CN118228089A - Improved fault diagnosis method for dissolved gas in transformer oil - Google Patents
Improved fault diagnosis method for dissolved gas in transformer oil Download PDFInfo
- Publication number
- CN118228089A CN118228089A CN202410645446.3A CN202410645446A CN118228089A CN 118228089 A CN118228089 A CN 118228089A CN 202410645446 A CN202410645446 A CN 202410645446A CN 118228089 A CN118228089 A CN 118228089A
- Authority
- CN
- China
- Prior art keywords
- fault
- data
- gas
- fault diagnosis
- dissolved gas
- 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
- 238000000034 method Methods 0.000 title claims abstract description 157
- 238000003745 diagnosis Methods 0.000 title claims abstract description 114
- 239000007789 gas Substances 0.000 claims abstract description 145
- 238000009833 condensation Methods 0.000 claims abstract description 32
- 230000005494 condensation Effects 0.000 claims abstract description 32
- 238000000926 separation method Methods 0.000 claims abstract description 26
- 230000008569 process Effects 0.000 claims abstract description 23
- 238000011156 evaluation Methods 0.000 claims abstract description 21
- 238000013210 evaluation model Methods 0.000 claims abstract description 15
- 239000012159 carrier gas Substances 0.000 claims abstract description 5
- 239000000203 mixture Substances 0.000 claims description 26
- 238000004364 calculation method Methods 0.000 claims description 17
- 238000004422 calculation algorithm Methods 0.000 claims description 12
- 238000012545 processing Methods 0.000 claims description 8
- 238000012549 training Methods 0.000 claims description 6
- 238000010801 machine learning Methods 0.000 claims description 5
- 238000007781 pre-processing Methods 0.000 claims description 4
- 230000002159 abnormal effect Effects 0.000 claims description 3
- 230000015556 catabolic process Effects 0.000 claims description 3
- 238000004140 cleaning Methods 0.000 claims description 3
- 238000009413 insulation Methods 0.000 claims description 3
- 238000007637 random forest analysis Methods 0.000 claims description 3
- -1 content variation Substances 0.000 claims description 2
- 238000002790 cross-validation Methods 0.000 claims description 2
- 238000013021 overheating Methods 0.000 claims description 2
- 238000004458 analytical method Methods 0.000 description 15
- 238000013135 deep learning Methods 0.000 description 9
- 238000001514 detection method Methods 0.000 description 9
- 238000013528 artificial neural network Methods 0.000 description 6
- 230000002265 prevention Effects 0.000 description 6
- 238000013461 design Methods 0.000 description 4
- 230000008859 change Effects 0.000 description 3
- 238000012423 maintenance Methods 0.000 description 3
- KFZMGEQAYNKOFK-UHFFFAOYSA-N Isopropanol Chemical compound CC(C)O KFZMGEQAYNKOFK-UHFFFAOYSA-N 0.000 description 2
- 238000012790 confirmation Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000005457 optimization Methods 0.000 description 2
- 230000003449 preventive effect Effects 0.000 description 2
- 238000005070 sampling Methods 0.000 description 2
- 238000013459 approach Methods 0.000 description 1
- 238000004590 computer program Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000000354 decomposition reaction Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 239000004744 fabric Substances 0.000 description 1
- 230000036541 health Effects 0.000 description 1
- 239000011810 insulating material Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000000737 periodic effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 230000035945 sensitivity Effects 0.000 description 1
- 239000002904 solvent Substances 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000002123 temporal effect Effects 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 230000008646 thermal stress Effects 0.000 description 1
- 238000011269 treatment regimen Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N30/00—Investigating or analysing materials by separation into components using adsorption, absorption or similar phenomena or using ion-exchange, e.g. chromatography or field flow fractionation
- G01N30/02—Column chromatography
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
-
- 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
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2337—Non-hierarchical techniques using fuzzy logic, i.e. fuzzy clustering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Life Sciences & Earth Sciences (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Computation (AREA)
- Evolutionary Biology (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Artificial Intelligence (AREA)
- Software Systems (AREA)
- Fuzzy Systems (AREA)
- Mathematical Physics (AREA)
- Automation & Control Theory (AREA)
- Health & Medical Sciences (AREA)
- Chemical & Material Sciences (AREA)
- Analytical Chemistry (AREA)
- Biochemistry (AREA)
- General Health & Medical Sciences (AREA)
- Immunology (AREA)
- Pathology (AREA)
- Housings And Mounting Of Transformers (AREA)
Abstract
The invention discloses an improved fault diagnosis method for dissolved gas in transformer oil, which relates to the technical field of fault diagnosis for dissolved gas and comprises the following steps: extracting a transformer oil sample, and analyzing the transformer oil sample based on the combination of carrier gas and a gas chromatograph to obtain first data of dissolved gas in oil; constructing a fault evaluation model according to an analytic hierarchy process, inputting first data of dissolved gas into the fault evaluation model, and generating a fault evaluation coefficient; selecting an optimal fault diagnosis method based on the fault evaluation coefficient, performing first fault diagnosis, and classifying fault types; when the fault type diagnosed by the FCM clustering method is a slight fault, a clustering result is obtained, the clustering result is analyzed to obtain condensation and separation data, and a second fault diagnosis is carried out according to the condensation and separation data.
Description
Technical Field
The invention relates to the technical field of fault diagnosis of dissolved gas in oil, in particular to an improved fault diagnosis method of dissolved gas in transformer oil.
Background
The dissolved gas in the oil refers to various gases generated by decomposition of the oil and the insulating material due to electric and thermal stresses during operation of the transformer. These gases are dissolved in transformer oil, and by detecting the type and concentration of these gases, the operating state of the transformer and the type of potential faults can be determined.
The composition and concentration of dissolved gases in transformer oil can reflect the operating conditions of power transformers substantially. The fault diagnosis method based on the data can conduct fault analysis in the running process of the transformer, timely detect early faults and track the development trend of the faults, so that disastrous accidents are prevented. This approach helps to shift from periodic maintenance to state-based maintenance, increasing the operational maintenance level of the transformer. Therefore, research on a data-based transformer fault diagnosis method has important practical significance.
Fault diagnosis of dissolved gas in transformer oil is carried out on DGA data of analysis of the dissolved gas in the transformer oil through two technologies of FCM clustering and deep belief network DBN. Specifically, after preprocessing DGA data by using a 9 ratio method, fault type discrimination is respectively carried out by adopting an FCM clustering algorithm and a DBN deep neural network.
For example, publication number: the invention application of CN113988426A discloses an electric vehicle charging load prediction method and system based on FCM clustering and LSTM, comprising the steps of collecting daily load data of an electric vehicle and acquiring daily meteorological data in the time period; establishing an FCM (fuzzy c-means) similar day clustering model, and carrying out load similar day clustering; summing the daily load data of the same date in each similar load day corresponding to the charging time to obtain the total daily load data of each date of the class; training to obtain a trained LSTM neural network aiming at similar days of different loads; and respectively predicting the electric vehicle loads corresponding to each load similar day by adopting the improved LS (least squares) neural network, so as to obtain a total electric vehicle load predicted value.
For example, publication number: the invention of CN110796281B discloses a state parameter prediction method of a wind turbine based on an improved deep belief network, and particularly relates to a state parameter prediction method of a wind turbine based on an improved deep belief network.
In the process of realizing the technical scheme in the embodiment of the application, the inventor discovers that the above technology at least has the following technical problems:
FCM clustering has lower accuracy in processing complex and high-dimensional data, and DBN deep neural networks can accurately judge most fault types, but the FCM clustering does not perform as well as FCM in low-energy discharge faults, and the prior art lacks a preferred implementation method of a fault diagnosis method of dissolved gas in efficient transformer oil.
The present invention proposes a solution to the above-mentioned problems.
Disclosure of Invention
In order to overcome the above-mentioned drawbacks of the prior art, embodiments of the present invention provide an improved method for diagnosing faults of dissolved gas in transformer oil, which solves the problems set forth in the above-mentioned background art by diagnosing faults of dissolved gas in transformer oil.
In order to achieve the above purpose, the present invention provides the following technical solutions:
An improved fault diagnosis method for dissolved gas in transformer oil comprises the following steps:
Extracting a transformer oil sample, and analyzing the transformer oil sample based on a carrier gas combined with a gas chromatograph to obtain first data of dissolved gas in oil, wherein the first data of the dissolved gas in oil comprises gas composition data and condition index data; constructing a fault evaluation model according to an analytic hierarchy process, inputting first data of dissolved gas into the fault evaluation model, and generating a fault evaluation coefficient; selecting an optimal fault diagnosis method based on a fault evaluation coefficient, performing first fault diagnosis, and classifying fault types, wherein the fault diagnosis method comprises a fault diagnosis method of a deep belief network and a fault diagnosis method of FCM clustering; when the fault type diagnosed by the FCM clustering method is a slight fault, a clustering result is obtained, the clustering result is analyzed to obtain condensation separation data, and a second fault diagnosis is performed according to the condensation separation data.
In a preferred embodiment, the gas composition data comprises a gas complexity factor, and the condition index data comprises a data dimension impact factor and a historical fault impact factor.
In a preferred embodiment, the selecting an optimal fault diagnosis method based on the fault evaluation coefficient performs a first fault diagnosis, and classifies fault types specifically as follows: training the fault evaluation coefficient based on a machine learning algorithm to obtain an optimal fault diagnosis method, performing fault diagnosis according to the corresponding fault diagnosis method, and classifying fault types into slight faults, medium faults and serious faults according to the corresponding fault diagnosis method.
In a preferred embodiment, the obtaining the clustering result, analyzing the clustering result to obtain the condensation and separation data, and performing the second fault diagnosis according to the condensation and separation data specifically includes: when the FCM clustering method is diagnosed as an optimal fault diagnosis method and the fault type is a slight fault; performing condensation separation operation on samples in the cluster to obtain condensation separation data, and combining the condensation separation data with higher similarity into a condensation set; and calculating variance and mean value of the condensation separation data in the condensation collection, and performing secondary fault diagnosis on the transformer by using a fault diagnosis method of the deep belief network when the variance and the mean value are respectively larger than a preset mean value standard value and a preset variance standard value.
In a preferred embodiment, the specific method for obtaining the gas complexity factor is as follows: collecting concentration values of various gases in transformer oil, and establishing a set; preprocessing the concentration values of various gases in the transformer oil, including cleaning and removing abnormal values; calculating the relative proportion of each dissolved gas in the total dissolved gas; and calculating the relative proportion through the information entropy to obtain the gas complexity coefficient.
In a preferred embodiment, the specific method for obtaining the data dimension influence coefficient is as follows:
Acquiring data of dissolved gas in oil when a transformer fails, and extracting and processing characteristics of the data to obtain a first characteristic set; acquiring the occurrence ratio of the features; calculating and evaluating the weight of each feature based on the random forest model and the occurrence ratio of the feature; acquiring the average value among all the features; and calculating the square of the deviation of each feature according to the weight of each feature and the average value among the features, and adding the square of the deviation of each feature to obtain the data dimension influence coefficient.
In a preferred embodiment, the method for obtaining the historical fault influence coefficient is as follows:
Acquiring historical fault data and historical faults of a transformer; respectively establishing fault diagnosis models based on FCM clustering and a deep belief network; inputting historical fault data of the transformer, combining with the performance of the cross-validation fault evaluation model, and calculating the accuracy and recall rate of each model in the process of diagnosing different fault types; analyzing and calculating a historical fault influence coefficient type by combining the accuracy rate, the recall rate and the historical fault data of the transformer; the historical fault types include partial discharge, high temperature overheating, and insulation breakdown, and the fault data includes gas composition, content variation, and oil temperature variation for each fault type.
The invention discloses a method for diagnosing faults of dissolved gas in improved transformer oil, which has the technical effects and advantages that:
1. The present invention evaluates the complexity of gas composition by analyzing the gas composition data using gas complexity coefficients and processes complex and simple gas composition data in combination with deep learning methods such as DBN and conventional clustering methods such as FCM. And secondly, analyzing the complexity of the data and the influence on a fault diagnosis method by using the condition index data through the data dimension influence coefficient, and providing guidance for model selection and parameter optimization. Finally, the historical fault influence coefficient is analyzed, the influence of different fault types on the fault diagnosis method is evaluated, and guidance is provided for decision support, resource utilization and fault prevention. The technical means together form a comprehensive fault diagnosis framework, and provide important support for improving diagnosis precision, improving detection efficiency and guiding preventive measures.
2. According to the invention, a fault evaluation model is constructed by using an analytic hierarchy process, and a fault evaluation coefficient is generated by using dissolved gas data, data dimension influence coefficients and historical fault influence coefficients. These coefficients enable accurate selection of the most appropriate fault diagnosis methods, including deep belief networks and FCM clustering methods. The method has the advantages that various factors are comprehensively considered, and the optimal method is accurately selected through the machine learning algorithm, so that the diagnosis precision and the comprehensiveness are improved. In particular, for a slight fault, a secondary diagnosis is performed, variance and mean are calculated by using the condensed and separated data to further verify the fault type, and if necessary, the fault diagnosis method of the deep belief network is used for further confirmation, thereby improving the reliability and accuracy of the fault diagnosis.
Drawings
FIG. 1 is a schematic structural diagram of a method for diagnosing faults of dissolved gas in improved transformer oil.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
Example 1, fig. 1 shows an improved method for diagnosing faults of dissolved gas in transformer oil according to the present invention, comprising the steps of:
S1, extracting a transformer oil sample, and analyzing the transformer oil sample based on carrier gas combined with a gas chromatograph to obtain first data of dissolved gas in oil, wherein the first data of the dissolved gas in the oil comprises gas composition data and condition index data; s2, constructing a fault evaluation model according to an analytic hierarchy process, inputting first data of dissolved gas into the fault evaluation model, and generating a fault evaluation coefficient; s3, selecting an optimal fault diagnosis method based on the fault evaluation coefficient, performing first fault diagnosis, and classifying fault types; and S4, when the fault type diagnosed by the FCM clustering method is a slight fault, obtaining a clustering result, analyzing the clustering result to obtain condensation separation data, and performing secondary fault diagnosis according to the condensation separation data.
It should be noted that: the application of the 9 ratio method combined with FCM clustering in DGA fault diagnosis is a fault diagnosis method based on dissolved gas components in oil; according to the method, gas ratio features in transformer oil are extracted through a9 ratio method, then clustering analysis is carried out on the features through an FCM clustering algorithm, data points of similar features are classified into the same category, and therefore different fault types or operation states which possibly exist are identified. By the method, the faults of the transformer can be found and diagnosed in time, and effective support is provided for the stable operation of the power system; the FCM clustering algorithm is simple, has high calculation speed, is suitable for quick preliminary screening, and is suitable for field application with limited calculation resources; the FCM clustering algorithm has lower accuracy when processing complex and high-dimensional data, and particularly for fault types which are difficult to distinguish, the selection of the number of clusters has larger influence on the result and needs experience and attempt;
The application of the DBN deep neural network in analysis and fault detection of dissolved gas in transformer oil is to realize automatic detection and diagnosis of transformer faults by carrying out feature extraction and fault diagnosis model construction on the dissolved gas data in the pretreated transformer oil. The method utilizes the DBN model to automatically learn and extract advanced features in the data, and performs fault detection and diagnosis on the data of the dissolved gas in the new transformer oil through the trained model, thereby providing effective support for the stable operation of the power system; when the DBN deep neural network processes complex and high-dimensional data, more characteristic information can be extracted, the accuracy of fault diagnosis is improved, parameters are not required to be set manually, and the data characteristics are automatically learned through training; however, the training process is complex and requires a large amount of computing resources, is not suitable for environments with limited resources, requires a long time for training and model adjustment, and is not suitable for scenes with high real-time requirements.
Extracting a transformer oil sample, and analyzing the transformer oil sample based on a carrier gas combined with a gas chromatograph to obtain first data of dissolved gas in oil, wherein the first data of the dissolved gas in oil comprises gas composition data and condition index data;
specifically, when fault diagnosis of dissolved gas in relevant transformer oil is carried out, an oil sample is required to be extracted for analysis;
Before sampling, the transformer is required to be confirmed to have power failure and grounding, so that safety is ensured; the sampling port of the transformer is required to be wiped by a high-purity solvent (such as isopropanol) and clean cloth, so that an oil sample is prevented from being polluted;
in this embodiment, the gas composition data includes a gas complexity factor.
If the composition of dissolved gases in transformer oil is very complex, contains multiple gas components, and their interrelationships are not obvious, deep learning methods (such as DBN) may be more suitable because they can better capture the nonlinear relationship between complex data; if the composition of the dissolved gas is relatively simple and there is a significant linear or simple relationship, conventional clustering methods (such as FCM) may already be sufficiently accurate.
The fault diagnosis method for analyzing the dissolved gas in the transformer oil with optimal gas complexity coefficient has the following advantages:
Improving the diagnosis precision: the gas complexity coefficient can reflect the composition and the complexity of the dissolved gas in the oil, and the condition of the dissolved gas in the oil can be known more accurately by analyzing the gas complexity coefficient, so that the precision and the accuracy of fault diagnosis are improved;
and (3) guiding feature selection: according to the analysis result of the gas complexity coefficient, the characteristic selection process can be guided, and the gas characteristic with the most representation and distinction is selected for constructing a fault diagnosis model, so that the sensitivity and the stability of the model are improved;
optimizing model design: according to the analysis result of the gas complexity coefficient, the model design can be optimized, and a fault diagnosis method suitable for complex gas composition, such as a deep learning-based method, is selected, so that the adaptability and generalization capability of the model are improved;
and the fault detection efficiency is improved: by analyzing the gas complexity coefficient, the characteristics and the change rule of the dissolved gas in the oil can be better understood, so that a fault detection algorithm and a fault detection model are designed more effectively, and the efficiency and the accuracy of fault detection are improved;
The manual intervention is reduced: based on the analysis result of the gas complexity coefficient, an automatic fault diagnosis system can be established, manual intervention is reduced, operation cost is reduced, and diagnosis efficiency and reliability are improved.
Therefore, the analysis of the gas complexity coefficient can provide important reference for the selection and design of the fault diagnosis method of the dissolved gas in the optimal transformer oil, so that the accuracy, efficiency and automation degree of fault diagnosis are improved.
In this embodiment, the specific method for obtaining the gas complexity coefficient is as follows:
Collecting concentration values of various gases in transformer oil, and establishing a set; preprocessing the concentration values of various gases in the transformer oil, including cleaning and removing abnormal values; and calculating the relative proportion of each dissolved gas in the total dissolved gas, and calculating the relative proportion through information entropy to obtain the gas complexity coefficient.
Wherein the relative proportionsThe specific calculation formula of (2) is as follows:
the specific calculation formula of the gas complexity coefficient is as follows:
the final calculation formula of the gas complexity coefficient is as follows:
In the method, in the process of the invention, Is the gas complexity factor,/>Is the total number of gas species,/>Is the gas type label,/>Is the concentration of the i-th gas.
It should be noted that the calculated information entropy reflects the degree of diversity of the composition of the dissolved gas, and that the larger the information entropy is, the more complicated the composition of the dissolved gas is.
The calculation formula of the gas complexity coefficient shows that when the expression value of the gas complexity coefficient is larger, the composition of the dissolved gas is more complex and is more suitable for a fault diagnosis method of a deep belief network, otherwise, when the expression value of the gas complexity coefficient is smaller, the composition of the dissolved gas is simpler and is more suitable for a fault diagnosis method of FCM clustering.
The condition index data in this embodiment includes a data dimension influence coefficient and a history fault influence coefficient.
Deep learning methods are generally more advantageous for high dimensional and complex data, such as multiple gas ratios and other related features, because they can automatically extract and learn advanced features; if the data dimension is relatively low and the data features are simple, conventional clustering methods may be able to provide sufficiently accurate results and are easier to implement and interpret.
The method for diagnosing the faults of the dissolved gas in the transformer oil with optimal analysis by analyzing the data dimension influence coefficient has the advantages that:
Decision support: the influence of the data dimension on the fault diagnosis method is known, and important reference information can be provided for a decision maker; if the data dimension is higher, the deep learning method may be more suitable, while if the data dimension is lower, the traditional clustering method may be more suitable; such knowledge can help a decision maker select an appropriate method to process data and solve a problem;
And (3) resource utilization: deep learning algorithms typically require a large amount of data and computational resources to train the model, especially when processing high-dimensional and complex data; in contrast, traditional clustering methods are generally simpler, lower in computational cost, and easier to implement and deploy; by analyzing the data dimension influence coefficients, a proper method can be reasonably selected according to the resource conditions, so that the resources are more effectively utilized;
interpretation of results: conventional clustering methods generally produce intuitive and easily understood results that can clearly divide a data set and provide explicit fault diagnosis information; in contrast, deep learning methods may produce complex models and features, and the results may be relatively difficult to interpret; thus, traditional clustering methods may be easier to implement and interpret when the data dimensions are low and the data features are relatively simple;
time efficiency: conventional clustering methods generally have lower temporal complexity, and can provide faster computation speeds when processing smaller-scale datasets; whereas deep learning methods generally require more computation time to train the model, especially when dealing with large-scale data and high-dimensional features; by analyzing the data dimension influence coefficients, a method suitable for the data scale and time requirements can be selected, and the calculation efficiency is improved.
Therefore, analysis of the data dimension influence coefficient is helpful for selection of an optimal fault diagnosis method for dissolved gas in transformer oil, so that method selection is more reasonable and effective.
In this embodiment, a specific method for obtaining the data dimension influence coefficient is as follows:
Acquiring data of dissolved gas in oil when a transformer fails, and extracting and processing characteristics of the data to obtain a first characteristic set; acquiring the occurrence ratio of the features; calculating and evaluating the weight of each feature based on the random forest model and the occurrence ratio of the feature; acquiring the average value among all the features; and calculating the square of the deviation of each feature according to the weight of each feature and the average value among the features, and adding the square of the deviation of each feature to obtain the data dimension influence coefficient.
Wherein the weight of each featureThe specific calculation formula of (2) is as follows:
The specific calculation formula of the mean value between the features is as follows
The specific calculation formula of the data dimension influence coefficient is as follows:
In the method, in the process of the invention, For the appearance duty cycle of feature z,/>And/>Are all feature labels,/>As the difference between the feature c and the feature z,Is the characteristic mean value/>For the total number of features,/>The coefficients are affected for the data dimension.
The calculation formula of the data dimension influence coefficient shows that the larger the expression value of the data dimension influence coefficient is, the higher and complex the data dimension is, the more suitable for the fault diagnosis method of the deep belief network is, and on the contrary, the smaller the expression value of the data dimension influence coefficient is, the lower and simple the data dimension is, and the more suitable for the fault diagnosis method of the FCM cluster is.
Historical fault data refers to records and related information of dissolved gas faults in transformer oil that occurred over a period of time. Such data typically includes the time of occurrence of the fault, the type of fault, the specifics of the fault, the method of fault handling, and the associated environmental conditions, etc. The historical fault data records various fault conditions of the transformer in the operation process, and is an important basis for analyzing and evaluating the performance and health state of the transformer. Through analysis of the historical fault data, the characteristics and rules of different types of faults can be known, and references are provided for formulating effective fault prevention and treatment strategies
The method for diagnosing the dissolved gas faults in the transformer oil with optimal analysis by analyzing the historical fault influence coefficient has the advantages that:
And (3) establishing a guide model: through analysis of the influence coefficient of the historical faults, the characteristics of different fault types in the composition and content of dissolved gas in the transformer oil can be known; this helps to guide the establishment of the fault diagnosis model, selecting appropriate features and parameters;
Evaluating model performance: the performance of different fault diagnosis methods can be evaluated by using the historical fault influence coefficients; the historical fault data is substituted into the model for testing, so that the performance of various methods in practical application can be known, and a method with better performance is selected;
Optimizing model parameters: based on the analysis result of the historical fault influence coefficient, the parameters of the fault diagnosis model can be optimized; according to the characteristics of different fault types in the historical data, parameters such as a threshold value, weight and the like of the model are adjusted so as to improve the accuracy and stability of the model;
Improving the diagnosis accuracy: by analyzing the historical fault influence coefficients, the typical characteristics of different fault types can be identified, so that faults of dissolved gas in the transformer oil can be diagnosed more accurately; this helps to improve the accuracy and reliability of the fault diagnosis;
Guiding fault prevention: by knowing the distribution and rules of different fault types in the historical fault data, the fault prevention work in the running process of the transformer can be guided; measures are timely taken to prevent common fault types from happening, and reliability and stability of the transformer are improved;
Therefore, analyzing the historical fault influence coefficient is a key step of optimizing a fault diagnosis method of dissolved gas in transformer oil, and can guide the establishment of a model, evaluate performance, optimize parameters, improve diagnosis accuracy and guide fault prevention work, thereby improving the operation efficiency and reliability of the transformer.
In this embodiment, the method for obtaining the historical fault influence coefficient is as follows:
Obtaining historical fault data and historical fault types of the transformer, wherein the historical fault types comprise partial discharge, high-temperature overheat and insulation breakdown, and the fault data comprise gas composition, content change and oil temperature change of each fault type; respectively establishing fault diagnosis models based on FCM clustering and deep belief networks, inputting historical fault data of a transformer, combining the performance of the fault evaluation models with cross verification, and calculating the accuracy and recall rate of each model in the process of diagnosing different fault types; analyzing and calculating a historical fault influence coefficient by combining the accuracy rate, the recall rate and the historical fault data of the transformer;
The specific calculation formula of the historical fault influence coefficient is as follows:
In the method, in the process of the invention, For historic failure influence coefficient/>For the accuracy of the y-th fault FCM cluster,/>Recall rate for FCM cluster for the y-th fault,/>For accuracy in diagnosing the ith fault type for a deep belief network,Recall for deep belief network in diagnosing the ith fault type,/>For the failure type label, t is the total number of failure types.
The calculation formula of the data dimension influence coefficient shows that the larger the expression value of the historical fault influence coefficient is, the more suitable for the fault diagnosis method of the deep belief network is, and on the contrary, the smaller the expression value of the historical fault influence coefficient is, the more suitable for the fault diagnosis method of the FCM cluster is.
The present embodiment evaluates the complexity of gas composition by analyzing the gas composition data using gas complexity coefficients, and processes complex and simple gas composition data in combination with deep learning methods (e.g., DBN) and conventional clustering methods (e.g., FCM). And secondly, analyzing the complexity of the data and the influence on a fault diagnosis method by using the condition index data through the data dimension influence coefficient, and providing guidance for model selection and parameter optimization. Finally, the historical fault influence coefficient is analyzed, the influence of different fault types on the fault diagnosis method is evaluated, and guidance is provided for decision support, resource utilization and fault prevention. The technical means together form a comprehensive fault diagnosis framework, and provide important support for improving diagnosis precision, improving detection efficiency and guiding preventive measures.
In example 2, a fault evaluation model is constructed according to an analytic hierarchy process, and the first data of the dissolved gas is input into the fault evaluation model to generate a fault evaluation coefficient.
The specific calculation formula of the fault evaluation coefficient is as follows:
In the middle of For failure evaluation coefficient,/>Is a preset proportionality coefficient of a gas complexity coefficient,/>As a coefficient of complexity of the gas,Preset proportionality coefficient of data dimension influence coefficient,/>For data dimension influence coefficient,/>Preset proportionality coefficient of history fault influence coefficient,/>Is a historical fault impact coefficient.
And selecting an optimal fault diagnosis method based on the fault evaluation coefficient, performing first fault diagnosis, and classifying fault types.
Specifically, the fault diagnosis method comprises a fault diagnosis method of a deep belief network and a fault diagnosis method of FCM clustering, the fault evaluation coefficient is trained based on a machine learning algorithm to obtain an optimal fault diagnosis method, fault diagnosis is carried out according to the corresponding fault diagnosis method, and fault types are classified into slight faults, medium faults and serious faults according to the corresponding fault diagnosis method.
When the fault type diagnosed by the FCM clustering method is a slight fault, a clustering result is obtained, the clustering result is analyzed to obtain condensation separation data, and a second fault diagnosis is carried out according to the condensation separation data;
Specifically, when the FCM clustering method is diagnosed as the optimal fault diagnosis method and the fault type is a slight fault; performing condensation separation operation on samples in the cluster to obtain condensation separation data, combining the condensation separation data with higher similarity into a condensation set, performing variance and mean calculation on the condensation separation data in the condensation set, and performing secondary fault diagnosis on the transformer by using a fault diagnosis method of a deep belief network when the variance and the mean are respectively larger than a preset mean standard value and a preset variance standard value.
In the embodiment, a fault evaluation model is constructed by using an analytic hierarchy process, and a fault evaluation coefficient is generated by using dissolved gas data, a data dimension influence coefficient and a historical fault influence coefficient. These coefficients enable accurate selection of the most appropriate fault diagnosis methods, including deep belief networks and FCM clustering methods. The method has the advantages that various factors are comprehensively considered, and the optimal method is accurately selected through the machine learning algorithm, so that the diagnosis precision and the comprehensiveness are improved. In particular, for a slight fault, a secondary diagnosis is performed, variance and mean are calculated by using the condensed and separated data to further verify the fault type, and if necessary, the fault diagnosis method of the deep belief network is used for further confirmation, thereby improving the reliability and accuracy of the fault diagnosis.
The above formulas are all formulas with dimensions removed and numerical values calculated, the formulas are formulas with a large amount of data collected for software simulation to obtain the latest real situation, and preset parameters in the formulas are set by those skilled in the art according to the actual situation.
The above embodiments may be implemented in whole or in part by software, hardware, firmware, or any other combination. When implemented in software, the above-described embodiments may be implemented in whole or in part in the form of a computer program product.
Those of ordinary skill in the art will appreciate that the various illustrative modules and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware, or combinations of computer software and electronic hardware. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the solution. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present application.
In addition, each functional module in each embodiment of the present application may be integrated into one processing module, or each module may exist alone physically, or two or more modules may be integrated into one module.
The foregoing is merely illustrative of the present application, and the present application is not limited thereto, and any person skilled in the art will readily recognize that variations or substitutions are within the scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.
Finally: the foregoing description of the preferred embodiments of the invention is not intended to limit the invention to the precise form disclosed, and any such modifications, equivalents, and alternatives falling within the spirit and principles of the invention are intended to be included within the scope of the invention.
Claims (10)
1. An improved fault diagnosis method for dissolved gas in transformer oil is characterized by comprising the following steps:
obtaining a transformer oil sample, and analyzing the transformer oil sample based on a carrier gas combined with a gas chromatograph to obtain first data of dissolved gas in oil, wherein the first data of the dissolved gas in oil comprises gas composition data and condition index data;
Constructing a fault evaluation model according to an analytic hierarchy process, inputting first data of dissolved gas into the fault evaluation model, and generating a fault evaluation coefficient;
selecting an optimal fault diagnosis method based on a fault evaluation coefficient, performing first fault diagnosis, judging fault types, and classifying the fault types, wherein the fault diagnosis method comprises a fault diagnosis method of FCM clustering;
and when the fault type is a slight fault, acquiring and analyzing a clustering result to obtain condensation separation data, and performing a second fault diagnosis according to the condensation separation data.
2. The method of claim 1, wherein the gas composition data comprises gas complexity coefficients and the condition index data comprises data dimension impact coefficients and historical fault impact coefficients.
3. The method for diagnosing faults of dissolved gas in transformer oil according to claim 2, wherein the method for diagnosing faults optimally selected based on the fault evaluation coefficients performs a first fault diagnosis, judges the fault type, and classifies the fault type specifically as:
Training the fault evaluation coefficient based on a machine learning algorithm to obtain an optimal fault diagnosis method, performing fault diagnosis according to the corresponding fault diagnosis method, and classifying fault types into slight faults, medium faults and serious faults according to the corresponding fault diagnosis method.
4. The method for diagnosing faults of dissolved gas in transformer oil as claimed in claim 3, wherein the steps of obtaining and analyzing the clustering result to obtain the condensation and separation data, and performing the second fault diagnosis according to the condensation and separation data are specifically as follows:
when the optimal fault diagnosis method is diagnosed and the fault type is a slight fault, performing condensation and separation operation on samples in the cluster to obtain condensation and separation data, and combining the condensation and separation data with higher similarity into a condensation set;
And calculating variance and mean value of the condensation separation data in the condensation collection, and performing secondary fault diagnosis on the transformer by using a fault diagnosis method of the deep belief network when the variance and the mean value are respectively larger than a preset mean value standard value and a preset variance standard value.
5. The method for diagnosing a fault in dissolved gas in transformer oil as claimed in claim 4, wherein the specific calculation formula of the fault evaluation coefficient is as follows:
in the above, the ratio of/> For failure evaluation coefficient,/>Is a preset proportionality coefficient of a gas complexity coefficient,/>Is the gas complexity factor,/>Preset proportionality coefficient of data dimension influence coefficient,/>For data dimension influence coefficient,/>Preset proportionality coefficient of history fault influence coefficient,/>Is a historical fault impact coefficient.
6. The method for diagnosing faults of dissolved gas in transformer oil as claimed in claim 5, wherein the specific acquisition method of the gas complexity coefficients is as follows:
Collecting concentration values of various gases in transformer oil, and establishing a set;
Preprocessing the concentration values of various gases in the transformer oil, including cleaning and removing abnormal values;
calculating the relative proportion of each dissolved gas in the total dissolved gas;
and calculating the relative proportion through the information entropy to obtain the gas complexity coefficient.
7. The method for diagnosing faults of dissolved gas in transformer oil as claimed in claim 5, wherein the specific acquisition method of the data dimension influence coefficient is as follows:
Acquiring data of dissolved gas in oil when a transformer fails, and extracting and processing characteristics of the data to obtain a first characteristic set;
Acquiring the occurrence ratio of the features, and calculating and evaluating the weight of each feature based on the occurrence ratio of the random forest model combined features; acquiring the average value among all the features;
and calculating the square of the deviation of each feature according to the weight of each feature and the average value among the features, and adding the square of the deviation of each feature to obtain the data dimension influence coefficient.
8. The method for diagnosing a dissolved gas failure in transformer oil as recited in claim 5, wherein the method for obtaining the historical failure influence coefficient is as follows:
Acquiring historical fault data and historical faults of a transformer;
Respectively establishing fault diagnosis models based on FCM clustering and a deep belief network;
inputting historical fault data of the transformer, combining with the performance of the cross-validation fault evaluation model, and calculating the accuracy and recall rate of each model in the process of diagnosing different fault types;
Analyzing and calculating a historical fault influence coefficient type by combining the accuracy rate, the recall rate and the historical fault data of the transformer;
The historical fault types include partial discharge, high temperature overheating, and insulation breakdown, and the fault data includes gas composition, content variation, and oil temperature variation for each fault type.
9. The method for diagnosing a fault in dissolved gas in transformer oil as claimed in claim 6, wherein the formula of the gas complexity factor is as follows:
in the above, the ratio of/> Is the gas complexity factor,/>Is the total number of gas species,/>Is the gas type label,/>Is the concentration of the gas in the i.
10. The method for diagnosing a dissolved gas failure in transformer oil as recited in claim 7, wherein the specific calculation formula of the data dimension influence coefficient is as follows:
in the above, the ratio of/> For the appearance duty cycle of feature z,/>And/>Are all feature labels,/>For the difference between feature c and feature z,/>Is the characteristic mean value/>For the total number of features,/>The coefficients are affected for the data dimension.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410645446.3A CN118228089B (en) | 2024-05-23 | 2024-05-23 | Improved fault diagnosis method for dissolved gas in transformer oil |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410645446.3A CN118228089B (en) | 2024-05-23 | 2024-05-23 | Improved fault diagnosis method for dissolved gas in transformer oil |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN118228089A true CN118228089A (en) | 2024-06-21 |
| CN118228089B CN118228089B (en) | 2024-08-13 |
Family
ID=91498320
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410645446.3A Active CN118228089B (en) | 2024-05-23 | 2024-05-23 | Improved fault diagnosis method for dissolved gas in transformer oil |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN118228089B (en) |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110058103A (en) * | 2019-05-23 | 2019-07-26 | 国电南京自动化股份有限公司 | Intelligent transformer fault diagnosis system based on Vxworks platform |
| KR20200014129A (en) * | 2018-07-31 | 2020-02-10 | 오토시맨틱스 주식회사 | Diagnosis method of electric transformer using Deep Learning |
| CN112115638A (en) * | 2020-08-28 | 2020-12-22 | 合肥工业大学 | Transformer fault diagnosis method based on improved Adam algorithm optimization neural network |
| CN113887623A (en) * | 2021-09-30 | 2022-01-04 | 国网甘肃省电力公司检修公司 | IFCM-BB-based transformer fault diagnosis method |
| CN115329908A (en) * | 2022-10-14 | 2022-11-11 | 西安热工研究院有限公司 | Power transformer fault diagnosis method based on deep learning |
| WO2024007580A1 (en) * | 2022-07-07 | 2024-01-11 | 南京国电南自电网自动化有限公司 | Power equipment parallel fault diagnosis method and apparatus based on hybrid clustering |
| CN117972571A (en) * | 2023-11-09 | 2024-05-03 | 国网河南省电力公司三门峡供电公司 | Transformer fault diagnosis method based on dissolved gas in oil |
-
2024
- 2024-05-23 CN CN202410645446.3A patent/CN118228089B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR20200014129A (en) * | 2018-07-31 | 2020-02-10 | 오토시맨틱스 주식회사 | Diagnosis method of electric transformer using Deep Learning |
| CN110058103A (en) * | 2019-05-23 | 2019-07-26 | 国电南京自动化股份有限公司 | Intelligent transformer fault diagnosis system based on Vxworks platform |
| CN112115638A (en) * | 2020-08-28 | 2020-12-22 | 合肥工业大学 | Transformer fault diagnosis method based on improved Adam algorithm optimization neural network |
| CN113887623A (en) * | 2021-09-30 | 2022-01-04 | 国网甘肃省电力公司检修公司 | IFCM-BB-based transformer fault diagnosis method |
| WO2024007580A1 (en) * | 2022-07-07 | 2024-01-11 | 南京国电南自电网自动化有限公司 | Power equipment parallel fault diagnosis method and apparatus based on hybrid clustering |
| CN115329908A (en) * | 2022-10-14 | 2022-11-11 | 西安热工研究院有限公司 | Power transformer fault diagnosis method based on deep learning |
| CN117972571A (en) * | 2023-11-09 | 2024-05-03 | 国网河南省电力公司三门峡供电公司 | Transformer fault diagnosis method based on dissolved gas in oil |
Non-Patent Citations (3)
| Title |
|---|
| JIEJIE DAI, ET AL.: "Dissolved gas analysis of insulating oil for power transformer fault diagnosis with deep belief network", 《IEEE TRANSACTIONS ON DIELECTRICS AND ELECTRICAL INSULATION》, vol. 24, no. 5, 31 October 2017 (2017-10-31), XP011673624, DOI: 10.1109/TDEI.2017.006727 * |
| 代杰杰;宋辉;杨;陈玉峰;盛戈;江秀臣;: "基于油中气体分析的变压器故障诊断ReLU-DBN方法", 《电网技术》, no. 2, 5 February 2018 (2018-02-05) * |
| 滕文辉: "电力变压器油中溶解气体浓度预测与故障诊断的研究", 《中国优秀硕士学位论文全文数据库》, 15 May 2021 (2021-05-15) * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN118228089B (en) | 2024-08-13 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Hsu et al. | Wind turbine fault diagnosis and predictive maintenance through statistical process control and machine learning | |
| CN116559598B (en) | Smart distribution network fault positioning method and system | |
| CN116625438A (en) | Gas pipe network safety on-line monitoring system and method thereof | |
| CN114723285B (en) | Power grid equipment safety evaluation prediction method | |
| CN109389325B (en) | Method for evaluating state of electronic transformer of transformer substation based on wavelet neural network | |
| CN101738998B (en) | System and method for monitoring industrial process based on local discriminatory analysis | |
| CN112417766B (en) | Fault diagnosis method based on fault-free data | |
| CN117408162B (en) | Power grid fault control method based on digital twin | |
| CN110879586A (en) | Phase modulator fault diagnosis and state monitoring method and system | |
| CN114519382A (en) | Nuclear power plant key operation parameter extraction and abnormity monitoring method | |
| CN110968703B (en) | Method and system for constructing abnormal metering point knowledge base based on LSTM end-to-end extraction algorithm | |
| Renga et al. | Data-driven exploratory models of an electric distribution network for fault prediction and diagnosis | |
| CN117391458B (en) | Safety production risk detection and early warning method and system based on data analysis | |
| CN118193954A (en) | Power distribution network abnormal data detection method and system based on edge calculation | |
| CN114581699A (en) | Transformer state evaluation method based on deep learning model in consideration of multi-source information | |
| CN118015812A (en) | Equipment abnormality alarm method | |
| CN114597886A (en) | Power distribution network operation state evaluation method based on interval type two fuzzy clustering analysis | |
| CN113313365A (en) | Degradation early warning method and device for primary air fan | |
| CN118228089B (en) | Improved fault diagnosis method for dissolved gas in transformer oil | |
| KR20220089853A (en) | Method for Failure prediction and prognostics and health management of renewable energy generation facilities using machine learning technology | |
| CN111934903A (en) | Docker container fault intelligent prediction method based on time sequence evolution genes | |
| CN111506636A (en) | System and method for analyzing residential electricity consumption behavior based on autoregressive and neighbor algorithm | |
| CN111814325B (en) | Single-variable alarm threshold optimization method for non-stationary operation of coal-fired power generation equipment | |
| CN114553756A (en) | Equipment fault detection method based on joint generation countermeasure network and electronic equipment | |
| CN114112390A (en) | Early fault diagnosis method for nonlinear complex 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 |