpose of evaluating the state of the oil immersed transformer, the key parameters are selected differently, and those skilled in the art select according to the actual situation.
The optical fiber sensor has the advantages of high sensitivity, strong anti-interference capability, small transmission loss and the like, and is very suitable for real-time monitoring in severe environments. The data collected by the optical fiber sensor is transmitted to a data processing center for further analysis and processing.
S102, determining the variation trend of each type of key parameter through a time sequence analysis method.
For each type of the key parameters, an autoregressive model (AR model) is adopted for time series analysis. The autoregressive model is a linear model that includes representing the parameter values at the current time instant as a linear combination of parameter values at the past p time instants, plus an error term.
In this application, the autoregressive model is expressed as:
where y _ t denotes the parameter value at time t,is an autoregressive coefficient, p is the order of the model, and ε_t is the error term. And acquiring the variation trend of the parameters by fitting the model.
The order p of the autoregressive model is estimated by statistical methods, such as autocorrelation function (ACF) or partial autocorrelation function (PACF). After the order p is determined, the estimation of the parameter trend is performed using the least squares method (OL S) or other estimation method.
In particular, the autocorrelation function describes the correlation of a time series with itself at different time delays. For the AR model, ACF should be significantly reduced after p delays. Therefore, the value of p is selected by observing the ACF map. In general, the delay at which the ACF first decreases to near zero is selected as the order p of the model.
The partial autocorrelation function describes the partial correlation of a time series with itself at different time delays, i.e. the correlation between the current point in time and the past point in time after controlling the influence of other points in time. For the AR model, PACF should decrease significantly after p delays. Thus, the value of p is selected by observing the PACF diagram. In general, the delay for the first decrease of PACF to near zero is chosen as the order p of the model.
The variation trend of each type of key parameter is known through a time series analysis method, such as whether periodical variation, long-term trend or abnormal fluctuation exists or not. The information has important significance for state evaluation and fault early warning of the transformer. For example, if a trend of a certain critical parameter is found to deviate from the normal range, it is necessary to further check whether there is a potential failure or abnormality of the transformer.
S103, according to the change trend, determining self-change parameters and factor-change parameters in multiple types of key parameters by adopting a causal graph.
Causal relationships between key parameters are further analyzed using a causal graph to determine which parameters are self-changing and which are causal.
A causal graph is a graphical tool for representing causal relationships between variables. Which helps identify and understand interactions between different variables and the effect of these interactions on the resulting variables. In a causal graph, the result variables are typically represented as root nodes of the graph, while the process variables and the confusion variables are represented as other nodes. Arrows represent causal relationships between variables, i.e. pointing from dependent variables to result variables.
In this application, a causal graph is represented by a conditional probability distribution:
P(Y|do(X))=∑_ZP(Y|X,Z)P(Z);
wherein Y represents a result variable, X represents a process variable, and Z represents a set of confounding variables; p (y|do (X)) represents the probability distribution of the result Y given the process X, said probability distribution describing the nature of the occurrence of the result Y after the process X has been carried out; p (Y|X, Z) represents the probability distribution of the result Y given the process X and the confounding variable Z, said probability distribution describing the nature of the occurrence of the result Y under consideration of the common influence of the process X and the confounding variable Z; p (Z) represents a probability distribution of the confounding variable Z, the probability distribution describing the occurrence of the confounding variable Z; Σz represents the summation of all confounding variables Z, the significance of which is that the total probability of occurrence of the result Y after the execution of the process X is calculated taking into account the values of all confounding variables Z.
The above formula expresses the probability distribution of the result Y given the process X.
By constructing the causal graph, the self-changing parameters and the causal changing parameters are identified.
The self-changing parameter refers to a parameter that is not affected by other parameters, or that changes are independent. In the causal graph, the self-changing parameter is considered as a node without a parent node. The change-due parameter refers to a parameter affected by other parameters. In the causal graph, the change parameter is a node with a parent node, and the change is caused by the change of the parent node.
S104, setting an influence weight for the self-change parameter, and setting the influence weight for the self-change parameter according to the association degree of the self-change parameter and the self-change parameter.
Impact weights are set for the self-changing parameters and the dependent-changing parameters to reflect the extent of their impact on the outcome variable. The greater the weight, the greater the impact of the parameter on the outcome variable.
Specifically, for the self-varying parameters, fixed weights are set for them according to domain knowledge and experience. For example, if a certain self-changing parameter is a key determinant of the outcome variable, it is weighted more heavily.
For the factor change parameter, the weight thereof needs to be determined in consideration of the association degree between the factor change parameter and the self-change parameter. The higher the degree of correlation, the more the change parameter is affected by the self-changing parameter, and therefore the greater the effect on the resulting variable.
In this application, the expression is expressed as follows:
weight=f(correlation);
where weight is the impact weight and corrivation is the degree of association between the change parameter and the self-change parameter.
In this application, the higher the association degree, the higher the influence weight should be, so the f-function is a monotonically increasing function. If the influence of the degree of correlation is nonlinear, the f-function will be a nonlinear function.
For example, in one specific embodiment, the above expression is a function of:
f(correlation)=log(1+correlation);
where corridation represents the degree of association between the dependent variable parameter and the resulting variable. The function increases faster when corridation is smaller and slower when corridation is larger, reflecting that the effect of the variation parameters on the result variable is larger at the beginning, but gradually decreases as the degree of correlation increases.
In fact, since the cause of different anomaly problems is different when the anomaly occurs in the transformer, even if two identical parameters have correlation, the relationship is not exactly the same when different anomaly problems occur, resulting in the f-functions being also different. Therefore, the independent variable parameters and the dependent variable parameters need to be determined again, and the association degree is determined, and these steps are realized by computer technology, using a machine learning algorithm, or using a data fitting method, which are not described in detail herein.
Next, an influence weight is set for each of the cause change parameters.
For example, if the degree of correlation between temperature and the result variable is 0.8, the degree of correlation between pressure and the result variable is 0.6, the degree of correlation between humidity and the result variable is 0.4, the degree of correlation between voltage and the result variable is 0.7, the degree of correlation between current and the result variable is 0.5, and the degree of correlation between oil level and the result variable is 0.3. Then, the influence weight of each of the change-due parameters is calculated using the above function:
weight_temperature=log(1+0.8)=0.5878;
weight_pressure=log(1+0.6)=0.4055;
weight_humidity=log(1+0.4)=0.2877;
weight_voltage=log(1+0.7)=0.4817;
weight_current=log(1+0.5)=0.3979;
weight_oil_level=log(1+0.3)=0.2624;
s105, extracting the data characteristics of the self-change parameters and the factor-change parameters.
Data features are extracted from the self-varying parameters and the dependent varying parameters for subsequent analysis and processing. A data feature is some transformation or summary of the original parameters, reflecting some important properties or characteristics of the parameters.
In this application, the following formula is used:
feature=g(parameter)
where feature is a data feature and parameter is an original parameter value.
In a specific embodiment, for continuous parameters, the function g is used to obtain a mean, a variance, a maximum, a minimum, etc.; for discrete parameters, the function g is used to extract features using frequency distribution or histograms. In addition, some feature extraction methods such as Principal Component Analysis (PCA), wavelet transform, fourier transform, etc. are also employed to extract more complex features.
And then, carrying out normalization processing on the data characteristics. The normalization process normalizes the different types of data features to within the same reference range. The method is particularly carried out in a scaling mode.
S106, calculating transformer state bias of the abnormal characteristics according to the extracted data characteristics and the determined influence weight.
And calculating the transformer state bias of the abnormal characteristic according to the extracted data characteristic and the determined influence weight.
In the present application, the expression of the transformer state deflection is as follows:
bias=∑_iweight_i*feature_i;
the weights and features of all parameters are multiplied and then the products are added to obtain the state bias of the transformer.
The following parameters are assumed:
weight= [0.2,0.3,0.5] # this is the weight of each parameter;
features= [10,20,30] # this is the data characteristic of each parameter;
according to a given formula, bias is calculated as:
bias=0.2×10+0.3×20+0.5×30;
the calculation result is as follows: bias=23.
Therefore, the transformer state bias for the anomaly is: 23.
s107 evaluates the probability of an abnormal problem based on the state bias.
In the present application, the probability expression for evaluating the abnormal problem is as follows:
P(problem)=h(bias);
the bias value is mapped to the anomaly problem probability by a function h.
The present application provides a simple h-function, which is a linear function:
h(bias)=0.1×bias;
according to the given bias value, the abnormal problem probability is calculated as follows:
P(problem)=0.1×23;
the calculation result is as follows: p (problem) =2.3;
therefore, the probability of an anomaly problem is: 2.3. this means that there is a 2.3% probability that an anomaly will occur based on the current transformer state bias.
For another example, the present application also provides a simple h-function, which is a linear function:
h(bias)=bias/D;
wherein D is a transformer deflection value when an abnormality occurs in the transformer.
In practice, the h function is determined based on the independent parameters and the dependent parameter types, and the relationship of these variables to transformer anomalies. The person skilled in the art collects data in advance and determines the specific form of the h-function by deductive or inductive methods.
Further, the following parameters are assumed:
weight= [0.2,0.3,0.5] # this is the weight of each parameter;
features= [10,20,30] # this is the data characteristic of each parameter;
the first parameter is 0.2×10=2; the second parameter is 0.3×20=6; the third parameter is 0.5×30=16.
And sorting according to the parameters, determining key parameters causing the abnormality according to the sorting, and further presuming the occurrence reason of the abnormality according to the key parameters.
The application also provides an optical fiber sensing evaluation device for the state of the oil immersed transformer.
As shown in fig. 3, an optical fiber sensing evaluation device for the state of an oil immersed transformer includes:
the acquisition module 301 is configured to monitor various types of key parameters inside the transformer in real time by using an optical fiber sensor.
Optical fiber sensors are used to monitor various types of critical parameters inside the transformer in real time, including but not limited to temperature, pressure, humidity, voltage, current, oil level, etc. These parameters are important criteria for evaluating the state of the transformer. It should be noted that, according to the purpose of evaluating the state of the oil immersed transformer, the key parameters are selected differently, and those skilled in the art select according to the actual situation.
The optical fiber sensor has the advantages of high sensitivity, strong anti-interference capability, small transmission loss and the like, and is very suitable for real-time monitoring in severe environments. The data collected by the optical fiber sensor is transmitted to a data processing center for further analysis and processing.
An analysis module 302, configured to determine a trend of the change of each type of key parameter by using a time series analysis method.
For each type of the key parameters, an autoregressive model (AR model) is adopted for time series analysis. The autoregressive model is a linear model that includes representing the parameter values at the current time instant as a linear combination of parameter values at the past p time instants, plus an error term.
In this application, the autoregressive model is expressed as:
where y _ t denotes the parameter value at time t,is an autoregressive coefficient, p is the order of the model, and ε_t is the error term. And acquiring the variation trend of the parameters by fitting the model.
The order p of the autoregressive model is estimated by statistical methods, such as autocorrelation function (ACF) or partial autocorrelation function (PACF). After the order p is determined, the estimation of the parameter trend is performed using the least squares method (OL S) or other estimation method.
In particular, the autocorrelation function describes the correlation of a time series with itself at different time delays. For the AR model, ACF should be significantly reduced after p delays. Therefore, the value of p is selected by observing the ACF map. In general, the delay at which the ACF first decreases to near zero is selected as the order p of the model.
The partial autocorrelation function describes the partial correlation of a time series with itself at different time delays, i.e. the correlation between the current point in time and the past point in time after controlling the influence of other points in time. For the AR model, PACF should decrease significantly after p delays. Thus, the value of p is selected by observing the PACF diagram. In general, the delay for the first decrease of PACF to near zero is chosen as the order p of the model.
The variation trend of each type of key parameter is known through a time series analysis method, such as whether periodical variation, long-term trend or abnormal fluctuation exists or not. The information has important significance for state evaluation and fault early warning of the transformer. For example, if a trend of a certain critical parameter is found to deviate from the normal range, it is necessary to further check whether there is a potential failure or abnormality of the transformer.
And the parameter module 303 is configured to determine, according to the variation trend, a self-variation parameter and a factor-variation parameter of the multiple types of key parameters by using a causal graph.
Causal relationships between key parameters are further analyzed using a causal graph to determine which parameters are self-changing and which are causal.
A causal graph is a graphical tool for representing causal relationships between variables. Which helps identify and understand interactions between different variables and the effect of these interactions on the resulting variables. In a causal graph, the result variables are typically represented as root nodes of the graph, while the process variables and the confusion variables are represented as other nodes. Arrows represent causal relationships between variables, i.e. pointing from dependent variables to result variables.
In this application, a causal graph is represented by a conditional probability distribution:
P(Y|do(X))=∑_ZP(Y|X,Z)P(Z);
wherein Y represents a result variable, X represents a process variable, and Z represents a set of confounding variables; p (y|do (X)) represents the probability distribution of the result Y given the process X, said probability distribution describing the nature of the occurrence of the result Y after the process X has been carried out; p (Y|X, Z) represents the probability distribution of the result Y given the process X and the confounding variable Z, said probability distribution describing the nature of the occurrence of the result Y under consideration of the common influence of the process X and the confounding variable Z; p (Z) represents a probability distribution of the confounding variable Z, the probability distribution describing the occurrence of the confounding variable Z; Σz represents the summation of all confounding variables Z, the significance of which is that the total probability of occurrence of the result Y after the execution of the process X is calculated taking into account the values of all confounding variables Z.
The above formula expresses the probability distribution of the result Y given the process X.
By constructing the causal graph, the self-changing parameters and the causal changing parameters are identified.
The self-changing parameter refers to a parameter that is not affected by other parameters, or that changes are independent. In the causal graph, the self-changing parameter is considered as a node without a parent node. The change-due parameter refers to a parameter affected by other parameters. In the causal graph, the change parameter is a node with a parent node, and the change is caused by the change of the parent node.
The weight module 304 is configured to set an impact weight for the self-change parameter, and set an impact weight for the self-change parameter according to the association degree between the self-change parameter and the self-change parameter.
Impact weights are set for the self-changing parameters and the dependent-changing parameters to reflect the extent of their impact on the outcome variable. The greater the weight, the greater the impact of the parameter on the outcome variable.
Specifically, for the self-varying parameters, fixed weights are set for them according to domain knowledge and experience. For example, if a certain self-changing parameter is a key determinant of the outcome variable, it is weighted more heavily.
For the factor change parameter, the weight thereof needs to be determined in consideration of the association degree between the factor change parameter and the self-change parameter. The higher the degree of correlation, the more the change parameter is affected by the self-changing parameter, and therefore the greater the effect on the resulting variable.
In this application, the expression is expressed as follows:
weight=f(correlation);
where weight is the impact weight and corrivation is the degree of association between the change parameter and the self-change parameter.
In this application, the higher the association degree, the higher the influence weight should be, so the f-function is a monotonically increasing function. In this application, the f-function is a nonlinear function.
For example, in one specific embodiment, the above expression is a function of:
f(correlation)=log(1+correlation);
where corridation represents the degree of association between the dependent variable parameter and the resulting variable. The function increases faster when corridation is smaller and slower when corridation is larger, reflecting that the effect of the variation parameters on the result variable is larger at the beginning, but gradually decreases as the degree of correlation increases.
In fact, since the cause of different anomaly problems is different when the anomaly occurs in the transformer, even if two identical parameters have correlation, the relationship is not exactly the same when different anomaly problems occur, resulting in the f-functions being also different. Therefore, the independent variable parameters and the dependent variable parameters need to be determined again, and the association degree is determined, and these steps are realized by a computer technology, a machine learning algorithm or a data fitting method, which are not described in detail herein.
Next, an influence weight is set for each of the cause change parameters.
For example, if the degree of correlation between temperature and the result variable is 0.8, the degree of correlation between pressure and the result variable is 0.6, the degree of correlation between humidity and the result variable is 0.4, the degree of correlation between voltage and the result variable is 0.7, the degree of correlation between current and the result variable is 0.5, and the degree of correlation between oil level and the result variable is 0.3. Then, the influence weight of each of the change-due parameters is calculated using the above function:
weight_temperature=log(1+0.8)=0.5878;
weight_pressure=log(1+0.6)=0.4055;
weight_humidity=log(1+0.4)=0.2877;
weight_voltage=log(1+0.7)=0.4817;
weight_current=log(1+0.5)=0.3979;
weight_oil_level=log(1+0.3)=0.2624;
and the feature module 305 is used for extracting the data features of the self-variation parameters and the factor-variation parameters.
Data features are extracted from the self-varying parameters and the dependent varying parameters for subsequent analysis and processing. A data feature is some transformation or summary of the original parameters, reflecting some important properties or characteristics of the parameters.
In this application, the following formula is used:
feature=g(parameter)
where feature is a data feature and parameter is an original parameter value.
In a specific embodiment, for continuous parameters, the function g is used to obtain a mean, a variance, a maximum, a minimum, etc.; for discrete parameters, the function g is used to extract features using frequency distribution or histograms. In addition, some feature extraction methods such as Principal Component Analysis (PCA), wavelet transform, fourier transform, etc. are also employed to extract more complex features.
And then, carrying out normalization processing on the data characteristics. The normalization process normalizes the different types of data features to within the same reference range. The method is particularly carried out in a scaling mode.
A status module 306 for calculating transformer status bias for the anomaly signature based on the extracted data signature and the determined impact weight.
And calculating the transformer state bias of the abnormal characteristic according to the extracted data characteristic and the determined influence weight.
In the present application, the expression of the transformer state deflection is as follows:
bias=∑_iweight_i*feature_i;
the weights and features of all parameters are multiplied and then the products are added to obtain the state bias of the transformer.
The following parameters are assumed:
weight= [0.2,0.3,0.5] # this is the weight of each parameter;
features= [10,20,30] # this is the data characteristic of each parameter;
according to a given formula, bias is calculated as:
bias=0.2×10+0.3×20+0.5×30;
the calculation result is as follows: bias=23.
Therefore, the transformer state bias for the anomaly is: 23.
a probability module 307 for evaluating the probability of an abnormal problem based on the state bias.
In the present application, the probability expression for evaluating the abnormal problem is as follows:
P(problem)=h(bias);
the bias value is mapped to the anomaly problem probability by a function h.
The present application provides a simple h-function, which is a linear function:
h(bias)=0.1×bias;
according to the given bias value, the abnormal problem probability is calculated as follows:
P(problem)=0.1×23;
the calculation result is as follows: p (problem) =2.3;
therefore, the probability of an anomaly problem is: 2.3. this means that there is a 2.3% probability that an anomaly will occur based on the current transformer state bias.
Further, the following parameters are assumed:
weight= [0.2,0.3,0.5] # this is the weight of each parameter;
features= [10,20,30] # this is the data characteristic of each parameter;
the first parameter is 0.2×10=2; the second parameter is 0.3×20=6; the third parameter is 0.5×30=16.
And sorting according to the parameters, determining key parameters causing the abnormality according to the sorting, and further presuming the occurrence reason of the abnormality according to the key parameters.
The application also provides an optical fiber sensing evaluation device for the state of the oil immersed transformer, which comprises:
a memory;
and the processor is used for reading the computer-implemented program of the optical fiber sensing evaluation method of the oil immersed transformer state stored in the memory and executing the computer-implemented program.
The application also provides a storage medium, which stores a computer execution program and is used for being called by a processor to execute the steps of the optical fiber sensing evaluation method for the state of the oil immersed transformer.
Claims (10)
1. The optical fiber sensing evaluation method for the state of the oil immersed transformer is characterized by comprising the following steps of:
monitoring various key parameters in the transformer in real time by adopting an optical fiber sensor;
determining the variation trend of key parameters of each type by a time sequence analysis method:
according to the variation trend, determining self-variation parameters and factor variation parameters in various types of key parameters by adopting a causal graph;
setting an influence weight for the self-variation parameter, and setting the influence weight for the self-variation parameter according to the association degree of the self-variation parameter and the self-variation parameter;
extracting the data characteristics of the self-variation parameters and the factor-variation parameters;
and calculating transformer state bias of the abnormal characteristics according to the extracted data characteristics and the determined influence weights, wherein the expression is as follows:
bias=∑_iweight_i*feature_i;
wherein bias is the transformer state bias, weight_i is the influence weight of the ith parameter, and feature_i is the data characteristic of the ith parameter;
and evaluating the probability of the abnormal problem according to the state deviation.
2. The optical fiber sensing evaluation method of the state of the oil immersed transformer according to claim 1, wherein the change trend of each type of key parameter is determined by a time series analysis method, and the expression is as follows:
y_t=φ1y_{t-1}+φ2y_{t-2}+...+φp*y_{t-p}+ε_t;
where y_t represents the parameter value at time t, Φ1, Φ2,..phi.p is the autoregressive coefficient, p is the order of the model, ε_t is the error term.
3. The method for evaluating the optical fiber sensing of the state of an oil immersed transformer according to claim 1, wherein the expression of the causal graph is as follows:
P(Y|do(X))=∑_ZP(Y|X,Z)P(Z);
wherein Y represents a result variable, X represents a process variable, and Z represents a set of confounding variables;
identifying from the result variable and the process variable: the parameters without the parent node are independent variable parameters, and the parameters with the parent node are dependent variable parameters.
4. The method for evaluating the state of an oil immersed transformer according to claim 1, wherein the influence weight set for the dependent variable parameter according to the degree of association between the dependent variable parameter and the self-variable parameter is expressed as:
weight=f(correlation);
where weight is the independent variable influence weight, corrivation is the degree of correlation between the change parameter and the self-change parameter, and the function f is determined by fitting the data.
5. An optical fiber sensing evaluation device for the state of an oil immersed transformer, which is characterized by comprising:
the acquisition module is used for monitoring various key parameters in the transformer in real time by adopting an optical fiber sensor;
the analysis module is used for determining the variation trend of each type of key parameter through a time sequence analysis method:
the parameter module is used for determining self-variation parameters and factor variation parameters in various types of key parameters by adopting a causal graph according to the variation trend;
the weight module is used for setting influence weights for the self-change parameters and setting the influence weights for the self-change parameters according to the association degree of the self-change parameters and the self-change parameters;
the characteristic module is used for extracting the data characteristics of the self-change parameters and the change-caused parameters;
the state module is used for calculating transformer state deflection of the abnormal characteristics according to the extracted data characteristics and the determined influence weights, and the expression is as follows:
bias=∑_iweight_i*feature_i;
wherein bias is the transformer state bias, weight_i is the influence weight of the ith parameter, and feature_i is the data characteristic of the ith parameter;
and the probability module is used for evaluating the probability of the abnormal problem according to the state deviation.
6. The optical fiber sensing evaluation device for the state of the oil immersed transformer according to claim 5, wherein the analysis module determines the variation trend of each type of key parameter through a time sequence analysis method, and the expression is as follows:
y_t=φ1y_{t-1}+φ2y_{t-2}+...+φp*y_{t-p}+ε_t;
where y_t represents the parameter value at time t, Φ1, Φ2,..phi.p is the autoregressive coefficient, p is the order of the model, ε_t is the error term.
7. The device for evaluating the state of an oil immersed transformer according to claim 5, wherein the causal graph is expressed as follows:
P(Y|do(X))=∑_ZP(Y|X,Z)P(Z);
wherein Y represents a result variable, X represents a process variable, and Z represents a set of confounding variables;
identifying from the result variable and the process variable: the parameters without the parent node are independent variable parameters, and the parameters with the parent node are dependent variable parameters.
8. The optical fiber sensing evaluation device for the state of the oil immersed transformer according to claim 5, wherein the weight module sets an influence weight for the change-caused parameter according to the association degree of the change-caused parameter and the self-change parameter, and the expression is:
weight=f(correlation);
where weight is the independent variable influence weight, corrivation is the degree of correlation between the change parameter and the self-change parameter, and the function f is determined by fitting the data.
9. An optical fiber sensing evaluation device for the state of an oil immersed transformer, comprising:
a memory;
a processor for reading a computer-implemented program of the optical fiber sensing evaluation method of the oil immersed transformer state according to any one of claims 1 to 4 stored in a memory, and executing the program.
10. A storage medium storing a computer-executable program for causing a processor to invoke the steps of performing the optical fiber sensing evaluation method for the state of the oil immersed transformer according to any one of claims 1 to 4.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202311417399.9A CN117633504B (en) | 2023-10-30 | 2023-10-30 | Optical fiber sensing evaluation method and device for state of oil immersed transformer |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202311417399.9A CN117633504B (en) | 2023-10-30 | 2023-10-30 | Optical fiber sensing evaluation method and device for state of oil immersed transformer |
Publications (2)
Publication Number | Publication Date |
---|---|
CN117633504A true CN117633504A (en) | 2024-03-01 |
CN117633504B CN117633504B (en) | 2024-06-21 |
Family
ID=90034661
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202311417399.9A Active CN117633504B (en) | 2023-10-30 | 2023-10-30 | Optical fiber sensing evaluation method and device for state of oil immersed transformer |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN117633504B (en) |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20130046721A1 (en) * | 2011-08-19 | 2013-02-21 | International Business Machines Corporation | Change point detection in causal modeling |
US20210216681A1 (en) * | 2020-01-15 | 2021-07-15 | Technical Centre for Soil, Agricultural&Rural Ecology&Environment, Ministry of Ecology&Environment | Method for designing sve process parameters in petroleum-type polluted field |
CN114970362A (en) * | 2022-06-08 | 2022-08-30 | 中交机电工程局有限公司 | Power grid load scheduling prediction method and system under multi-energy structure |
CN115358110A (en) * | 2022-07-25 | 2022-11-18 | 国网江苏省电力有限公司淮安供电分公司 | Transformer fault detection system based on acoustic sensor array |
WO2023080379A1 (en) * | 2021-11-02 | 2023-05-11 | 주식회사 바스젠바이오 | Disease onset information generating apparatus based on time-dependent correlation using polygenic risk score and method therefor |
-
2023
- 2023-10-30 CN CN202311417399.9A patent/CN117633504B/en active Active
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20130046721A1 (en) * | 2011-08-19 | 2013-02-21 | International Business Machines Corporation | Change point detection in causal modeling |
US20210216681A1 (en) * | 2020-01-15 | 2021-07-15 | Technical Centre for Soil, Agricultural&Rural Ecology&Environment, Ministry of Ecology&Environment | Method for designing sve process parameters in petroleum-type polluted field |
WO2023080379A1 (en) * | 2021-11-02 | 2023-05-11 | 주식회사 바스젠바이오 | Disease onset information generating apparatus based on time-dependent correlation using polygenic risk score and method therefor |
CN114970362A (en) * | 2022-06-08 | 2022-08-30 | 中交机电工程局有限公司 | Power grid load scheduling prediction method and system under multi-energy structure |
CN115358110A (en) * | 2022-07-25 | 2022-11-18 | 国网江苏省电力有限公司淮安供电分公司 | Transformer fault detection system based on acoustic sensor array |
Also Published As
Publication number | Publication date |
---|---|
CN117633504B (en) | 2024-06-21 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Li et al. | Wiener-based remaining useful life prediction of rolling bearings using improved Kalman filtering and adaptive modification | |
US7565262B2 (en) | Bayesian sensor estimation for machine condition monitoring | |
CA2634470C (en) | Method and system for trend detection and analysis | |
Lin et al. | Nonlinear dynamic principal component analysis for on-line process monitoring and diagnosis | |
JP3651693B2 (en) | Plant monitoring diagnosis apparatus and method | |
KR20140058501A (en) | Monitoring system using kernel regression modeling with pattern sequences | |
KR20140041767A (en) | Monitoring method using kernel regression modeling with pattern sequences | |
CN112116198B (en) | Data-driven process industrial state perception network key node screening method | |
Aremu et al. | A Relative Entropy Weibull-SAX framework for health indices construction and health stage division in degradation modeling of multivariate time series asset data | |
CN108399434B (en) | Analysis and prediction method of high-dimensional time series data based on feature extraction | |
Zhong et al. | Multimode non‐Gaussian process monitoring based on local entropy independent component analysis | |
CN113269327A (en) | Flow anomaly prediction method based on machine learning | |
CN113537328A (en) | Rotary machine fault diagnosis method and device based on deep learning | |
CN112000081B (en) | Fault monitoring method and system based on multi-block information extraction and Mahalanobis distance | |
Lucke et al. | From oscillatory to non-oscillatory disturbances: A comparative review of root cause analysis methods | |
Zhang et al. | Self-organising neural networks for automated machinery monitoring systems | |
Mishra et al. | Hybrid models for rotating machinery diagnosis and prognosis: estimation of remaining useful life | |
CN117633504B (en) | Optical fiber sensing evaluation method and device for state of oil immersed transformer | |
CN117408422A (en) | Natural gas flowmeter performance evaluation method and system based on deep learning | |
KR20220040659A (en) | Anomaly detection system of time-series data | |
CN114674511B (en) | Bridge modal anomaly early warning method for eliminating time-varying environmental factor influence | |
CN116627116A (en) | Process industry fault positioning method and system and electronic equipment | |
Castagliola et al. | Evaluation of the statistical performance of a variable sampling interval R EWMA control chart | |
Kabiri et al. | A bayesian approach for recognition of control chart patterns | |
Febriansyah et al. | Outlier detection and decision tree for wireless sensor network fault diagnosis |
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 |