Transformer operation monitoring method and device, electronic equipment, medium and product

Abstract

The application relates to the technical field of transformer detection, in particular to a transformer operation monitoring method and device, electronic equipment, medium and product, the method comprising: collecting multi-dimensional state parameters in the current transformer operation process; performing feature extraction on the multi-dimensional state parameters to obtain initial multi-dimensional features, and performing feature fusion on the initial multi-dimensional features to obtain a fusion feature vector; inputting the fusion feature vector into a preset transformer state evaluation model to obtain a comprehensive state evaluation result of the current transformer, wherein the preset transformer state evaluation model is obtained by training a target neural network. Thus, the problem that fault diagnosis is not comprehensive and low in accuracy caused by the isolation of existing transformer monitoring means and insufficient data fusion is solved, intelligent comprehensive monitoring and early accurate early warning based on multi-dimensional parameter fusion are realized, and the reliability and operation and maintenance efficiency of transformer state evaluation are improved.

Description

Technical Field

[0001] This application relates to the field of transformer testing technology, and in particular to a transformer operation monitoring method, device, electronic equipment, medium and product. Background Technology

[0002] Power transformers are core high-voltage equipment in power supply and distribution systems, and their operating status directly affects the reliability and security of the power grid. Due to the multiple stresses from electricity, heat, and mechanical factors during long-term operation, transformers are prone to various faults such as partial discharge, insulation aging, winding deformation, and abnormal core grounding. In severe cases, these faults can lead to equipment breakdown, explosions, and other accidents, causing significant economic losses and power outages. Therefore, continuous and accurate monitoring and assessment of transformer operating status is crucial for achieving fault early warning and improving operation and maintenance levels.

[0003] In related technologies, current online transformer monitoring technologies have various independent monitoring methods, analyzing each parameter separately. These methods typically collect and analyze single parameters, lacking the coordination and fusion of multi-dimensional data, making it difficult to comprehensively reflect the overall operating status of the transformer. In practical applications, because the monitoring systems are independent and the data is heterogeneous, they are susceptible to environmental interference and fluctuations in operating conditions, leading to biased fault diagnosis, high false alarm rates, and difficulty in providing accurate and integrated decision support for maintenance personnel. These issues urgently need to be addressed. Summary of the Invention

[0004] This application provides a transformer operation monitoring method, device, electronic equipment, medium, and product to solve the problems of incomplete fault diagnosis and low accuracy caused by the isolation of existing transformer monitoring methods and insufficient data fusion. It realizes intelligent comprehensive monitoring and early accurate warning based on multi-dimensional parameter fusion, thereby improving the reliability of transformer condition assessment and operation and maintenance efficiency.

[0005] The first aspect of this application provides a transformer operation monitoring method, including the following steps: Collect multi-dimensional state parameters during the current operation of the transformer; Feature extraction is performed on the multidimensional state parameters to obtain initial multidimensional features, and feature fusion is performed on the initial multidimensional features to obtain a fused feature vector; The fused feature vector is input into a preset transformer condition assessment model to obtain the comprehensive condition assessment result of the current transformer, wherein the preset transformer condition assessment model is obtained by training a target neural network.

[0006] According to one embodiment of this application, the multidimensional state parameters include at least one of the following: dissolved gas concentration in transformer oil, transformer partial discharge signal, transformer core grounding current, and transformer body vibration signal.

[0007] According to one embodiment of this application, the step of extracting features from the multidimensional state parameters to obtain initial multidimensional features includes: If the multidimensional state parameters include the dissolved gas concentration in the transformer oil, then the concentrations of multiple target gases are calculated based on the preset gas ratio method to obtain at least one set of gas ratios, and the multiple target gas concentrations and the gas ratios are used as the first type of initial features; If the multidimensional state parameters include the transformer partial discharge signal, then statistical features are extracted from the phase-resolved partial discharge spectrum corresponding to the partial discharge signal as the second type of initial features; If the multidimensional state parameters include the transformer core grounding current, then the effective value, total harmonic distortion rate and harmonic content of the transformer core grounding current are calculated as the third type of initial features. If the multidimensional state parameters include the transformer body vibration signal, then at least one of the following is calculated as the effective value of the transformer body vibration signal, the energy proportion of a specific fault frequency band, and the kurtosis index, as the fourth type of initial feature; The initial multidimensional features are obtained based on the first type of initial features, the second type of initial features, the third type of initial features, and the fourth type of initial features.

[0008] According to one embodiment of this application, the step of inputting the fused feature vector into a preset transformer condition assessment model to obtain the comprehensive condition assessment result of the current transformer includes: The fused feature vectors are respectively input into the unsupervised anomaly detection submodule and the supervised fault diagnosis submodule of the transformer condition assessment model for parallel processing; Using the unsupervised anomaly detection submodule, the reconstruction error is calculated based on the fused feature vector, and a first initial evaluation result is obtained based on the reconstruction error; The fused feature vector is classified using the supervised fault diagnosis submodule to obtain a second initial evaluation result, which includes the fault type of the current transformer and the corresponding confidence probability. The comprehensive state assessment result is generated based on the first initial assessment result and the second initial assessment result.

[0009] According to one embodiment of this application, the comprehensive condition assessment result includes a health score and a fault type probability. After inputting the fused feature vector into a preset transformer condition assessment model to obtain the comprehensive condition assessment result of the current transformer, the method further includes: The target warning level is determined based on the health score and the probability of the fault type. A diagnostic report is generated based on the target warning level. The diagnostic report includes at least one of the following: target abnormal parameters, fault type, confidence level of the fault type, fault location, and maintenance recommendations.

[0010] According to the transformer operation monitoring method provided in this application, features are extracted from the multi-dimensional state parameters during the current transformer operation to obtain initial multi-dimensional features. These features are then fused to obtain a fused feature vector, which is input into a preset transformer state assessment model to obtain a comprehensive state assessment result for the current transformer. This solves the problems of incomplete fault diagnosis and low accuracy caused by isolated existing transformer monitoring methods and insufficient data fusion. It achieves intelligent comprehensive monitoring and early accurate warning based on multi-dimensional parameter fusion, improving the reliability of transformer state assessment and operation and maintenance efficiency.

[0011] A second aspect of this application provides a transformer operation monitoring device, comprising: The data acquisition module is used to collect multi-dimensional state parameters during the current operation of the transformer. The processing module is used to extract features from the multidimensional state parameters to obtain initial multidimensional features, and to fuse the initial multidimensional features to obtain a fused feature vector. The monitoring module is used to input the fused feature vector into a preset transformer condition assessment model to obtain the comprehensive condition assessment result of the current transformer, wherein the preset transformer condition assessment model is obtained by training a target neural network.

[0012] According to one embodiment of this application, the multidimensional state parameters include at least one of the following: dissolved gas concentration in transformer oil, transformer partial discharge signal, transformer core grounding current, and transformer body vibration signal.

[0013] According to one embodiment of this application, the processing module is configured to: If the multidimensional state parameters include the dissolved gas concentration in the transformer oil, then the concentrations of multiple target gases are calculated based on the preset gas ratio method to obtain at least one set of gas ratios, and the multiple target gas concentrations and the gas ratios are used as the first type of initial features; If the multidimensional state parameters include the transformer partial discharge signal, then statistical features are extracted from the phase-resolved partial discharge spectrum corresponding to the partial discharge signal as the second type of initial features; If the multidimensional state parameters include the transformer core grounding current, then the effective value, total harmonic distortion rate and harmonic content of the transformer core grounding current are calculated as the third type of initial features. If the multidimensional state parameters include the transformer body vibration signal, then at least one of the following is calculated as the effective value of the transformer body vibration signal, the energy proportion of a specific fault frequency band, and the kurtosis index, as the fourth type of initial feature; The initial multidimensional features are obtained based on the first type of initial features, the second type of initial features, the third type of initial features, and the fourth type of initial features.

[0014] According to one embodiment of this application, the monitoring module is used for: The fused feature vectors are respectively input into the unsupervised anomaly detection submodule and the supervised fault diagnosis submodule of the transformer condition assessment model for parallel processing; Using the unsupervised anomaly detection submodule, the reconstruction error is calculated based on the fused feature vector, and a first initial evaluation result is obtained based on the reconstruction error; The fused feature vector is classified using the supervised fault diagnosis submodule to obtain a second initial evaluation result, which includes the fault type of the current transformer and the corresponding confidence probability. The comprehensive state assessment result is generated based on the first initial assessment result and the second initial assessment result.

[0015] According to one embodiment of this application, the comprehensive condition assessment result includes a health score and a fault type probability. After inputting the fused feature vector into a preset transformer condition assessment model to obtain the comprehensive condition assessment result of the current transformer, the monitoring module is further configured to: The target warning level is determined based on the health score and the probability of the fault type. A diagnostic report is generated based on the target warning level. The diagnostic report includes at least one of the following: target abnormal parameters, fault type, confidence level of the fault type, fault location, and maintenance recommendations.

[0016] According to the transformer operation monitoring device provided in this application embodiment, features are extracted from the multi-dimensional state parameters during the current transformer operation to obtain initial multi-dimensional features. These features are then fused to obtain a fused feature vector, which is input into a preset transformer state assessment model to obtain the comprehensive state assessment result of the current transformer. This solves the problems of incomplete fault diagnosis and low accuracy caused by isolated existing transformer monitoring methods and insufficient data fusion. It achieves intelligent comprehensive monitoring and early accurate warning based on multi-dimensional parameter fusion, improving the reliability of transformer state assessment and operation and maintenance efficiency.

[0017] A third aspect of this application provides an electronic device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the transformer operation monitoring method as described in the above embodiments.

[0018] A fourth aspect of this application provides a computer-readable storage medium storing computer instructions for causing the computer to execute the transformer operation monitoring method as described in the above embodiments.

[0019] A fifth aspect of this application provides a computer program product, including a computer program that, when executed by a processor, implements the transformer operation monitoring method as described in the above embodiments.

[0020] Additional aspects and advantages of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application. Attached Figure Description

[0021] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein: Figure 1 This is a flowchart of a transformer operation monitoring method provided according to an embodiment of this application; Figure 2 This is a flowchart illustrating the operation of a comprehensive monitoring model for transformer operation according to an embodiment of this application. Figure 3 This is a block diagram of a transformer operation monitoring device according to an embodiment of this application; Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0022] The embodiments of this application are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this application, and should not be construed as limiting this application.

[0023] As those skilled in the art will understand, current online transformer monitoring devices generally employ methods such as partial discharge detection, dissolved gas analysis in oil (oil chromatography), and core grounding current monitoring, each independently assessing the transformer's operating status for anomalies. These methods typically rely on a single parameter for judgment and have not yet formed a comprehensive diagnostic system that effectively integrates the aforementioned multi-source monitoring methods. Because the acquisition and analysis of each parameter are fragmented and susceptible to environmental interference, load fluctuations, and changes in operating conditions, the assessment results are often one-sided and uncertain. In practical applications, field testing personnel find it difficult to make comprehensive and accurate fault diagnoses based on isolated and scattered data, thus limiting the reliability and intelligence level of condition-based maintenance.

[0024] Based on the technical problems existing in the above-mentioned related technologies, this application establishes a comprehensive diagnostic platform for transformer partial discharge, grounding current, oil chromatography and vibration monitoring devices. By preprocessing the collected data, effective data for transformer fault analysis is screened, and by comprehensively analyzing the effective data, accurate abnormal transformer conditions can be obtained. This platform can quickly, efficiently and comprehensively detect transformer faults and potential faults.

[0025] The following description, with reference to the accompanying drawings, describes a transformer operation monitoring method, apparatus, electronic device, medium, and product according to embodiments of this application.

[0026] Specifically, Figure 1 This is a flowchart illustrating a transformer operation monitoring method provided in an embodiment of this application.

[0027] like Figure 1 As shown, the transformer operation monitoring method includes the following steps: In step S101, multi-dimensional state parameters of the current transformer operation process are collected.

[0028] In some embodiments, the multidimensional state parameters include at least one of the following: dissolved gas concentration in transformer oil, transformer partial discharge signal, transformer core grounding current, and transformer body vibration signal.

[0029] Specifically, the embodiments of this application employ gas chromatography analysis technology combined with vacuum degassing to separate dissolved gases in the oil. Based on a MEMS integrated sensor and a self-made composite chromatographic column, the concentrations of gases such as H2, CO, CH4, C2H6, C2H4, and C2H2 are separated and detected, with errors controlled within ±10%. A cyclic sampling mode is supported to avoid oil loss, and a multi-stage automatic gas replenishment device replaces the carrier gas cylinder, achieving true maintenance-free operation.

[0030] Optionally, embodiments of this application can use an ultra-high frequency sensor (300MHz-3GHz band) to capture electromagnetic wave signals generated by partial discharge inside the transformer. An anti-interference algorithm is used to suppress external noise, display the discharge spectrum in real time, and record the waveforms before and after the alarm (one power frequency cycle) and the discharge characteristics within 30 minutes. The discharge type is identified by combining parameters such as discharge amplitude and frequency.

[0031] Optionally, embodiments of this application may use a high-precision active zero-flux current sensor to acquire the core grounding current value in real time, with a resolution down to the μA level. A trend analysis algorithm is used to determine sudden current changes or exceedances, identify grounding faults, and upload data in real time, ensuring the timeliness and responsiveness of the monitoring data.

[0032] Optionally, embodiments of this application may also employ a horn-shaped sound-gathering cover, a noise sensor, and an address encoder bound to the noise sensor. A sound-gathering hole is provided on the rear end face of the horn cover directly opposite the horn opening of the sound-gathering cover, and the diameter of the horn opening of the sound-gathering cover is larger than the diameter of the sound-gathering hole. The edge of the horn opening of the sound-gathering cover is sealed and adsorbed or fixedly connected to the outer surface of the oil tank. The noise sensor is set inside the sound-gathering hole of the sound-gathering cover. The address encoder and the noise sensor are respectively connected to the controller to collect the vibration signal of the transformer body during the current operation of the transformer.

[0033] In step S102, feature extraction is performed on the multidimensional state parameters to obtain initial multidimensional features, and feature fusion is performed on the initial multidimensional features to obtain a fused feature vector.

[0034] Further, in some embodiments, feature extraction is performed on the multidimensional state parameters to obtain initial multidimensional features, including: if the multidimensional state parameters include the dissolved gas concentration in the transformer oil, then the concentrations of multiple target gases are calculated based on a preset gas ratio method to obtain at least one set of gas ratios, and the multiple target gas concentrations and gas ratios are used as the first type of initial features; if the multidimensional state parameters include the transformer partial discharge signal, then statistical features are extracted from the phase-resolved partial discharge spectrum corresponding to the partial discharge signal as the second type of initial features; if the multidimensional state parameters include the transformer core grounding current, then the effective value, total harmonic distortion rate, and harmonic content of the transformer core grounding current are calculated as the third type of initial features; if the multidimensional state parameters include the transformer body vibration signal, then at least one of the effective value, energy proportion of a specific fault frequency band, and kurtosis index of the transformer body vibration signal is calculated as the fourth type of initial features; the initial multidimensional features are obtained based on the first type of initial features, the second type of initial features, the third type of initial features, and the fourth type of initial features.

[0035] Specifically, in the feature extraction stage, this invention first performs structured processing on various multidimensional state parameters to generate an initial multidimensional feature set with characterization capabilities. If the monitored state parameters include the concentration of dissolved gases in transformer oil, the concentrations of target gases such as H2, CO, CO2, CH4, C2H6, C2H4, C2H2, and total hydrocarbons (TH) are combined and calculated based on the gas ratio method to obtain multiple sets of gas ratio features. At the same time, the original concentrations of these gases and their calculated ratios are used together as the first type of initial features, thereby simultaneously preserving the original information and the combination relationships used for fault diagnosis in industry standards.

[0036] If the monitored parameters include transformer partial discharge signals, then physically meaningful statistical features are extracted from the phase-resolved partial discharge spectrum acquired and generated by the partial discharge sensor. Examples include skewness, kurtosis, cross-correlation coefficient, discharge asymmetry, and Weibull distribution parameters. Alternatively, CNNs can be used to automatically extract PRPD image features; these features are categorized as second-class initial features and used to quantify the intensity, pattern, and statistical regularity of the discharge.

[0037] For the parameter of transformer core grounding current, the extraction process includes calculating the effective value of its current signal, the total harmonic distortion rate, and the ratio of each harmonic (especially odd harmonics) to the fundamental frequency. For example, the relative amplitudes of the 3rd, 5th, and 7th harmonics can be given special attention as key indicators for identifying multi-point grounding or magnetic circuit anomalies in the core, constituting the third type of initial feature.

[0038] If the monitored object includes the vibration signal of the transformer body, the effective value (RMS), peak value, peak-to-peak value, kurtosis, amplitude of specific frequency bands (such as bearing fault frequency band, gear meshing frequency band), energy ratio, center of gravity frequency, etc. are extracted. These features are summarized as the fourth type of initial features.

[0039] Furthermore, the above four types of initial features are integrated to form a unified initial multidimensional feature vector, thereby realizing the standardized structured transformation from multi-source heterogeneous monitoring data.

[0040] Furthermore, cross-modal feature fusion is performed, including the following steps: Early fusion: Concatenates all extracted features into a single massive feature vector. Simple, but may suffer from the "curse of dimensionality" and redundancy.

[0041] Intermediate fusion: First, a sub-neural network is built for each data source to extract features and reduce dimensionality. Then, these high-order feature representations are concatenated or weighted and fused.

[0042] Late-stage fusion: Each data source is judged independently first, and then the judgment results are voted on or combined.

[0043] In step S103, the fused feature vector is input into the preset transformer state assessment model to obtain the comprehensive state assessment result of the current transformer. The preset transformer state assessment model is obtained by training the target neural network.

[0044] Furthermore, in some embodiments, the fused feature vector is input into a preset transformer condition assessment model to obtain a comprehensive condition assessment result of the current transformer. This includes: inputting the fused feature vector into an unsupervised anomaly detection submodule and a supervised fault diagnosis submodule of the transformer condition assessment model for parallel processing; using the unsupervised anomaly detection submodule to calculate the reconstruction error based on the fused feature vector, and obtaining a first initial assessment result based on the reconstruction error; using the supervised fault diagnosis submodule to classify the fused feature vector to obtain a second initial assessment result, the second initial assessment result including the fault type of the current transformer and the corresponding confidence probability; and generating a comprehensive condition assessment result based on the first initial assessment result and the second initial assessment result.

[0045] In this embodiment, the unsupervised anomaly detection submodule learns the fusion feature representation of multi-source data under normal conditions. When new data is input, its reconstruction error is calculated. An excessively large error indicates an anomaly that cannot be understood by existing models and can be used to discover unknown faults.

[0046] The supervised fault diagnosis submodule in this application embodiment can be a deep learning classifier, and the backbone can be a fully connected network, a Transformer, or a graph neural network. The input to the supervised fault diagnosis submodule is the fused feature vector, and the output is the fault type (e.g., overheating, discharge, dampness, etc.) and confidence probability. Simultaneously, the model should be able to output interpretable results (e.g., using an attention mechanism to show which features contribute most to the current judgment).

[0047] In addition, embodiments of this application can also utilize the health index prediction and degradation trend prediction module (LSTM / Transformer + regression model) to input time series data into the LSTM or Transformer model, predict feature values ​​(such as C2H2 content and PD amplitude) at future time points, and then calculate the health index (HI) and remaining useful life (RUL).

[0048] In this application, physical mechanisms (such as "C2H2 is a characteristic gas of discharge" and "increased IC and contain a large number of harmonics indicate that the core is grounded at multiple points") can also be added to the model as constraints or loss functions, so that the AI ​​learning process conforms to physical laws and improves generalization ability and interpretability.

[0049] Furthermore, in some embodiments, the comprehensive condition assessment result includes a health score and a fault type probability. After inputting the fused feature vector into a preset transformer condition assessment model to obtain the comprehensive condition assessment result of the current transformer, the method further includes: determining the target warning level based on the health score and the fault type probability; generating a diagnostic report based on the target warning level, wherein the diagnostic report includes at least one of the following: target abnormal parameters, fault type, confidence level of the fault type, fault location, and maintenance recommendations.

[0050] Specifically, the health status score in this application embodiment can be a score from 0 to 100, which intuitively reflects the overall condition of the transformer. The fault diagnosis report clearly indicates the fault type, possible location, severity, and rate of development.

[0051] Optionally, the multi-level early warning mechanism in this application embodiment may include: Level 1 Warning (Attention): Slight abnormality in a single indicator; Level 2 warning: Multiple indicators are abnormal and conform to a certain failure mode; Level 3 Warning (Danger): Indicators are severely exceeding limits, the fault mode is clear, and immediate action is required; Maintenance decision recommendations: Based on the risk assessment results, we recommend "continue to observe", "planned maintenance" or "immediate shutdown".

[0052] The following section provides a detailed description of the pre-defined transformer condition assessment model in this application.

[0053] The pre-defined transformer condition assessment model includes a data layer, a feature layer, a model layer, and an application layer. Among them, for example... Figure 2 As shown, the data layer is used for multi-source heterogeneous data acquisition and management. The feature layer performs feature engineering and fusion representation. The model layer performs core algorithms and model construction. The application layer performs state evaluation and decision support. It should be noted that the model layer in this embodiment does not refer to a simple LLM, but rather to a complex deep neural network model with numerous parameters, which can employ a multi-task learning framework.

[0054] Specifically, the data layer preprocesses parameters across various dimensions. When the parameter is oil chromatography data (including the content and production rate of gases such as H2, CH4, C2H2, C2H4, C2H6, CO, CO2, and total hydrocarbons (TH), the oil chromatography data undergoes data cleaning (outlier removal), normalization, and calculation of absolute and relative production rates. When the parameter is partial discharge data (including discharge amplitude, discharge phase, discharge frequency, and PRPD spectrum (phase-amplitude-frequency three-dimensional spectrum)), the partial discharge data undergoes noise reduction processing (wavelet transform, deep learning denoising) and PRPD spectrum feature extraction (statistical features, graphical features, fractal features, etc.). When the parameter is core grounding current data (including current RMS value and current waveform (including harmonic components)), the core grounding current data is filtered, and the fundamental frequency and harmonic content (especially odd harmonics) are calculated. When the parameters are transformer vibration data (including vibration amplitude, vibration period, vibration wavelength and vibration frequency, etc.), the transformer vibration data is verified, cleaned, conditioned and denoised, calibrated and shaped, and features are initially extracted.

[0055] However, since the sampling frequencies of the four types of data are completely different, time series alignment is required. Typically, the lowest frequency DGA data is used as the reference point to construct a "time slice" and aggregate the features of PD and IC (such as maximum value, mean, variance, etc.) within the slice.

[0056] Furthermore, feature extraction is performed on single data sources to obtain features from oil chromatography data, partial discharge data, core grounding current data, and transformer vibration data. The extracted features are then fused to generate a unified fused feature vector. Subsequently, at the core model layer, an unsupervised module is used for anomaly detection, a supervised module for fault diagnosis, and a prediction module for health status assessment. The accuracy of decision-making is improved by combining a knowledge base and a mechanism model. Finally, at the application output layer, health status scores, fault type identification and probability, risk warning levels, remaining useful life (RUL) prediction, and maintenance strategy suggestions are provided, realizing a closed loop from multi-source data to intelligent operation and maintenance decision-making.

[0057] The following describes the multi-dimensional parameter-based transformer operation integrated monitoring system of this application.

[0058] The system consists of four types of equipment: an online monitoring module for dissolved gases in oil, an online monitoring module for ultra-high frequency partial discharge of transformers, an online monitoring module for iron core grounding current, and an online monitoring module for transformer vibration, as well as an integrated monitoring IED. Its technical approach mainly revolves around four core monitoring modules: dissolved gases in oil, ultra-high frequency partial discharge, iron core grounding current, and transformer vibration. Through multi-dimensional data acquisition, intelligent analysis, and system linkage, it achieves a comprehensive condition assessment of the transformer.

[0059] For the monitoring of dissolved gases in oil, this application develops a rapid piston vacuum degassing technology to improve the degassing efficiency to over 95%, enabling the separation of all components within 40 minutes. It also develops a multi-stage automatic gas replenishment device to achieve stable output of carrier gas. The key feature is that it can automatically replenish gas at all stages without affecting sampling. Furthermore, it develops a dedicated chromatographic column mixed packing material to increase peak elution rate and separation effect, meeting the usage requirements of a one-hour sampling cycle.

[0060] For UHF partial discharge monitoring, this application develops an UHF sensor with directional reception enhancement to amplify the signal at the sensor's receiving surface while reducing sensor sensitivity in other directions, greatly improving the sensor's field usability and avoiding spatial interference from other directions. It achieves full-band coverage of the discharge signal, develops an anti-interference algorithm, and suppresses radio noise through a band-stop filter, improving the signal-to-noise ratio by 40%. Simultaneously, it constructs a discharge spectrum feature library to automatically identify six types of defects, including floating discharge and surface discharge.

[0061] For core grounding current monitoring, this application employs high-precision detection technology, developing a μA-level resolution current sensor that supports a wide dynamic range measurement of 0.1mA-10A. Based on a temperature compensation algorithm, it eliminates the influence of ambient temperatures ranging from -40℃ to 85℃ on measurement accuracy. Simultaneously, an adaptive threshold algorithm is developed to dynamically adjust the alarm threshold based on historical data, reducing the false alarm rate to below 5%.

[0062] For transformer vibration monitoring, this application establishes a vibration "baseline" for the transformer under normal conditions, studies the attenuation, resonance, and modal characteristics of vibration signals in the structure, and develops a vibration spectrum feature library, a vibration-electrical quantity relationship model, and standardized measurement point layout guidelines under normal conditions. Furthermore, it analyzes simulated / historical fault data, establishing a mapping relationship between fault characteristics and specific fault types (such as winding deformation, inter-turn short circuits, and multi-point grounding of the core) based on experimental platform simulations (e.g., winding loosening, deformation, insufficient core clamping force) or collected historical fault cases. Simultaneously, it develops sensor selection and optimized layout, an anti-interference data acquisition unit, and an edge computing unit for real-time data acquisition, storage, preprocessing (as discussed in the first step), and automatic feature calculation.

[0063] Therefore, this invention provides a fully automated transformer operation monitoring system based on multi-dimensional parameters, eliminating the need for manual intervention and significantly reducing maintenance manpower costs. Simultaneously, the system's early warning function effectively reduces the probability of sudden equipment failures, with an estimated failure rate reduction of approximately 30%, resulting in annual maintenance cost savings of 150,000 to 300,000 yuan for a single substation. Furthermore, early intervention in faults can delay equipment aging and reduce lifespan loss due to faults (approximately 20%), thereby lowering depreciation costs associated with frequent equipment replacements. Finally, by implementing proactive fault management, the system reduces the frequency of power outage repairs, significantly improving grid power continuity and providing sustained guarantees for socio-economic benefits such as industrial production.

[0064] The transformer operation monitoring method proposed in this application extracts features from the multi-dimensional state parameters during the current transformer operation process to obtain initial multi-dimensional features. These features are then fused to obtain a fused feature vector, which is input into a preset transformer state assessment model to obtain a comprehensive state assessment result for the current transformer. This solves the problems of incomplete fault diagnosis and low accuracy caused by isolated existing transformer monitoring methods and insufficient data fusion. It achieves intelligent comprehensive monitoring and early accurate warning based on multi-dimensional parameter fusion, improving the reliability of transformer state assessment and operation and maintenance efficiency.

[0065] Next, the transformer operation monitoring device proposed according to the embodiments of this application is described with reference to the accompanying drawings.

[0066] Figure 3 This is a block diagram of a transformer operation monitoring device according to an embodiment of this application.

[0067] like Figure 3 As shown, the transformer operation monitoring device 10 includes: a data acquisition module 100, a processing module 200, and a monitoring module 300.

[0068] The acquisition module 100 is used to acquire multi-dimensional state parameters during the current operation of the transformer; the processing module 200 is used to extract features from the multi-dimensional state parameters to obtain initial multi-dimensional features, and to fuse the initial multi-dimensional features to obtain a fused feature vector; the monitoring module 300 is used to input the fused feature vector into a preset transformer state assessment model to obtain the comprehensive state assessment result of the current transformer, wherein the preset transformer state assessment model is obtained by training a target neural network.

[0069] Furthermore, in some embodiments, the multidimensional state parameters include at least one of the following: dissolved gas concentration in transformer oil, transformer partial discharge signal, transformer core grounding current, and transformer body vibration signal.

[0070] Further, in some embodiments, the processing module 200 is configured to: if the multidimensional state parameters include the dissolved gas concentration in the transformer oil, calculate the concentrations of multiple target gases based on a preset gas ratio method to obtain at least one set of gas ratios, and use the multiple target gas concentrations and gas ratios as a first type of initial feature; if the multidimensional state parameters include the transformer partial discharge signal, extract statistical features from the phase-resolved partial discharge spectrum corresponding to the partial discharge signal as a second type of initial feature; if the multidimensional state parameters include the transformer core grounding current, calculate the effective value of the transformer core grounding current, the total harmonic distortion rate, and the content of each harmonic as a third type of initial feature; if the multidimensional state parameters include the transformer body vibration signal, calculate at least one of the effective value of the transformer body vibration signal, the energy proportion of a specific fault frequency band, and the kurtosis index as a fourth type of initial feature; and obtain initial multidimensional features based on the first type of initial feature, the second type of initial feature, the third type of initial feature, and the fourth type of initial feature.

[0071] Furthermore, in some embodiments, the monitoring module 300 is configured to: input the fused feature vector into the unsupervised anomaly detection submodule and the supervised fault diagnosis submodule of the transformer condition assessment model for parallel processing; use the unsupervised anomaly detection submodule to calculate the reconstruction error based on the fused feature vector, and obtain a first initial assessment result based on the reconstruction error; use the supervised fault diagnosis submodule to classify the fused feature vector to obtain a second initial assessment result, the second initial assessment result including the current transformer fault type and the corresponding confidence probability; and generate a comprehensive condition assessment result based on the first initial assessment result and the second initial assessment result.

[0072] Furthermore, in some embodiments, the comprehensive condition assessment result includes a health score and a fault type probability. After inputting the fused feature vector into a preset transformer condition assessment model to obtain the current comprehensive condition assessment result of the transformer, the monitoring module 300 is further used for: The target warning level is determined based on the health score and the probability of the fault type. A diagnostic report is generated based on the target warning level. The diagnostic report includes at least one of the following: target abnormal parameters, fault type, confidence level of fault type, fault location, and maintenance recommendations.

[0073] It should be noted that the foregoing explanation of the embodiment of the transformer operation monitoring method also applies to the transformer operation monitoring device of this embodiment, and will not be repeated here.

[0074] The transformer operation monitoring device proposed in this application extracts features from multi-dimensional state parameters during the current transformer operation to obtain initial multi-dimensional features. These features are then fused to obtain a fused feature vector, which is input into a preset transformer state assessment model to obtain a comprehensive state assessment result for the current transformer. This solves the problems of incomplete fault diagnosis and low accuracy caused by isolated existing transformer monitoring methods and insufficient data fusion. It achieves intelligent comprehensive monitoring and early accurate warning based on multi-dimensional parameter fusion, improving the reliability of transformer state assessment and operation and maintenance efficiency.

[0075] Figure 4 A schematic diagram of the structure of an electronic device provided in an embodiment of this application. The electronic device may include: The memory 401, the processor 402, and the computer program stored on the memory 401 and capable of running on the processor 402.

[0076] When the processor 402 executes the program, it implements the transformer operation monitoring method provided in the above embodiments.

[0077] Furthermore, electronic devices also include: Communication interface 403 is used for communication between memory 401 and processor 402.

[0078] The memory 401 is used to store computer programs that can run on the processor 402.

[0079] Memory 401 may include high-speed RAM memory, and may also include non-volatile memory, such as at least one disk storage device.

[0080] If the memory 401, processor 402, and communication interface 403 are implemented independently, then the communication interface 403, memory 401, and processor 402 can be interconnected via a bus to complete communication between them. The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be categorized into address buses, data buses, control buses, etc. For ease of representation, Figure 4 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0081] Optionally, in a specific implementation, if the memory 401, processor 402, and communication interface 403 are integrated on a single chip, then the memory 401, processor 402, and communication interface 403 can communicate with each other through an internal interface.

[0082] Processor 402 may be a central processing unit (CPU), an application specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of this application.

[0083] This application also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the above-described transformer operation monitoring method.

[0084] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the above-described transformer operation monitoring method.

[0085] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0086] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this application, "N" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0087] Any process or method described in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or more N executable instructions for implementing custom logic functions or processes, and the scope of the preferred embodiments of this application includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as should be understood by those skilled in the art to which embodiments of this application pertain.

[0088] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of computer-readable media include: an electrical connection having one or more wires (electronic device), a portable computer disk drive (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Alternatively, the computer-readable medium may be paper or other suitable media on which the program can be printed, since the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in a computer memory.

[0089] It should be understood that the various parts of this application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, the N steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0090] Those skilled in the art will understand that all or part of the steps of the methods in the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, the program includes one or a combination of the steps of the method embodiments.

[0091] Furthermore, the functional units in the various embodiments of this application can be integrated into a processing module, or each unit can exist physically separately, or two or more units can be integrated into a module. The integrated module can be implemented in hardware or as a software functional module. If the integrated module is implemented as a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium.

[0092] The storage medium mentioned above can be a read-only memory, a disk, or an optical disk, etc. Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of this application.

Claims

1. A method for monitoring transformer operation, characterized in that, Includes the following steps: Collect multi-dimensional state parameters during the current operation of the transformer; Feature extraction is performed on the multidimensional state parameters to obtain initial multidimensional features, and feature fusion is performed on the initial multidimensional features to obtain a fused feature vector; The fused feature vector is input into a preset transformer condition assessment model to obtain the comprehensive condition assessment result of the current transformer, wherein the preset transformer condition assessment model is obtained by training a target neural network.

2. The method according to claim 1, characterized in that, The multidimensional state parameters include at least one of the following: dissolved gas concentration in transformer oil, transformer partial discharge signal, transformer core grounding current, and transformer body vibration signal.

3. The method according to claim 2, characterized in that, The step of extracting features from the multidimensional state parameters to obtain initial multidimensional features includes: If the multidimensional state parameters include the dissolved gas concentration in the transformer oil, then the concentrations of multiple target gases are calculated based on the preset gas ratio method to obtain at least one set of gas ratios, and the multiple target gas concentrations and the gas ratios are used as the first type of initial features; If the multidimensional state parameters include the transformer partial discharge signal, then statistical features are extracted from the phase-resolved partial discharge spectrum corresponding to the partial discharge signal as the second type of initial features; If the multidimensional state parameters include the transformer core grounding current, then the effective value, total harmonic distortion rate and harmonic content of the transformer core grounding current are calculated as the third type of initial features. If the multidimensional state parameters include the transformer body vibration signal, then at least one of the following is calculated as the effective value of the transformer body vibration signal, the energy proportion of a specific fault frequency band, and the kurtosis index, as the fourth type of initial feature; The initial multidimensional features are obtained based on the first type of initial features, the second type of initial features, the third type of initial features, and the fourth type of initial features.

4. The method according to claim 1, characterized in that, The step of inputting the fused feature vector into a preset transformer condition assessment model to obtain the comprehensive condition assessment result of the current transformer includes: The fused feature vectors are respectively input into the unsupervised anomaly detection submodule and the supervised fault diagnosis submodule of the transformer condition assessment model for parallel processing; Using the unsupervised anomaly detection submodule, the reconstruction error is calculated based on the fused feature vector, and a first initial evaluation result is obtained based on the reconstruction error; The fused feature vector is classified using the supervised fault diagnosis submodule to obtain a second initial evaluation result, which includes the fault type of the current transformer and the corresponding confidence probability. The comprehensive state assessment result is generated based on the first initial assessment result and the second initial assessment result.

5. The method according to claim 1, characterized in that, The comprehensive condition assessment result includes a health score and a fault type probability. After inputting the fused feature vector into a preset transformer condition assessment model to obtain the comprehensive condition assessment result of the current transformer, it also includes: The target warning level is determined based on the health score and the probability of the fault type. A diagnostic report is generated based on the target warning level. The diagnostic report includes at least one of the following: target abnormal parameters, fault type, confidence level of the fault type, fault location, and maintenance recommendations.

6. A transformer operation monitoring device, characterized in that, include: The data acquisition module is used to collect multi-dimensional state parameters during the current operation of the transformer. The processing module is used to extract features from the multidimensional state parameters to obtain initial multidimensional features, and to fuse the initial multidimensional features to obtain a fused feature vector. The monitoring module is used to input the fused feature vector into a preset transformer condition assessment model to obtain the comprehensive condition assessment result of the current transformer, wherein the preset transformer condition assessment model is obtained by training a target neural network.

7. The apparatus according to claim 6, characterized in that, The multidimensional state parameters include at least one of the following: dissolved gas concentration in transformer oil, transformer partial discharge signal, transformer core grounding current, and transformer body vibration signal.

8. An electronic device, characterized in that, include: A memory, a processor, and a computer program stored in the memory and capable of running on the processor, wherein the processor executes the computer program to implement the transformer operation monitoring method as described in any one of claims 1-5.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, The computer program is executed by a processor to implement the transformer operation monitoring method as described in any one of claims 1-5.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the transformer operation monitoring method as described in any one of claims 1-5.

Images