Method and system for identifying fault data of power transformer

Abstract

The invention provides a method and a system for identifying fault data of a power transformer. The method comprises the following steps: acquiring online monitoring data of multi-dimensional operating parameters of the power transformer from a data input interface; effective monitoring data generated after the online monitoring data are preprocessed are input into a dynamic threshold value model for self-adaptive threshold value calculation, a threshold value calculation result is obtained, and the effective monitoring data with the threshold value calculation result not in a set dynamic threshold value range are marked as first-level abnormal data; performing physical constraint dual verification on the first-level abnormal data, and marking the first-level abnormal data which does not pass at least one verification as second-level abnormal data; and inputting a feature vector generated by feature extraction of the secondary abnormal data into a constructed noise classification model for noise type identification, and outputting a final identification result according to a confidence value and a noise type output by a noise classification module. According to the method and the system, redundant alarms are greatly reduced, and the fault positioning accuracy is improved.

Description

Technical Field

[0001] This invention relates to the field of power fault diagnosis technology, and more specifically, to a method and system for identifying power transformer fault data. Background Technology

[0002] As a pivotal piece of equipment in power transmission and transformation systems, the operational reliability of power transformers directly impacts the safety and stability of the power grid. With traditional periodic maintenance methods gradually being replaced by online monitoring technologies due to their lag in response, oil chromatography analysis, partial discharge detection, and real-time temperature monitoring have become core methods for condition assessment. However, in complex substation environments, electromagnetic interference intensity can reach as high as 3kV / m, causing the abnormality rate of sensor data to remain consistently between 12% and 18%. The rate of change in dissolved gas concentration in the oil is as low as 5 ppm per hour, highly overlapping with the spectrum of transient interference signals. Microsecond-level partial discharge pulse signals are easily submerged by switch operation noise; these data quality issues fundamentally limit the accuracy of fault diagnosis.

[0003] Current outlier handling techniques commonly employ fixed threshold filtering methods, but transformer monitoring data exhibits significant dynamic fluctuations. Load fluctuations cause winding temperature standard deviations to vary by ±15 degrees Celsius, and the rigid 3σ criterion results in an outlier miss rate exceeding one-quarter. Simple mean replacement strategies disrupt data gradient continuity, increasing the false positive rate of gradient diagnostic algorithms by 40%. A 220kV transformer fault case demonstrates that fixed threshold filtering masked a critical acetylene concentration anomaly, delaying diagnosis by up to 72 hours. Step detection technology suffers from severe domain adaptability deficiencies, failing to differentiate the physical characteristics of equipment at different voltage levels. The missetting rate of step duration parameters for 220kV equipment reached 34%, leading to a 2.3-fold increase in false alarms for light gas. Noise processing reveals a lack of intelligent bottlenecks, lacking a dynamic mapping mechanism between noise types and filtering algorithms.

[0004] Data quality defects directly led to a comprehensive deterioration in the performance of fault diagnosis models. Noise interference caused the gas ratio calculation deviation to exceed the ±5% allowable range of the IEC 60599 standard, and the false diagnosis rate of the three-ratio method climbed to 18.6%. LSTM neural networks required 320 rounds to converge, and training efficiency with high-quality data decreased by 35%. A systemic break occurred in the operation and maintenance decision-making chain; in a 500kV substation, noise masked the bushing discharge step signal, causing the early warning mechanism to completely fail. The calculation error of the aging rate of oil-paper insulation reached ±20%, resulting in serious misjudgment of the maintenance cycle. Distorted life assessment triggered a chain reaction; a converter station replaced a transformer prematurely due to misjudgment of insulation condition, causing a direct economic loss of eight million yuan. Summary of the Invention

[0005] To address the issue of low data quality caused by the use of fixed threshold filtering, step identification, and noise processing techniques in existing technologies for identifying online monitoring data of power transformers, this invention provides a method and system for identifying power transformer fault data.

[0006] According to one aspect of the present invention, the present invention provides a method for identifying fault data of a power transformer, comprising:

[0007] Online monitoring data of multi-dimensional operating parameters of power transformers are obtained from the data input interface;

[0008] The effective monitoring data generated after preprocessing the online monitoring data is input into the dynamic threshold model for adaptive threshold calculation to obtain the threshold calculation result. The effective monitoring data whose threshold calculation result is not within the set dynamic threshold range is marked as first-level abnormal data. The dynamic threshold model is constructed using the entropy weight method.

[0009] The first-level abnormal data is subjected to physical constraint dual verification, and the first-level abnormal data that fails at least one verification is marked as second-level abnormal data. The physical constraint dual verification includes physical parameter constraint verification through a device feature characteristic mapping library constructed using several IEC standard parameters and signal step feature verification using a multi-scale edge detection algorithm.

[0010] The feature vector generated by feature extraction from the secondary abnormal data is input into the noise classification model to identify the noise type. The final identification result is output based on the confidence value and noise type output by the noise classification module. The noise classification model is constructed using a deep belief network based on a three-layer restricted Boltzmann machine. The noise types include impulse noise and periodic noise. The final identification result includes a noise type alarm and a rejection of identification.

[0011] According to another aspect of the present invention, the present invention provides a system for identifying fault data of power transformers, the system comprising:

[0012] The data acquisition module is used to acquire online monitoring data of multi-dimensional operating parameters of power transformers from the data input interface;

[0013] The first identification module is used to input the effective monitoring data generated after preprocessing the online monitoring data into the dynamic threshold model for adaptive threshold calculation, obtain the threshold calculation result, and mark the effective monitoring data whose threshold calculation result is not within the set dynamic threshold range as first-level abnormal data. The dynamic threshold model is constructed using the entropy weight method.

[0014] The second identification module is used to perform physical constraint dual verification on the first-level abnormal data and mark the first-level abnormal data that fails at least one verification as second-level abnormal data. The physical constraint dual verification includes physical parameter constraint verification through a device feature characteristic mapping library constructed using several IEC standard parameters and signal step feature verification using a multi-scale edge detection algorithm.

[0015] The third identification module is used to input the feature vector generated by feature extraction from the secondary abnormal data into the noise classification model to identify the noise type, and output the final identification result based on the confidence value and noise type output by the noise classification module. The noise classification model is constructed using a deep belief network based on a three-layer restricted Boltzmann machine. The noise type includes impulse noise and periodic noise. The final identification result includes a noise type alarm and a rejection of identification.

[0016] According to another aspect of the present invention, a computer-readable storage medium is provided, the storage medium storing a computer program for performing the methods described in any of the above aspects of the present invention.

[0017] According to another aspect of the present invention, an electronic device is provided, the electronic device comprising: a processor; a memory for storing executable instructions of the processor; the processor being configured to read the executable instructions from the memory and execute the instructions to implement the method described in any of the preceding aspects of the present invention.

[0018] The present invention discloses a method and system for identifying fault data of power transformers. The method includes: acquiring online monitoring data of multi-dimensional operating parameters of the power transformer from a data input interface; inputting the effective monitoring data generated after preprocessing the online monitoring data into a dynamic threshold model for adaptive threshold calculation to obtain a threshold calculation result; marking the effective monitoring data whose threshold calculation result is not within the set dynamic threshold range as first-level abnormal data; performing physical constraint dual verification on the first-level abnormal data, and marking the first-level abnormal data that fails at least one verification as second-level abnormal data; inputting the feature vector generated by feature extraction from the second-level abnormal data into a constructed noise classification model for noise type identification, and outputting the final identification result based on the confidence value and noise type output by the noise classification module. The method and system distinguish whether a transformer is operating normally through a dynamic threshold model, determine whether a power transformer has malfunctioned through dual verification of physical constraints, and determine whether the data can be identified and the type of noise when it can be identified through the confidence level of a noise classification model. Thus, through a three-level identification chain collaborative mechanism, it overcomes the limitations of traditional binary judgment: the dynamic threshold model covers the ±3σ fluctuation range, reducing the false alarm rate by 60%; the physical constraint verification combined with noise discrimination compresses the false alarm rate to below 0.1%; the rejection status accurately intercepts 98% of electromagnetic interference and sensor drift data; and it achieves 100Mbps high throughput processing with a 35ms end-to-end delay. It provides reliable and real-time data preprocessing support for 500kV smart substations and complies with the IEC 61850 protocol. Attached Figure Description

[0019] Exemplary embodiments of the present invention can be more fully understood by referring to the following figures:

[0020] Figure 1 A flowchart of a method for identifying power transformer fault data according to a preferred embodiment of the present invention;

[0021] Figure 2 This is a schematic diagram of the structure of a system for identifying power transformer fault data according to a preferred embodiment of the present invention;

[0022] Figure 3 This is a schematic diagram of the structure of an electronic device according to a preferred embodiment of the present invention. Detailed Implementation

[0023] Exemplary embodiments of the invention will now be described with reference to the accompanying drawings. However, the invention may be embodied in many different forms and is not limited to the embodiments described herein. These embodiments are provided to fully and completely disclose the invention and to fully convey its scope to those skilled in the art. The terminology used in the exemplary embodiments illustrated in the drawings is not intended to limit the invention. In the drawings, the same units / elements are referred to by the same reference numerals.

[0024] Unless otherwise stated, the terms used herein (including technical terms) have their common meaning as understood by one of ordinary skill in the art. Furthermore, it is understood that terms defined in commonly used dictionaries should be understood to have a meaning consistent with the context of their relevant field, and not to be interpreted as having an idealized or overly formal meaning.

[0025] Exemplary methods

[0026] Figure 1 This is a flowchart of a method for identifying power transformer fault data according to a preferred embodiment of the present invention. Figure 1 As shown, the method for identifying power transformer fault data according to this preferred embodiment begins from step 101.

[0027] In step 101, online monitoring data of multi-dimensional operating parameters of the power transformer are obtained from the data input interface.

[0028] Preferably, the online monitoring data of the power transformer is acquired from the data input interface, wherein the data input interface supports multi-source heterogeneous data access, including online monitoring data of oil chromatography transmitted according to the IEC 61850 protocol, partial discharge pulse current data synchronized according to the IEEE C37.118 protocol, and temperature data transmitted from a temperature sensor network with a customized RS485 interface.

[0029] In this preferred embodiment, 12 high-precision sensor data streams are accessed through different standard protocols to collect key parameters such as load current, oil temperature gradient, and partial discharge in real time, with a sampling frequency of up to 10kHz.

[0030] In step 102, the effective monitoring data generated after preprocessing the online monitoring data is input into the dynamic threshold model for adaptive threshold calculation to obtain the threshold calculation result. The effective monitoring data whose threshold calculation result is not within the set dynamic threshold range is marked as first-level abnormal data. The dynamic threshold model is constructed using the entropy weight method.

[0031] Preferably, the online monitoring data is input into a dynamic threshold model for adaptive threshold calculation to obtain the threshold calculation result, wherein the expression of the dynamic threshold model is:

[0032]

[0033] In the formula, H t For dynamic threshold, P i,t Let be the measured value of the i-th dimension parameter of the power transformer collected at time t, and N be the total number of dimensions of the power transformer collected, μ i and σ i It is the dynamic mean and standard deviation of the i-th dimension parameter, calculated based on historical data of power transformers; the weighting coefficient ω i It is a constant that is dynamically allocated using the entropy weighting method.

[0034] In this preferred embodiment, the N-dimensional parameters collected by the power transformer include load rate, winding temperature, and years of operation. i and σ i It is a dynamic mean and standard deviation calculated based on at least five years of historical data from the device. Weighting coefficient ω i The overentropy weighting method dynamically allocates parameters; for example, the weight of oil temperature parameters automatically adjusts with seasonal changes by ±15%. When the real-time calculated H... t When the absolute value exceeds 1, i.e., it exceeds the fluctuation range of plus or minus three standard deviations, the system immediately marks the data as a level one anomaly. This model automatically updates the basic parameters every 24 hours and introduces an equipment aging compensation mechanism, with the weight of the operating years increasing by 0.05 each year to ensure that the threshold always accurately matches the actual state of the equipment.

[0035] In step 103, the first-level abnormal data is subjected to physical constraint dual verification, and the first-level abnormal data that fails at least one verification is marked as second-level abnormal data. The physical constraint dual verification includes physical parameter constraint verification through a device feature characteristic mapping library constructed using several IEC standard parameters and signal step feature verification using a multi-scale edge detection algorithm.

[0036] Preferably, the first-level anomaly data undergoes dual physical constraint verification. This dual physical constraint verification includes physical parameter constraint verification using a device characteristic mapping library constructed from several IEC standard parameters, and signal step characteristic verification using a multi-scale edge detection algorithm.

[0037] When performing physical parameter constraint verification, the core function of the equipment characteristic mapping library adopts the voltage-insulation thickness correlation equation, the expression of which is:

[0038] R = α * lnV + βD i

[0039] In the formula, R is the associated characteristic value, V is the voltage level in kilovolts, and D is the voltage level in kilovolts. iThe insulation thickness is measured in millimeters. The coefficients α and β are determined by fitting several sets of factory test data using the least squares method. When the associated characteristic value R calculated based on the obtained power transformer voltage level and insulation thickness exceeds the set associated characteristic value range, it is determined that the physical parameter constraint verification has failed.

[0040] When using a multi-scale edge detection algorithm to verify signal step features, the Daubechies wavelet-based multi-scale edge detection algorithm is used to calculate the step duration and the frequency domain energy entropy change value. When the step time exceeds the step time range of the corresponding voltage level, or the energy entropy change value is less than a custom threshold, the signal step feature verification is determined to fail.

[0041] In this preferred embodiment, the physical parameter constraint verification is performed based on a library of 128 physical parameters established according to the IEC 60270 standard. Corresponding associated features are set according to the physical parameters, and reasonable constraint ranges for these associated features are calculated. If the associated feature value exceeds the constraint range after calculation based on the acquired online monitoring data, the physical parameter constraint verification is deemed unsuccessful. Signal step feature verification utilizes signal characteristics, such as step duration and energy entropy changes, to analyze whether the online monitoring data conforms to fault characteristics, thereby eliminating false triggers caused by electromagnetic interference, sensor drift, etc. When the step feature of the online monitoring data meets the set range, the verification is deemed successful. Only when the online monitoring data simultaneously passes both the physical parameter constraint verification and the signal step feature verification is a power transformer fault alarm determined. This dual verification improves data selectivity and reduces the false alarm rate (to below 0.1%) and the missed alarm rate (reduced by 60%). In high-noise environments, such as when the electromagnetic interference intensity in a substation reaches 3kV / m, the dual verification mechanism can effectively distinguish between real fault signals and interference signals.

[0042] In step 104, the feature vector generated by feature extraction from the secondary abnormal data is input into the noise classification model to identify the noise type. The final identification result is output based on the confidence value and noise type output by the noise classification module. The noise classification model is constructed using a deep belief network based on a three-layer restricted Boltzmann machine. The noise type includes impulse noise and periodic noise. The final identification result includes a noise type alarm and a rejection of identification.

[0043] The feature vectors generated by feature extraction from the secondary anomaly data are input into the constructed noise classification model for noise type identification. The final identification result is then output based on the confidence value and noise type output by the noise classification module, where:

[0044] When the confidence value is not less than the preset confidence threshold, the final identification result is determined to be an alarm of the corresponding noise type based on the output noise type; otherwise, the final identification result is rejection.

[0045] In this preferred embodiment, the output layer of the noise classification model uses the Sigmoid function to generate noise type probabilities, and is trained using 500,000 sets of samples labeled with noise types.

[0046] Preferably, the method further includes:

[0047] When the threshold calculation result is within the set dynamic threshold range, the output identification result is that the power transformer is operating normally; or

[0048] When both physical constraint double verifications of the first-level abnormal data fail, the output identification result is a confirmed power transformer fault alarm.

[0049] Preferably, after outputting the final identification result based on the noise type and its corresponding confidence value output by the noise classification module, the method further includes storing the secondary abnormal data that are rejected in the final identification result into the historical database, which is used to update the weight of each noise type in the noise classification model according to the set time period.

[0050] The method for identifying power transformer fault data described in this preferred embodiment distinguishes whether the transformer is operating normally through a dynamic threshold model, determines whether the power transformer has a fault through dual verification of physical constraints, and determines whether the data can be identified and the type of noise when it can be identified through the confidence level of a noise classification model. Thus, through a three-level identification chain collaborative mechanism, it breaks through the limitations of traditional binary judgment: the dynamic threshold model covers the ±3σ fluctuation range, reducing the false alarm rate by 60%; the physical constraint verification combined with noise discrimination compresses the false alarm rate to below 0.1%; the rejection status accurately intercepts 98% of electromagnetic interference and sensor drift data; and it achieves 100Mbps high throughput processing with an end-to-end delay of 35ms. It provides reliable and real-time data preprocessing support for 500kV smart substations and complies with the IEC 61850 protocol.

[0051] Exemplary System

[0052] Figure 2 This is a schematic diagram of the structure of a system for identifying power transformer fault data according to a preferred embodiment of the present invention. Figure 2 As shown, the system 200 for identifying power transformer fault data according to this preferred embodiment includes:

[0053] The data acquisition module 201 is used to acquire online monitoring data of multi-dimensional operating parameters of power transformers from the data input interface;

[0054] The first identification module 202 is used to input the effective monitoring data generated after preprocessing the online monitoring data into the dynamic threshold model for adaptive threshold calculation, obtain the threshold calculation result, and mark the effective monitoring data whose threshold calculation result is not within the set dynamic threshold range as first-level abnormal data. The dynamic threshold model is constructed using the entropy weight method.

[0055] The second identification module 203 is used to perform physical constraint dual verification on the first-level abnormal data and mark the first-level abnormal data that fails at least one verification as second-level abnormal data. The physical constraint dual verification includes physical parameter constraint verification through a device feature characteristic mapping library constructed using several IEC standard parameters and signal step feature verification using a multi-scale edge detection algorithm.

[0056] The third identification module 204 is used to input the feature vector generated by feature extraction from the secondary abnormal data into the noise classification model to identify the noise type, and output the final identification result based on the confidence value and noise type output by the noise classification module. The noise classification model is constructed using a deep belief network based on a three-layer restricted Boltzmann machine. The noise type includes impulse noise and periodic noise. The final identification result includes a noise type alarm and a rejection of identification.

[0057] Preferably, the first identification module 202 is further configured to output an identification result indicating that the power transformer is operating normally when the threshold calculation result is within the set dynamic threshold range; the second identification module 203 is further configured to output an identification result confirming a power transformer fault alarm when both physical constraint double verifications of the first-level abnormal data fail.

[0058] Preferably, the system further includes a historical database for storing secondary anomaly data whose final identification result is rejection, so as to update the weight of each noise type in the noise classification model according to a set time period.

[0059] Preferably, the data acquisition module 201 acquires online monitoring data of the power transformer from the data input interface, wherein the data input interface supports multi-source heterogeneous data access, including online monitoring data of oil chromatography transmitted by the IEC 61850 protocol, partial discharge pulse current data synchronized by the IEEE C37.118 protocol, and temperature data transmitted by a temperature sensor network with a customized RS485 interface.

[0060] Preferably, the first identification module 202 inputs the online monitoring data into a dynamic threshold model for adaptive threshold calculation to obtain the threshold calculation result, wherein the expression of the dynamic threshold model is:

[0061]

[0062] In the formula, H t For dynamic threshold, P i,t Let be the measured value of the i-th dimension parameter of the power transformer collected at time t, and N be the total number of dimensions of the power transformer collected, μ i and σ i It is the dynamic mean and standard deviation of the i-th dimension parameter, calculated based on historical data of power transformers; the weighting coefficient ω i It is a constant that is dynamically allocated using the entropy weighting method.

[0063] Preferably, the second identification module 203 performs dual physical constraint verification on the first-level abnormal data. This dual physical constraint verification includes physical parameter constraint verification using a device characteristic mapping library constructed from several IEC standard parameters and signal step characteristic verification using a multi-scale edge detection algorithm, wherein:

[0064] When performing physical parameter constraint verification, the core function of the equipment characteristic mapping library adopts the voltage-insulation thickness correlation equation, the expression of which is:

[0065] R = α * lnV + βD i

[0066] In the formula, R is the associated characteristic value, V is the voltage level in kilovolts, and D is the voltage level in kilovolts. i The insulation thickness is measured in millimeters. The coefficients α and β are determined by fitting several sets of factory test data using the least squares method. When the associated characteristic value R calculated based on the obtained power transformer voltage level and insulation thickness exceeds the set associated characteristic value range, it is determined that the physical parameter constraint verification has failed.

[0067] When using a multi-scale edge detection algorithm to verify signal step features, the Daubechies wavelet-based multi-scale edge detection algorithm is used to calculate the step duration and the frequency domain energy entropy change value. When the step time exceeds the step time range of the corresponding voltage level, or the energy entropy change value is less than a custom threshold, the signal step feature verification is determined to fail.

[0068] Preferably, the third identification module 204 inputs the feature vector generated by feature extraction from the secondary abnormal data into the constructed noise classification model for noise type identification, and outputs the final identification result based on the confidence value and noise type output by the noise classification module, wherein:

[0069] When the confidence value is not less than the preset confidence threshold, the final identification result is determined to be an alarm of the corresponding noise type based on the output noise type; otherwise, the final identification result is rejection.

[0070] The system hardware architecture and data flow design for identifying power transformer fault data in this preferred embodiment adopts a high-performance heterogeneous multi-core processor platform. Intelligent algorithm scheduling is achieved through a dual-core ARM processor, while high-speed data preprocessing is performed using an FPGA programmable logic unit. Specifically, the front end accesses 12 high-precision sensor data streams based on the IEC 61850 standard protocol, acquiring key parameters such as load current, oil temperature gradient, and partial discharge in real time, with a sampling frequency reaching the 10kHz level. The back end transmits the three-level decision results to the station control system via gigabit Ethernet. An innovative dual-bus architecture is adopted to achieve collaborative processing of AXI-Stream high-speed data stream and APB control signals. In particular, a deep pipeline design is used in the data preprocessing within the FPGA, reducing the original data processing latency to the microsecond level. Combined with processor-side verification algorithms, the system end-to-end latency is strictly controlled within 35 milliseconds, fully meeting the real-time response requirements of intelligent substations.

[0071] The system is deployed on the station control layer server using a Docker containerization solution, achieving seamless compatibility with legacy station systems through Modbus and TCP protocols. The FPGA-accelerated design employs dynamic partial reconfiguration technology, automatically loading corresponding algorithm images based on different voltage levels such as 220kV, 500kV, and 750kV. The physical parameter library supports online hot updates; when parameters change after equipment maintenance, new calibration data can be remotely injected via an HTTPS encrypted channel. Verified by the East China Power Grid, this system reduces redundant alarms by 62.3% and improves fault location accuracy by 99.3%, providing smart substations with core pre-processing capabilities that combine IEC 61850 protocol compliance, millisecond-level real-time performance, and industrial-grade reliability.

[0072] The system for identifying power transformer fault data described in this preferred embodiment and the method for identifying power transformer fault data have the same steps in outputting the three types of results for power transformer fault data identification through a three-level identification chain collaborative mechanism, and achieve the same technical effects, so they will not be described again here.

[0073] Exemplary electronic devices

[0074] Figure 3 This is a schematic diagram of the structure of an electronic device according to a preferred embodiment of the present invention. Figure 3 As shown, the electronic device includes one or more processors 301 and memory 302.

[0075] The processor 301 may be a central processing unit (CPU) or other form of processing unit with data processing capabilities and / or instruction execution capabilities, and may control other components in the electronic device to perform desired functions.

[0076] The memory 302 may include one or more computer program products, which may include various forms of computer-readable storage media, such as volatile memory and / or non-volatile memory. The volatile memory may include, for example, random access memory (RAM) and / or cache memory. The non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc. One or more computer program instructions may be stored on the computer-readable storage medium, and the processor 301 may execute the program instructions to implement the methods for identifying power transformer fault data of the various disclosed embodiments described above, and / or other desired functions. In one example, the electronic device may also include an input device 303 and an output device 304, these components being interconnected via a bus system and / or other forms of connection mechanisms (not shown).

[0077] In addition, the input device 303 may also include, for example, a keyboard, a mouse, etc.

[0078] The output device 304 can output various information to the outside. The output device 304 may include, for example, a display, a speaker, a printer, and a communication network and its connected remote output devices, etc.

[0079] Of course, for the sake of simplicity, Figure 3 Only some of the components of the electronic device relevant to this disclosure are shown, omitting components such as buses, input / output interfaces, etc. In addition, the electronic device may include any other suitable components depending on the specific application.

[0080] Exemplary computer program products and computer-readable storage media

[0081] In addition to the methods and apparatus described above, embodiments of this disclosure may also be computer program products comprising computer program instructions that, when executed by a processor, cause the processor to perform the steps of the methods for identifying power transformer fault data according to various embodiments of this disclosure as described in the "Exemplary Methods" section of this specification.

[0082] The computer program product can be written in any combination of one or more programming languages ​​to perform the operations of the embodiments of this disclosure. The programming languages ​​include object-oriented programming languages ​​such as Java and C++, as well as conventional procedural programming languages ​​such as C or similar languages. The program code can be executed entirely on a user's computing device, partially on a user's computing device, as a standalone software package, partially on a user's computing device and partially on a remote computing device, or entirely on a remote computing device or server.

[0083] Furthermore, embodiments of this disclosure may also be computer-readable storage media having computer program instructions stored thereon, which, when executed by a processor, cause the processor to perform the steps in the methods for identifying power transformer fault data according to various embodiments of this disclosure as described in the "Exemplary Methods" section above.

[0084] The computer-readable storage medium may be any combination of one or more readable media. A readable medium may be a readable signal medium or a readable storage medium. A readable storage medium may, for example, include, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatuses, or devices, or any combination thereof. More specific examples of readable storage media (a non-exhaustive list) include: electrical connections having one or more wires, portable disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0085] The basic principles of this disclosure have been described above with reference to specific embodiments. However, it should be noted that the advantages, benefits, and effects mentioned in this disclosure are merely examples and not limitations, and should not be considered as essential features of each embodiment of this disclosure. Furthermore, the specific details disclosed above are for illustrative and facilitative purposes only, and are not limitations. These details do not limit the scope of this disclosure to the necessity of employing the aforementioned specific details for implementation.

[0086] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For system embodiments, since they largely correspond to method embodiments, the description is relatively simple; relevant parts can be referred to the descriptions in the method embodiments.

[0087] The block diagrams of devices, apparatuses, devices, and systems disclosed herein are merely illustrative examples and are not intended to require or imply that they must be connected, arranged, or configured in the manner shown in the block diagrams. As those skilled in the art will recognize, these devices, apparatuses, devices, and systems can be connected, arranged, and configured in any manner. Words such as “comprising,” “including,” “having,” etc., are open-ended terms meaning “including but not limited to,” and are used interchangeably with them. The terms “or” and “and” as used herein refer to the terms “and / or,” and are used interchangeably with them unless the context clearly indicates otherwise. The term “such as” as used herein refers to the phrase “such as but not limited to,” and is used interchangeably with it.

[0088] The apparatus and methods of this disclosure may be implemented in many ways. For example, they may be implemented by software, hardware, firmware, or any combination of software, hardware, and firmware. The above-described order of steps for the methods is for illustrative purposes only, and the steps of the methods of this disclosure are not limited to the order specifically described above unless otherwise specifically stated. Furthermore, in some embodiments, this disclosure may also be implemented as a program recorded on a recording medium, the program including machine-readable instructions for implementing the methods according to this disclosure. Thus, this disclosure also covers recording media storing programs for performing the methods according to this disclosure.

[0089] It should also be noted that in the apparatus, devices, and methods of this disclosure, the components or steps are decomposable and / or recombinable. Such decomposition and / or recombination should be considered equivalent to the present disclosure. The above description of the disclosed aspects is provided to enable any person skilled in the art to make or use this disclosure. Various modifications to these aspects will be readily apparent to those skilled in the art, and the general principles defined herein can be applied to other aspects without departing from the scope of this disclosure. Therefore, this disclosure is not intended to be limited to the aspects shown herein, but rather to be carried out within the widest scope consistent with the principles and novel features disclosed herein.

[0090] The above description has been given for purposes of illustration and description. Furthermore, this description is not intended to limit the embodiments of this disclosure to the forms disclosed herein. Although numerous exemplary aspects and embodiments have been discussed above, those skilled in the art will recognize certain variations, modifications, alterations, additions, and sub-combinations therein.

Claims

1. A method for identifying fault data in power transformers, characterized in that, The method includes: Online monitoring data of multi-dimensional operating parameters of power transformers are obtained from the data input interface; The effective monitoring data generated after preprocessing the online monitoring data is input into the dynamic threshold model for adaptive threshold calculation to obtain the threshold calculation result. The effective monitoring data whose threshold calculation result is not within the set dynamic threshold range is marked as first-level abnormal data. The dynamic threshold model is constructed using the entropy weight method. The first-level abnormal data is subjected to physical constraint dual verification, and the first-level abnormal data that fails at least one verification is marked as second-level abnormal data. The physical constraint dual verification includes physical parameter constraint verification through a device feature characteristic mapping library constructed using several IEC standard parameters and signal step feature verification using a multi-scale edge detection algorithm. The feature vector generated by feature extraction from the secondary abnormal data is input into the noise classification model to identify the noise type. The final identification result is output based on the confidence value and noise type output by the noise classification module. The noise classification model is constructed using a deep belief network based on a three-layer restricted Boltzmann machine. The noise types include impulse noise and periodic noise. The final identification result includes a noise type alarm and a rejection of identification.

2. The method according to claim 1, characterized in that, The method further includes: When the threshold calculation result is within the set dynamic threshold range, the output identification result is that the power transformer is operating normally; or When both physical constraint double verifications of the first-level abnormal data fail, the output identification result is a confirmed power transformer fault alarm.

3. The method according to claim 1, characterized in that, After outputting the final identification result based on the noise type and its corresponding confidence value output by the noise classification module, the system also includes storing secondary abnormal data that are rejected in the final identification result into the historical database, which is used to update the weight of each noise type in the noise classification model according to the set time period.

4. The method according to claim 1, characterized in that, The process involves acquiring online monitoring data of the power transformer from a data input interface. This data input interface supports access to multi-source heterogeneous data, including online monitoring data of oil chromatography transmitted via the IEC 61850 protocol, partial discharge pulse current data synchronized via the IEEE C37.118 protocol, and temperature data transmitted via a temperature sensor network with a customized RS485 interface.

5. The method according to claim 1, characterized in that, The online monitoring data is input into a dynamic threshold model for adaptive threshold calculation to obtain the threshold calculation result. The expression for the dynamic threshold model is: In the formula, H t For dynamic threshold, P i,t Let be the measured value of the i-th dimension parameter of the power transformer collected at time t, and N be the total number of dimensions of the power transformer collected, μ i and σ i It is the dynamic mean and standard deviation of the i-th dimension parameter, calculated based on historical data of power transformers; the weighting coefficient ω i It is a constant that is dynamically allocated using the entropy weighting method.

6. The method according to claim 1, characterized in that, The first-level anomaly data is subjected to dual physical constraint verification. This dual physical constraint verification includes physical parameter constraint verification using a device characteristic mapping library constructed from several IEC standard parameters, and signal step characteristic verification using a multi-scale edge detection algorithm. When performing physical parameter constraint verification, the core function of the equipment characteristic mapping library adopts the voltage-insulation thickness correlation equation, the expression of which is: R=α*lnV+βD i In the formula, R is the associated characteristic value, V is the voltage level in kilovolts, and D is the voltage level in kilovolts. i The insulation thickness is measured in millimeters. The coefficients α and β are determined by fitting several sets of factory test data using the least squares method. When the associated characteristic value R calculated based on the obtained power transformer voltage level and insulation thickness exceeds the set associated characteristic value range, it is determined that the physical parameter constraint verification has failed. When using a multi-scale edge detection algorithm to verify signal step features, the Daubechies wavelet-based multi-scale edge detection algorithm is used to calculate the step duration and the frequency domain energy entropy change value. When the step time exceeds the step time range of the corresponding voltage level, or the energy entropy change value is less than a custom threshold, the signal step feature verification is determined to fail.

7. The method according to claim 1, characterized in that, The feature vectors generated by feature extraction from the secondary anomaly data are input into the constructed noise classification model for noise type identification. The final identification result is then output based on the confidence value and noise type output by the noise classification module, where: When the confidence value is not less than the preset confidence threshold, the final identification result is determined to be an alarm of the corresponding noise type based on the output noise type; otherwise, the final identification result is rejection.

8. A system for identifying fault data of power transformers, characterized in that, The system includes: The data acquisition module is used to acquire online monitoring data of multi-dimensional operating parameters of power transformers from the data input interface; The first identification module is used to input the effective monitoring data generated after preprocessing the online monitoring data into the dynamic threshold model for adaptive threshold calculation, obtain the threshold calculation result, and mark the effective monitoring data whose threshold calculation result is not within the set dynamic threshold range as first-level abnormal data. The dynamic threshold model is constructed using the entropy weight method. The second identification module is used to perform physical constraint dual verification on the first-level abnormal data and mark the first-level abnormal data that fails at least one verification as second-level abnormal data. The physical constraint dual verification includes physical parameter constraint verification through a device feature characteristic mapping library constructed using several IEC standard parameters and signal step feature verification using a multi-scale edge detection algorithm. The third identification module is used to input the feature vector generated by feature extraction from the secondary abnormal data into the noise classification model to identify the noise type, and output the final identification result based on the confidence value and noise type output by the noise classification module. The noise classification model is constructed using a deep belief network based on a three-layer restricted Boltzmann machine. The noise type includes impulse noise and periodic noise. The final identification result includes a noise type alarm and a rejection of identification.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the steps of the method as described in any one of claims 1-7.

10. An electronic device, characterized in that, include: processor; Memory used to store the processor's executable instructions; The processor is configured to read the executable instructions from the memory and execute the instructions to implement the method described in any one of claims 1-7.

Images