CN117236022A - Training method and application method of residual life prediction model of transformer and electronic equipment - Google Patents
Training method and application method of residual life prediction model of transformer and electronic equipment Download PDFInfo
- Publication number
- CN117236022A CN117236022A CN202311208255.2A CN202311208255A CN117236022A CN 117236022 A CN117236022 A CN 117236022A CN 202311208255 A CN202311208255 A CN 202311208255A CN 117236022 A CN117236022 A CN 117236022A
- Authority
- CN
- China
- Prior art keywords
- transformer
- prediction model
- domain
- training
- residual life
- 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.)
- Pending
Links
- 238000012549 training Methods 0.000 title claims abstract description 83
- 238000000034 method Methods 0.000 title claims abstract description 55
- 238000007781 pre-processing Methods 0.000 claims abstract description 7
- 230000015556 catabolic process Effects 0.000 claims description 43
- 238000006731 degradation reaction Methods 0.000 claims description 43
- 238000004590 computer program Methods 0.000 claims description 8
- 230000006870 function Effects 0.000 claims description 8
- 238000012545 processing Methods 0.000 claims description 8
- 238000005457 optimization Methods 0.000 claims description 5
- 238000010606 normalization Methods 0.000 claims description 3
- 238000010586 diagram Methods 0.000 description 6
- 230000000694 effects Effects 0.000 description 3
- 238000011156 evaluation Methods 0.000 description 2
- 238000012423 maintenance Methods 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 239000003990 capacitor Substances 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 230000004069 differentiation Effects 0.000 description 1
- 239000004973 liquid crystal related substance Substances 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 239000013307 optical fiber Substances 0.000 description 1
- 230000003449 preventive effect Effects 0.000 description 1
- 239000004065 semiconductor Substances 0.000 description 1
- 238000012163 sequencing technique Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
Abstract
The invention relates to a training method, an application method, electronic equipment and a computer readable storage medium of a mutual inductor residual life prediction model, which comprise the following steps: acquiring multi-dimensional time sequence data and working condition data of a transformer, and preprocessing the multi-dimensional time sequence data to obtain a multi-dimensional time sequence training set; grouping the specific domains of the multi-dimensional time sequence training set to obtain a specific domain training set; constructing an initial transformer residual life prediction model; inputting the specific domain training set, extracting common features by a common feature extractor, extracting specific domain features by the specific domain feature extractor, obtaining specific domain predicted life by a specific domain life predictor, and performing iterative training to obtain a complete transformer residual life prediction model. In summary, the invention captures the domain invariant feature of the sample by extracting the common feature of the sample, captures the differential feature of the sample by extracting the specific domain feature of the sample, and realizes accurate prediction of the residual life of the capacitive voltage transformer under multiple working conditions.
Description
Technical Field
The invention relates to the field of electric power metering monitoring, in particular to a training method, an application method, electronic equipment and a computer readable storage medium for a residual life prediction model of a transformer.
Background
Capacitive voltage transformers (CVT, capacitor voltage transformer) are an important component of high-voltage electric energy metering systems, and are stable and reliable as a basis for ensuring accurate electric energy metering and fair electric energy trading. The CVT is accurately predicted for the residual life, so that maintenance personnel can monitor the running state of the transformer in time, the health condition of the transformer can be effectively estimated, and preventive maintenance strategies can be made in advance.
In the prior art, a deep learning-based method is generally adopted to predict the residual life of CVT equipment, but a good prediction effect can be obtained only on the basis that training data and data to be detected are in the same distribution. However, in an actual application scene, due to the fact that differences exist in operation working conditions of the CVT, the obtained working condition data also have differences in distribution, in the prior art, although the distribution differences of the characteristic data under different working conditions are reduced by learning the characteristics of constant cross-domain, it is difficult to extract enough domain-invariant characteristics, and the characteristic differences of different degradation stages are ignored, so that the prediction effect is poor, and the residual life of the CVT cannot be accurately predicted under the condition of multiple working conditions.
Disclosure of Invention
In view of the foregoing, it is necessary to provide a training method, an application method, an electronic device and a computer readable storage medium for a residual life prediction model of a transformer, which are used for solving the technical problems that in the prior art, the working condition difference of the transformer is difficult to extract enough domain invariant features, the differentiated features of different working conditions and different degradation stages are ignored, and the residual life of the CVT can not be accurately predicted under multiple working conditions.
In order to solve the above problems, in one aspect, the present invention provides a training method for a residual life prediction model of a transformer, including:
acquiring multi-dimensional time sequence data and working condition data of a transformer, and preprocessing the multi-dimensional time sequence data to obtain a multi-dimensional time sequence training set;
clustering the working condition data based on a clustering algorithm to obtain working condition groups, grouping the remaining life to obtain degradation stage groups, and performing specific domain grouping on the multi-dimensional time sequence training set according to the working condition groups and the degradation stage groups to obtain specific domain training sets;
constructing an initial transformer residual life prediction model, wherein the initial transformer residual life prediction model comprises a public feature extractor, a specific domain feature extractor and a specific domain life predictor;
inputting the specific domain training set into an initial transformer residual life prediction model, extracting common features based on a common feature extractor, extracting specific domain features based on the specific domain feature extractor, predicting the specific domain life based on the specific domain life predictor to obtain the specific domain predicted life, and iteratively training the initial transformer residual life prediction model to obtain the fully trained transformer residual life prediction model.
Further, acquiring multidimensional time sequence data of the transformer includes:
collecting index data of a transformer in a period of time;
dividing the time interval into a plurality of time subintervals, dividing each index data according to the time subintervals, and calculating the maximum value, the minimum value, the average value and the standard deviation of the index data in each time subinterval to obtain the multidimensional time sequence data of the transformer.
Further, preprocessing the multi-dimensional time sequence data to obtain a multi-dimensional time sequence training set, including:
the characteristic data is formed into a characteristic parameter sequence, and a corresponding time sequence is established according to the time subinterval corresponding to each data in the characteristic parameter sequence;
calculating the correlation index, the monotonicity index and the discreteness index of each feature according to the feature parameter sequence and the time sequence;
and combining the correlation index, the monotonicity index and the discrete index to obtain a comprehensive index, setting a comprehensive index threshold, and eliminating characteristic data of which the comprehensive index is lower than the comprehensive index threshold.
Further, clustering processing is performed on the working condition data based on a clustering algorithm to obtain working condition groups, including:
carrying out numerical coding on the working condition data, and carrying out normalization processing to obtain a working condition sample;
and clustering the working condition samples based on a shared field density peak clustering algorithm, and obtaining working condition grouping of the working condition samples according to a clustering result.
Further, grouping the remaining life to obtain a degradation stage group includes:
dividing the duration of the remaining life into a plurality of degradation phases, and grouping the remaining life according to the divided degradation phases to obtain a degradation phase group.
Further, constructing an initial transformer residual life prediction model, which comprises the following steps:
an initial transformer remaining life prediction model is constructed consisting of a common feature extractor, a plurality of domain-specific feature extractors, and a plurality of domain-specific life predictors, the domain-specific feature extractors and domain-specific life predictors corresponding to respective domain groupings of the domain-specific training set.
Further, inputting the specific domain training set into an initial transformer residual life prediction model, extracting common features based on a common feature extractor, extracting specific domain features based on the specific domain feature extractor, predicting to obtain specific domain predicted life based on the specific domain life predictor, iteratively training the initial transformer residual life prediction model to obtain a fully trained transformer residual life prediction model, comprising:
inputting training samples in a specific domain training set into a common feature extractor to obtain common features of the training samples, inputting the common features into each specific domain feature extractor to obtain each specific domain feature of the training samples, inputting the specific domain features into each corresponding specific domain life predictor to obtain each specific domain predicted life of the training samples, and calculating the average value of each specific domain predicted life to obtain a residual life prediction result;
and taking the minimized loss function as an optimization target, and adjusting network parameters of the initial transformer residual life prediction model until the loss is not reduced, so as to obtain the fully trained transformer residual life prediction model.
On the other hand, the invention also provides an application method of the residual life prediction model of the transformer, which comprises the following steps:
acquiring real-time operation data and real-time working condition data of the transformer to be tested;
inputting real-time operation data and real-time working condition data of the transformer to be tested into a fully trained transformer residual life prediction model, wherein the fully trained transformer residual life prediction model is determined according to the training method of the transformer residual life prediction model;
and outputting the predicted residual life of the transformer to be tested by training a complete residual life prediction model of the transformer.
On the other hand, the invention also provides electronic equipment, which comprises a processor, a memory and a computer program stored in the memory and capable of running on the processor, wherein when the processor executes the program, the training method of the residual life prediction model of the transformer according to any one of the above methods and/or the application method of the residual life prediction model of the transformer are realized.
In another aspect, the present invention further provides a computer readable storage medium, on which a computer program is stored, where the computer program, when executed by a processor, implements a method for training a model for predicting remaining life of a transformer according to the above description, and/or a method for applying the model for predicting remaining life of a transformer according to the above description.
Compared with the prior art, the beneficial effects of adopting the embodiment are as follows: according to the invention, the multi-dimensional time sequence training set is subjected to specific domain grouping through the working condition grouping and the degradation stage grouping, the domain invariant features of the samples are captured through the common features of the extracted samples, and the differential features of the samples are captured through the specific domain features of the extracted samples, so that the residual life of the capacitive voltage transformer is accurately predicted under the multi-working condition.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings that are needed in the description of the embodiments will be briefly described below, it being evident that the drawings in the following description are only some embodiments of the present invention, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
FIG. 1 is a schematic flow chart of an embodiment of a training method for a residual life prediction model of a transformer;
FIG. 2 is a schematic flow chart of an embodiment of a method for applying a residual life prediction model of a transformer provided by the present invention;
fig. 3 is a schematic structural diagram of an embodiment of an electronic device provided by the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. It will be apparent that the described embodiments are only some, but not all, embodiments of the invention. 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.
It should be understood that the drawings of the schematic drawings are not drawn to scale. A flowchart, as used in this disclosure, illustrates operations implemented according to some embodiments of the present invention. It should be appreciated that the operations of the flow diagrams may be implemented out of order and that steps without logical context may be performed in reverse order or concurrently. Moreover, one or more other operations may be added to or removed from the flow diagrams by those skilled in the art under the direction of the present disclosure.
Some of the block diagrams shown in the figures are functional entities and do not necessarily correspond to physically or logically separate entities. These functional entities may be implemented in software or in one or more hardware modules or integrated circuits or in different networks and/or processor systems and/or microcontroller systems.
Reference herein to "an embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment may be included in at least one embodiment of the invention. The appearances of such phrases in various places in the specification are not necessarily all referring to the same embodiment, nor are separate or alternative embodiments mutually exclusive of other embodiments. Those of skill in the art will explicitly and implicitly appreciate that the embodiments described herein may be combined with other embodiments.
Fig. 1 is a flow chart of an embodiment of a method for training a model for predicting remaining life of a transformer according to the present invention, where, as shown in fig. 1, the method for training a model for predicting remaining life of a transformer includes:
s101, acquiring multi-dimensional time sequence data and working condition data of a transformer, and preprocessing the multi-dimensional time sequence data to obtain a multi-dimensional time sequence training set;
s102, clustering the working condition data based on a clustering algorithm to obtain working condition groups, grouping the residual life to obtain degradation stage groups, and grouping the multi-dimensional time sequence training set to a specific domain to obtain a specific domain training set according to the working condition groups and the degradation stage groups;
s103, constructing an initial transformer residual life prediction model, wherein the initial transformer residual life prediction model comprises a public feature extractor, a specific feature extractor and a specific domain life predictor;
s104, inputting the specific domain training set into an initial transformer residual life prediction model, extracting common features based on a common feature extractor, extracting specific domain features based on the specific domain feature extractor, predicting to obtain specific domain predicted life based on the specific domain life predictor, and iteratively training the initial transformer residual life prediction model to obtain a fully trained transformer residual life prediction model.
Specifically, in the training method of the residual life prediction model of the transformer, the multi-dimensional time sequence training set is subjected to specific domain grouping through the working condition grouping and the degradation stage grouping, the domain invariant features of the samples are captured through the common features of the extracted samples, and the differential features of the samples are captured through the specific domain features of the extracted samples, so that the residual life of the capacitive voltage transformer is accurately predicted under the multi-working condition.
In a specific embodiment of the present invention, acquiring multidimensional time series data of a transformer includes:
collecting index data of a transformer in a period of time;
dividing the time interval into a plurality of time subintervals, dividing each index data according to the time subintervals, and calculating the maximum value, the minimum value, the average value and the standard deviation of the index data in each time subinterval to obtain the multidimensional time sequence data of the transformer.
Specifically, during the acquisition of the data set, the transformer is acquired for a period of time [ T ] 1 ,T 2 ]The secondary output voltage, ambient temperature, humidity, magnetic field, load parameter, operating frequency and other indexes and operating period in the system are used to make the time interval T 1 ,T 2 ]Dividing each index into multiple cells on average, dividing each index into multiple groups of data according to time subintervals, calculating maximum value, minimum value, average value and standard deviation of index data in each time subinterval, taking maximum value, minimum value, average value and standard deviation of each index in each time subinterval as a group of characteristic data, and determining operation corresponding to the group of dataThe line years form training data to obtain multidimensional time sequence data of the transformer.
In a specific embodiment of the present invention, preprocessing the multi-dimensional time series data to obtain a multi-dimensional time series training set includes:
the characteristic data is formed into a characteristic parameter sequence, and a corresponding time sequence is established according to the time subinterval corresponding to each data in the characteristic parameter sequence;
calculating the correlation index, the monotonicity index and the discreteness index of each feature according to the feature parameter sequence and the time sequence;
and combining the correlation index, the monotonicity index and the discrete index to obtain a comprehensive index, setting a comprehensive index threshold, and eliminating characteristic data of which the comprehensive index is lower than the comprehensive index threshold.
Specifically, considering that not all features can well reflect the degradation trend of the CVT, feature selection is required to remove features with poor effects. Thus, the embodiments filter features by taking monotonicity, discreteness, and relatedness into account.
For time interval [ T ] 1 ，T 2 ]In, a certain characteristic of the ith sample, a parameter sequence is formedAccording to the time subintervals corresponding to the parameters, a time sequence T is established i ＝(t i1 ，t i2 ，…t in ) Then for this sample, the correlation index is:
wherein Corr i For the correlation index of the i-th sample,characteristic value representing characteristic j time in ith sample,/>Parameter sequence mean, t, representing features in the ith sample ij Sample time representing characteristic jth moment in ith sample, +.>Representing the time series mean of the feature in the ith sample.
Monotonicity index is:
wherein Mon is i As monotonicity index of the ith sample, N ia Representing the i-th sampleNumber N of (2) ib Represents +.>Number of (2), wherein>The difference between the previous and the next eigenvalues in each sample of the ith.
The discrete index is:
wherein,and->Is the maximum value, the minimum value, the +.>Is the standard deviation of the characteristic parameter,is the mean value of the characteristic parameters of the ith sample.
And adding and fusing the correlation index, the monotonicity index and the discrete index to obtain a comprehensive index Ccr, setting a comprehensive index threshold delta, retaining the characteristic when Ccr is more than or equal to delta, indicating that the characteristic can not well reflect the degradation trend of the CVT when Ccr is less than delta, and eliminating the characteristic.
In a specific embodiment of the present invention, clustering the working condition data based on a clustering algorithm to obtain a working condition group includes:
carrying out numerical coding on the working condition data, and carrying out normalization processing to obtain a working condition sample;
and clustering the working condition samples based on a shared field density peak clustering algorithm, and obtaining working condition grouping of the working condition samples according to a clustering result.
Specifically, due to the fact that differences exist in degradation characteristic distribution of the transformer under different working conditions, the prior art is difficult to extract enough domain invariant characteristics, and the difference characteristics of different working conditions and different degradation stages are ignored. In order to solve the problem, the invention divides the operation condition by a clustering algorithm and divides the degradation stage by the residual life. In the working condition grouping, the embodiment uses the voltage grade of the CVT group, the load type of each transformer substation, the zero sequence imbalance and the negative sequence imbalance of the CVT group site as working condition characteristics, and adopts DPC-SN (Density Peak Clustering Algorithm Based on Shared Neighborhood, density peak clustering algorithm based on the sharing field) to divide the operation working conditions.
In an embodiment, the zero sequence imbalance in the operating mode characteristics is:negative sequence imbalance is:wherein U0, U1 and U2 sequentially represent zero sequence component, positive sequence component and negative sequence component of three-phase voltageAmount of the components. In order to realize grouping of working conditions, the embodiment adopts a one-hot coding or manual assignment mode to preprocess the characteristic quantity, and normalizes the obtained codes according to the following formula:
wherein x represents feature code, x nor Representing normalized results, x max At the maximum value of x, x min Is the minimum value of x.
And then constructing a working condition sample from the normalized data: x= [ X ] 1 ，x 2 ，x 3 ，…，x L ]L represents the number of samples, whereWherein->h 1 ，V 1 ，Q 1 The characteristic quantities of zero sequence unbalance, negative sequence unbalance, load types of all substations, voltage grades of CVT groups, areas where the CVT group stations are located and the like are sequentially represented.
In the clustering process of the DPC-SN clustering algorithm, for sample points X in a sample set X j And x l Sample point x j Is denoted gamma (x) j ) Sample point x l Is denoted gamma (x) l ) Sample point x j And x l Is defined as:
Υ(x j ，x l )＝Υ(x j )∩Υ(x l )
based on the shared neighborhood concept, define sample point x j And x l Similarity between SIMs jl The method comprises the following steps:
where m represents a sample point in the shared neighborhood and d represents a sample pointEurope distance between y (x) j ，x l ) The i represents the number of samples in the shared neighborhood, the larger its value is the surface sample point x j And x l The greater the degree of similarity between the sample points x j ，x l Sample point x in shared neighborhood m The smaller the sum of the distances, the sample point x j And x l The greater the degree of similarity between the two.
For sample point x j Defining the local density ρ j The method comprises the following steps:wherein the weight function->
Defining sample point x j The relative distance d of (2) j ：
Wherein d c To cut off the distance d jl Is the sample point x j And x l Euclidean distance between them.
According to k j ＝d j *ρ j Sequencing samples according to the order from big to small, assuming the optimal clustering number is K, and selecting K j The first K sample points were taken as initial cluster centers and the values of the following evaluation functions were calculated:
wherein, and P θ Respectively is a cluster->And the average distance of the sample data within the cluster theta,for the distance between two clusters>And delta θ Respectively is a cluster->And average density of sample data within cluster θ, +.>The density average value of samples in two clusters is set, and alpha and beta are set weights.
And adjusting the K value until the evaluation function S value takes the minimum value to obtain the optimal clustering number K, and obtaining the working condition grouping of the working condition samples according to the optimal clustering number.
In a specific embodiment of the present invention, grouping remaining life into a degradation phase group includes:
dividing the duration of the remaining life into a plurality of degradation phases, and grouping the remaining life according to the divided degradation phases to obtain a degradation phase group.
Specifically, on the degradation phase grouping, embodiments divide the operational state of the transformer into different degradation phases: the method comprises four degradation stages of normal operation, slight degradation, moderate degradation and severe degradation, obtaining a degradation stage group, and dividing a multi-dimensional time sequence training set E according to the obtained working condition group and the degradation stage group as follows:
| normal state | Slightly degenerated | Moderate degradation | Severe degradation of | |
| Working condition source domain 1 | Data set 11 | Data set 12 | Data set 13 | Data set 14 |
| Working condition source domain 2 | Data set 21 | Data set 22 | Data set 23 | Data set 24 |
| …… | …… | …… | …… | …… |
| Working condition source domain M | Data set M1 | Data set M2 | Data set M3 | Data set M4 |
In a specific embodiment of the invention, constructing an initial transformer residual life prediction model comprises the following steps:
an initial transformer remaining life prediction model is constructed consisting of a common feature extractor, a plurality of domain-specific feature extractors, and a plurality of domain-specific life predictors, the domain-specific feature extractors and domain-specific life predictors corresponding to respective domain groupings of the domain-specific training set.
Specifically, to correspond to a specific domain training set, the initial transformer remaining life prediction model is constructed to include a common feature extractor F (), a plurality of specific domain feature extractors corresponding to specific domain groupingsAnd a plurality of domain-specific lifetime predictors corresponding to the domain-specific packets +.>
In a specific embodiment of the present invention, inputting a specific domain training set into an initial transformer residual life prediction model, extracting common features based on a common feature extractor, extracting specific domain features based on a specific domain life predictor, obtaining a specific domain predicted life based on the specific domain life predictor prediction, iteratively training the initial transformer residual life prediction model, and obtaining a fully trained transformer residual life prediction model, including:
inputting training samples in a specific domain training set into a common feature extractor to obtain common features of the training samples, inputting the common features into each specific domain feature extractor to obtain each specific domain feature of the training samples, inputting the specific domain features into each corresponding specific domain life predictor to obtain each specific domain predicted life of the training samples, and calculating the average value of each specific domain predicted life to obtain a residual life prediction result;
and taking the minimized loss function as an optimization target, and adjusting network parameters of the initial transformer residual life prediction model until the loss is not reduced, so as to obtain the fully trained transformer residual life prediction model.
Specifically, it is provided withThe q-th sample of the degradation stage of the M-th working condition source domain r of the input model is represented by m=1, 2, …, M, r=1, 2,3,4, and represents normal operation, slight degradation, moderate degradation and serious degradation. Then sample->
Wherein,and the characteristic value of the kth characteristic t moment of the qth sample of the degradation stage of the mth working condition source domain r is input.
Sample ofOutput as +.>Through a domain-specific extractor H m Output after (-) is->Defined as->Pass-domain-specific lifetime predictor P m Post output asDefined as->
After the input sample passes through the 4M specific domain networks, 4M prediction outputs can be obtained, and the average value is used to represent the prediction result of the input sample:
and adopts the minimized loss function as an optimization target:
wherein lambda is a weight parameter, L R For the predicted loss between the actual observed remaining life of the sample and the predicted remaining life, Q is the number of samples in the source domain; l (L) D For the distribution distance between the source domain and the target domain features, K represents a kernel function set, H q ，H p MMD representing features extracted by feature extractor in source domain and target domain k (H q ，H p ) Is the maximum mean difference obtained when kernel function k is used.
And finally, obtaining a complete training transformer residual life prediction model through iterative optimization until the loss is not reduced.
Compared with the prior art, the multi-dimensional time sequence training set is subjected to specific domain grouping through the working condition grouping and the degradation stage grouping, the common characteristics of the samples are extracted through the common characteristic extractor to capture the domain invariant characteristics of the samples, and the specific domain characteristics of the samples are extracted through the specific domain extractor to capture the differentiation characteristics of the samples, so that the residual life of the capacitive voltage transformer is accurately predicted under the condition of multiple working conditions.
The embodiment of the invention also provides a method for applying the residual life prediction model of the transformer, and referring to fig. 2, fig. 2 is a schematic flow chart of an embodiment of the method for applying the residual life prediction model of the transformer, which includes steps S201 to S203, wherein:
s201, acquiring real-time operation data and real-time working condition data of a transformer to be tested;
s202, inputting real-time operation data and real-time working condition data of the transformer to be tested into a fully trained transformer residual life prediction model, wherein the fully trained transformer residual life prediction model is determined by the training method of the transformer residual life prediction model;
s203, outputting the predicted residual life of the transformer to be tested by the trained complete residual life prediction model of the transformer.
In the embodiment of the invention, firstly, the real-time operation data and the real-time working condition data of the transformer to be tested are effectively obtained; and then, effectively identifying the real-time operation data and the real-time working condition data by using the fully trained mutual inductor residual life prediction model, and predicting the residual life of the mutual inductor, so that the corresponding mutual inductor residual life prediction can be output.
The present invention further provides an electronic device 300, as shown in fig. 3, fig. 3 is a schematic structural diagram of an embodiment of the electronic device provided by the present invention, where the electronic device 300 includes a processor 301, a memory 302, and a computer program stored in the memory 302 and capable of running on the processor 301, and when the processor 301 executes the program, the training method of the residual life prediction model of the transformer and/or the application method of the residual life prediction model of the transformer are implemented.
As a preferred embodiment, the electronic device further includes a display 303 for displaying the process of executing the method for training the residual life prediction model of the transformer and/or the method for applying the residual life prediction model of the transformer by the processor 301.
The processor 301 may be an integrated circuit chip, and has signal processing capability. The processor 301 may be a general-purpose processor, including a central processing unit (Central Processing Unit, CPU), a network processor (Network Processor, NP), etc.; also digital signal processors (Digital Signal Processor, DSP), application specific integrated circuits (Application Specific Integrated Circuit, ASIC). The disclosed methods, steps, and logic blocks in the embodiments of the present invention may be implemented or performed. A general purpose processor may also be a microprocessor or the processor may be any conventional processor or the like.
The Memory 302 may be, but is not limited to, a random access Memory (Random Access Memory, RAM), a Read Only Memory (ROM), a Secure Digital (SD Card), a Flash Card (Flash Card), etc. The memory 302 is configured to store a program, and the processor 301 executes the program after receiving an execution instruction, and the method for defining a flow disclosed in any of the foregoing embodiments of the present invention may be applied to the processor 301 or implemented by the processor 301.
The display 303 may be an LED display, a liquid crystal display, a touch display, or the like. The display 303 is used to display various information on the electronic device 300.
It is to be understood that the configuration shown in fig. 3 is merely a schematic diagram of one configuration of the electronic device 300, and that the electronic device 300 may also include more or fewer components than those shown in fig. 3. The components shown in fig. 3 may be implemented in hardware, software, or a combination thereof.
The embodiment of the invention also provides a computer readable storage medium, wherein a computer program is stored on the computer readable storage medium, and when the program is executed by a processor, the method for training the residual life prediction model of the transformer and/or the method for applying the residual life prediction model of the transformer are realized.
In general, the computer instructions for carrying out the methods of the present invention may be carried in any combination of one or more computer-readable storage media. The non-transitory computer-readable storage medium may include any computer-readable medium, except the signal itself in temporary propagation.
The computer readable storage medium can be, for example, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or a combination of any of the foregoing. More specific examples (a non-exhaustive list) of the computer-readable storage medium would include the following: an electrical connection having one or more wires, a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In the context of this document, a computer readable storage medium may be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device.
The present invention is not limited to the above-mentioned embodiments, and any changes or substitutions that can be easily understood by those skilled in the art within the technical scope of the present invention are intended to be included in the scope of the present invention.
Claims (10)
1. A method for training a prediction model of remaining life of a transformer, the method comprising:
acquiring multi-dimensional time sequence data and working condition data of a transformer, and preprocessing the multi-dimensional time sequence data to obtain a multi-dimensional time sequence training set;
clustering the working condition data based on a clustering algorithm to obtain working condition groups, grouping the residual life to obtain degradation stage groups, and grouping the multi-dimensional time sequence training set to a specific domain to obtain a specific domain training set according to the working condition groups and the degradation stage groups;
constructing an initial transformer residual life prediction model, wherein the initial transformer residual life prediction model comprises a public feature extractor, a specific domain feature extractor and a specific domain life predictor;
inputting the specific domain training set into the initial transformer residual life prediction model, extracting common features based on a common feature extractor, extracting specific domain features based on the specific domain feature extractor, predicting to obtain specific domain predicted life based on the specific domain life predictor, and iteratively training the initial transformer residual life prediction model to obtain a fully trained transformer residual life prediction model.
2. The method for training a residual life prediction model of a transformer according to claim 1, wherein the acquiring multi-dimensional time series data of the transformer comprises:
collecting index data of a transformer in a period of time;
dividing the time interval into a plurality of time subintervals, dividing each index data according to the time subintervals, and calculating the maximum value, the minimum value, the average value and the standard deviation of the index data in each time subinterval to obtain the multidimensional time sequence data of the transformer.
3. The method for training the residual life prediction model of the transformer according to claim 2, wherein the preprocessing the multi-dimensional time series data to obtain a multi-dimensional time series training set comprises:
the characteristic data is formed into a characteristic parameter sequence, and a corresponding time sequence is established according to the time subinterval corresponding to each data in the characteristic parameter sequence;
calculating the correlation index, the monotonicity index and the discreteness index of each feature according to the feature parameter sequence and the time sequence;
and combining the correlation index, the monotonicity index and the discrete index to obtain a comprehensive index, setting a comprehensive index threshold, and eliminating characteristic data of which the comprehensive index is lower than the comprehensive index threshold.
4. The method for training the residual life prediction model of the transformer according to claim 1, wherein the clustering processing of the working condition data based on the clustering algorithm to obtain working condition groups comprises the following steps:
carrying out numerical coding on the working condition data, and carrying out normalization processing to obtain a working condition sample;
and clustering the working condition samples based on a shared field density peak clustering algorithm, and obtaining working condition grouping of the working condition samples according to a clustering result.
5. The method of training a model for predicting remaining life of a transformer according to claim 1, wherein said grouping the remaining life into groups of degradation phases comprises:
dividing the duration of the remaining life into a plurality of degradation phases, and grouping the remaining life according to the divided degradation phases to obtain a degradation phase group.
6. The method for training the residual life prediction model of the transformer according to claim 1, wherein the constructing the initial residual life prediction model of the transformer comprises:
an initial transformer remaining life prediction model is constructed consisting of a common feature extractor, a plurality of domain-specific feature extractors, and a plurality of domain-specific life predictors, the domain-specific feature extractors and domain-specific life predictors corresponding to respective domain groupings of the domain-specific training set.
7. The method for training a model for predicting remaining life of a transformer according to claim 6, wherein inputting the domain-specific training set into the model for predicting remaining life of the initial transformer, extracting common features based on a common feature extractor, extracting domain-specific features based on a domain-specific feature extractor, predicting a domain-specific predicted life based on a domain-specific life predictor, iteratively training the model for predicting remaining life of the initial transformer, and obtaining a fully trained model for predicting remaining life of the transformer comprises:
inputting the training samples in the specific domain training set into a common feature extractor to obtain common features of the training samples, inputting the common features into each specific domain feature extractor to obtain each specific domain feature of the training samples, inputting the specific domain features into each corresponding specific domain life predictor to obtain each specific domain predicted life of the training samples, and calculating the average value of each specific domain predicted life to obtain a residual life prediction result;
and taking the minimized loss function as an optimization target, and adjusting network parameters of the initial transformer residual life prediction model until the loss is not reduced, so as to obtain the fully trained transformer residual life prediction model.
8. The application method of the residual life prediction model of the transformer is characterized by comprising the following steps of:
acquiring real-time operation data and real-time working condition data of the transformer to be tested;
inputting the real-time operation data and the real-time working condition data of the transformer to be tested into a fully trained transformer residual life prediction model, wherein the fully trained transformer residual life prediction model is determined according to the training method of the transformer residual life prediction model according to any one of claims 1 to 7;
and outputting the predicted residual life of the transformer to be tested by the trained complete residual life prediction model of the transformer.
9. An electronic device comprising a processor, a memory and a computer program stored on the memory and executable on the processor, characterized in that the processor implements the method for training the residual life prediction model of the transformer according to any one of claims 1 to 7 and/or the method for applying the residual life prediction model of the transformer according to claim 8 when the processor executes the program.
10. A computer readable storage medium having stored thereon a computer program, characterized in that the computer program, when executed by a processor, implements the method of training the mutual inductor remaining life prediction model according to any one of claims 1 to 7, and/or the method of applying the mutual inductor remaining life prediction model according to claim 8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311208255.2A CN117236022A (en) | 2023-09-18 | 2023-09-18 | Training method and application method of residual life prediction model of transformer and electronic equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311208255.2A CN117236022A (en) | 2023-09-18 | 2023-09-18 | Training method and application method of residual life prediction model of transformer and electronic equipment |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117236022A true CN117236022A (en) | 2023-12-15 |
Family
ID=89089109
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311208255.2A Pending CN117236022A (en) | 2023-09-18 | 2023-09-18 | Training method and application method of residual life prediction model of transformer and electronic equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117236022A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117912534A (en) * | 2024-03-20 | 2024-04-19 | 济南浪潮数据技术有限公司 | Disk state prediction method and device, electronic equipment and storage medium |
-
2023
- 2023-09-18 CN CN202311208255.2A patent/CN117236022A/en active Pending
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117912534A (en) * | 2024-03-20 | 2024-04-19 | 济南浪潮数据技术有限公司 | Disk state prediction method and device, electronic equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112699913B (en) | Method and device for diagnosing abnormal relationship of household transformer in transformer area | |
| CN111505433B (en) | Low-voltage transformer area indoor variable relation error correction and phase identification method | |
| CN109829497B (en) | Supervised learning-based station area user identification and discrimination method | |
| CN111080105A (en) | Transformer area user-to-user relationship identification method and system based on voltage time sequence data | |
| CN110601173B (en) | Distribution network topology identification method and device based on edge calculation | |
| CN109934301B (en) | Power load cluster analysis method, device and equipment | |
| CN112418687B (en) | User electricity utilization abnormity identification method and device based on electricity utilization characteristics and storage medium | |
| CN110244205A (en) | Cable accessory Recognition of Partial Discharge based on convolutional neural networks under a kind of small data quantity | |
| CN117236022A (en) | Training method and application method of residual life prediction model of transformer and electronic equipment | |
| CN113408341A (en) | Load identification method and device, computer equipment and storage medium | |
| CN112819649A (en) | Method and device for determining station area subscriber change relationship | |
| CN113659564B (en) | Low-voltage distribution network topology identification method and system based on voltage fluctuation feature clustering | |
| CN112508254B (en) | Method for determining investment prediction data of transformer substation engineering project | |
| CN107274025B (en) | System and method for realizing intelligent identification and management of power consumption mode | |
| CN112215482A (en) | Method and device for identifying user variable relationship | |
| CN111259965A (en) | Method and system for carrying out mean value clustering on electrical characteristic data based on dimension reduction | |
| CN116011158A (en) | Topology identification method, system and device for low-voltage transformer area | |
| CN115879799A (en) | Transformer substation electric energy quality analysis method | |
| CN109031057A (en) | A kind of high-tension electricity apparatus local discharge mode identification method and device | |
| CN113393099A (en) | Power distribution network project group value index evaluation method and device and terminal equipment | |
| CN109521312B (en) | Non-technical line loss detection method, device and system | |
| CN111626559A (en) | Main factor analysis method-based medium-voltage distribution network line loss key characteristic index extraction method and system | |
| CN112256766A (en) | Power consumption behavior analysis method for energy collection terminal | |
| CN111505445A (en) | Credibility detection method and device for mutual-user relationship of transformer area and computer equipment | |
| CN114462901B (en) | Active power distribution network electric energy quality evaluation method and device, terminal and storage medium |
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 |