Intelligent algorithm-based mutual inductor fault diagnosis method and system, and storage medium

Abstract

The application relates to the technical field of fault diagnosis, and discloses a mutual inductor fault diagnosis method and system based on an intelligent algorithm and a storage medium. The method comprises the following steps: acquiring an output signal of a secondary side of a mutual inductor at a preset sampling frequency, carrying out denoising processing on the output signal and converting the output signal into time sequence data; carrying out time domain feature extraction on the time sequence data, obtaining first data features and second data features of a target physical quantity, extracting a zero sequence current component and a zero sequence voltage component from the time sequence data, and calculating a phase difference between the zero sequence current and the zero sequence voltage as a zero sequence feature; judging whether the mutual inductor is abnormal, and if yes, constructing a multi-modal input vector based on the time domain features, the zero sequence feature, an environmental parameter, a temperature parameter and historical fault records, inputting the multi-modal input vector into a pre-trained multi-task deep learning model, and outputting a fault diagnosis result by the deep learning model. The application improves the precision of mutual inductor fault diagnosis.

Description

Technical Field

[0001] This application relates to the field of fault diagnosis technology, and in particular to a fault diagnosis method, system and storage medium for current transformers based on intelligent algorithms. Background Technology

[0002] With the increasing development and complexity of power systems, the safe and stable operation of power equipment is crucial to the reliability of the power system. Instrument transformers, as important measuring devices in power systems, are mainly used for current and voltage measurement and protection, and are widely used in relay protection, metering equipment, and automated control systems. The operating status of instrument transformers directly affects the normal operation of the power system; therefore, the diagnosis and prediction of instrument transformer faults are particularly important.

[0003] Existing patent applications, such as Chinese patent application CN117056725A, propose a fault diagnosis method and apparatus for fiber optic current transformers. The method includes: acquiring light source monitoring data of the fiber optic current transformer; oversampling the light source monitoring data; dividing the oversampled light source monitoring data into training samples, test samples, and verification samples; extracting feature vectors from the training samples, test samples, and verification samples respectively; constructing a training set from the feature vectors of the training samples, a test set from the feature vectors of the test samples, and a verification set from the feature vectors of the verification samples; training a pre-set fault diagnosis model using the training set and test set to obtain the trained fault diagnosis model; verifying the accuracy of the trained fault diagnosis model using the verification set and determining whether the accuracy of the trained fault diagnosis model meets the requirements; if it meets the requirements, using the trained fault diagnosis model to diagnose faults in the fiber optic current transformer. However, the aforementioned prior art fault diagnosis method for fiber optic current transformers relies on oversampling and feature extraction, which may lead to excessively high computational complexity and strong sensitivity to noise data.

[0004] There is also a Chinese patent application with publication number CN117171616A, which proposes a data-augmented voltage transformer fault diagnosis method, device, and storage medium, including: acquiring historical voltage data of the secondary side of the voltage transformer; establishing a transient data model of the voltage transformer to enhance the fault voltage data; constructing a graph convolutional neural network model and inputting the enhanced fault voltage data into the graph convolutional neural network model for training; and using the trained graph convolutional neural network model to diagnose the current voltage data of the voltage transformer. However, although the above-mentioned existing voltage transformer fault diagnosis methods utilize data augmentation, model training may still face problems of low efficiency and high computational requirements under the conditions of large data volume and complex calculations. Summary of the Invention

[0005] To address the aforementioned technical problems, this application provides a method, system, and storage medium for fault diagnosis of instrument transformers based on intelligent algorithms, which can improve the accuracy and efficiency of fault diagnosis of instrument transformers.

[0006] In a first aspect, this application provides a fault diagnosis method for current transformers based on intelligent algorithms, the method comprising:

[0007] Step S1: Obtain the output signal of the secondary side of the current transformer at a preset sampling frequency, perform noise reduction processing on the output signal and convert it into time series data;

[0008] Step S2: Extract time-domain features from the time series data to obtain the first and second data features of the target physical quantity. Simultaneously, extract the zero-sequence current component and the zero-sequence voltage component from the time series data. Use Fast Fourier Transform to calculate the frequency domain energy distribution and phase shift. Calculate the phase difference between the zero-sequence current and the zero-sequence voltage based on the frequency domain energy distribution as the zero-sequence feature.

[0009] Step S3: Determine whether the current transformer has malfunctioned based on the time domain features. If so, obtain the environmental parameters and temperature parameters of the current transformer, and construct a multimodal input vector based on the time domain features, zero-sequence features, environmental parameters, temperature parameters, and historical fault records.

[0010] Step S4: Establish a multi-task deep learning model. Input the multimodal input vector into the pre-trained multi-task deep learning model. The deep learning model outputs fault diagnosis results, including the classification results of the transformer fault type, the estimated value of the fault location, and the probability of the fault occurring.

[0011] In conjunction with the first aspect, in a first implementation of the first aspect of this application, the output signal is subjected to noise reduction processing, including:

[0012] The acquired output signal is initially filtered based on a bandpass filter. The output signal is current or voltage waveform data. The frequency range of the bandpass filter is set based on the operating frequency of the power system. Output signals outside the frequency range are identified, and noise components are determined to exist. The noise components are then removed.

[0013] The output signal after preliminary processing is further denoised based on an adaptive filter to remove external noise, environmental noise and interference noise from within the power system. Wavelet transform is used to convert the denoised signal from the time domain to the frequency domain, extract key frequency components from the signal, and remove noise generated by the external system and environment through frequency domain analysis.

[0014] In conjunction with the first aspect, in the second implementation of the first aspect of this application, obtaining the first data feature and the second data feature of the target physical quantity includes:

[0015] The latest preset number of time series data is extracted from the time series data. The first data feature is the average value and first standard deviation of the target physical quantity in the latest time series data. The second data feature includes the time change rate and second standard deviation of the target physical quantity in the latest time series data. Regression analysis is performed on the latest preset number of time series data. A linear regression equation is established based on the acquisition time point. The predicted value of the target physical quantity at each acquisition time point is calculated. The optimal slope and optimal intercept are calculated based on the error function. The optimal slope is defined as the time change rate of the target physical quantity.

[0016] In conjunction with the first aspect, in the third implementation of the first aspect of this application, regression analysis is performed on the latest preset amount of time series data, including:

[0017] The optimal linear regression equation is determined based on the optimal slope and the optimal intercept. The optimal predicted value of the target physical quantity is determined based on the optimal linear regression equation. The sum of squared residuals between the actual measured value and the optimal predicted value is obtained. The sum of the squared differences between each acquisition time point and the average value of the time series is calculated. The sum of squared residuals is divided by the degree of dispersion to define the second standard deviation of the rate of change over time.

[0018] In conjunction with the first aspect, in the fourth implementation of the first aspect of this application, the frequency domain energy distribution and phase shift are calculated using Fast Fourier Transform, including:

[0019] A fast Fourier transform is performed on the zero-sequence current and zero-sequence voltage to extract the frequency domain signal. The energy of each frequency component is calculated in the frequency domain. Based on the complex number result obtained from the fast Fourier transform, the phase of each frequency component is extracted, and the phase difference between the zero-sequence current and zero-sequence voltage is calculated as the phase offset.

[0020] In conjunction with the first aspect, in the fifth implementation of the first aspect of this application, determining whether the current transformer has malfunctioned based on time-domain characteristics includes:

[0021] If the average value of the first data feature and the second data feature are set, and the time change rate of the target physical quantity in the time series data is greater than the preset reference value, and the first standard deviation and the second standard deviation are less than or equal to the preset reference value, then it is determined that the transformer is abnormal.

[0022] In conjunction with the first aspect, in the sixth implementation of the first aspect of this application, a multi-task deep learning model is established, including:

[0023] The multi-task deep learning model includes at least one shared layer and multiple task-specific layers. The multimodal input vector is input into the pre-trained multi-task deep learning model. The multi-task deep learning model extracts general features through the shared layer, uses historical fault records as training data, and combines time-domain features, zero-order features, environmental parameters, and temperature parameters for training. The model parameters are optimized through backpropagation, and the model performance is tuned through cross-validation. Fault diagnosis is performed through multiple task-specific layers, and the fault diagnosis results are output.

[0024] Secondly, this application provides a fault diagnosis system for instrument transformers based on intelligent algorithms, the system comprising:

[0025] The preprocessing module acquires the output signal of the secondary side of the current transformer at a preset sampling frequency, performs noise reduction processing on the output signal, and converts it into time series data.

[0026] The extraction module performs time-domain feature extraction on the time series data to obtain the first and second data features of the target physical quantity. At the same time, it extracts the zero-sequence current component and the zero-sequence voltage component from the time series data, uses fast Fourier transform to calculate the frequency domain energy distribution and phase shift, and calculates the phase difference between the zero-sequence current and the zero-sequence voltage as the zero-sequence feature based on the frequency domain energy distribution.

[0027] The fusion module determines whether the current transformer is abnormal based on time-domain features. If so, it obtains the environmental and temperature parameters of the current transformer and constructs a multi-modal input vector based on time-domain features, zero-sequence features, environmental parameters, temperature parameters, and historical fault records.

[0028] The diagnostic module establishes a multi-task deep learning model. The multimodal input vector is input into the pre-trained multi-task deep learning model, and the deep learning model outputs fault diagnosis results, including the classification results of the transformer fault type, the estimated value of the fault location, and the probability of the fault occurrence.

[0029] A third aspect of this application provides a computer-readable storage medium storing instructions that, when executed on a computer, cause the computer to perform the aforementioned intelligent algorithm-based transformer fault diagnosis method.

[0030] Compared with the prior art, the beneficial effects of the present invention are at least as follows:

[0031] This invention achieves accurate fault diagnosis of instrument transformers through the application of multimodal data input and a multi-task deep learning model. The method, through steps S1 to S4, combines denoising of the instrument transformer output signal, time-domain feature extraction, zero-sequence feature calculation, and multi-task learning, effectively extracting key features from current and voltage signals and performing real-time analysis. First, various noises, including external noise, environmental noise, and internal interference of the power system, are removed using bandpass and adaptive filters, thereby improving signal quality. Then, wavelet transform is used for frequency domain analysis to accurately extract frequency domain features and further eliminate noise components. Second, regression analysis is used to calculate the time change rate of the target physical quantity, and combined with time-domain and zero-sequence features, the dynamic changes of the signal are accurately captured to determine whether an anomaly exists. Furthermore, a multimodal input vector is constructed by combining environmental parameters, temperature parameters, and historical fault records. A deep learning model is then jointly trained on multiple fault diagnosis tasks to output fault type classification, fault location estimation, and the probability of fault occurrence. This method enables early warning and precise location of transformer faults, significantly improving the accuracy and reliability of fault detection, reducing misdiagnosis and missed diagnosis, optimizing the operation and maintenance management of the power system, and ensuring the safe and stable operation of the system. Attached Figure Description

[0032] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0033] Figure 1 This is a schematic diagram of an embodiment of the current transformer fault diagnosis method based on intelligent algorithms in this application.

[0034] Figure 2 This is a schematic diagram of one embodiment of the instrument transformer fault diagnosis system based on intelligent algorithms in this application. Detailed Implementation

[0035] This application provides a method, system, and storage medium for fault diagnosis of instrument transformers based on intelligent algorithms. The terms "first," "second," "third," "fourth," etc. (if present) in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in a sequence other than that illustrated or described herein. Furthermore, the terms "comprising" or "having" and any variations thereof are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or device that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or devices.

[0036] For ease of understanding, the specific process of the embodiments of this application is described below. Please refer to [link / reference]. Figure 1 One embodiment of the current transformer fault diagnosis method based on intelligent algorithms in this application includes:

[0037] Step S1: Acquire the output signal of the secondary side of the current or voltage transformer at a preset sampling frequency, perform noise reduction processing on the output signal and convert it into time series data.

[0038] In the process of diagnosing instrument transformer faults, power systems face many complex challenges, mainly stemming from factors such as unstable signal quality, external interference, and equipment aging. As core measuring devices in power systems, instrument transformers are responsible for monitoring changes in current or voltage and providing necessary data for protection, control, and monitoring. However, due to their complex operating environment and prolonged exposure to high voltage, harsh weather conditions, and electromagnetic interference, the output signal of instrument transformers is easily affected by noise, leading to inaccurate or distorted measurement data. Failure to identify and address these problems in a timely manner can result in serious consequences, such as misdiagnosis or missed diagnosis of power system faults, affecting the safety and stability of the entire power system.

[0039] Therefore, one of the key issues in the fault diagnosis of instrument transformers is how to extract effective information from noise-affected current or voltage signals to ensure accurate and timely fault detection. Because the signals themselves contain a large amount of environmental noise and interference, traditional diagnostic methods are prone to failure in complex environments, failing to provide accurate fault diagnosis results.

[0040] Specifically, the output signal is acquired from the secondary side of the current or voltage transformer at a preset sampling frequency. For example, assuming the voltage signal monitored by the transformer operates at a frequency of 50Hz (i.e., the power system's operating frequency is 50Hz), in order to capture the signal details within this frequency range, the sampling frequency needs to be at least 100Hz. This means 100 data points are collected per second, and the output signal is the waveform data of the secondary side current or voltage. The acquired signal is often affected by external noise and inherent equipment noise. Therefore, this invention uses bandpass filters, adaptive filters, and wavelet transforms to denoise the signal. The specific denoising process will be explained later. The denoised signal is converted into time-series data. The time-series data format can accurately capture the trend and fluctuations of the signal over time. By acquiring the signal at a preset sampling frequency and performing denoising, the accuracy and reliability of the signal data are ensured, errors caused by noise are eliminated, thereby improving the accuracy of transformer fault diagnosis.

[0041] Step S2: Extract time-domain features from the time series data to obtain the first and second data features of the target physical quantity. Simultaneously, extract the zero-sequence current component and the zero-sequence voltage component from the time series data. Use Fast Fourier Transform to calculate the frequency domain energy distribution and phase shift. Calculate the phase difference between the zero-sequence current and the zero-sequence voltage based on the frequency domain energy distribution as the zero-sequence feature.

[0042] Specifically, the latest predetermined amount of time series data is determined from the time series data, and time-domain features are extracted from it. The target physical quantity includes secondary side current or secondary side voltage. The first data feature includes the average value and first standard deviation of the target physical quantity. The second data feature includes the time change rate of the target feature and the second standard deviation of the time change rate.

[0043] However, in power systems, faults such as grounding faults, short-circuit faults, or load imbalances often lead to imbalances in current and voltage signals. Faults in instrument transformers themselves are usually caused by equipment aging, damage, or other reasons that distort their measurement signals. Therefore, in the case of instrument transformer faults and system faults, both current and voltage output waveforms may be abnormal. However, the sources and natures of these two abnormalities are different. Therefore, relying solely on the time-domain characteristics in time series data may not be sufficient to accurately distinguish between system faults and instrument transformer faults. Thus, it is necessary to extract the zero-sequence current component and the zero-sequence voltage component from the time series data.

[0044] In a three-phase power system, the zero-sequence component reflects the imbalance in the system's current and voltage. Under normal circumstances, the three-phase current and voltage are balanced, so the zero-sequence current and voltage should be very close to zero. A significant increase in zero-sequence current and voltage usually indicates an imbalance in the system. This application uses Fast Fourier Transform (FFT) to convert the time-domain signal into frequency-domain data, extracting the energy distribution and phase shift of each frequency component. The specific extraction method will be explained later. Relying solely on time-domain characteristics may lead to misjudgments of signal fluctuations because time-domain characteristics cannot capture changes in the frequency components of the signal, which are often obvious characteristics of fault occurrence. Introducing the analysis of zero-sequence current and voltage can effectively distinguish between anomalies caused by transformer faults and power system faults, thereby reducing the probability of misdiagnosis and missed diagnosis.

[0045] Step S3: Determine whether the current transformer is abnormal based on the time domain features. If so, obtain the environmental parameters and temperature parameters of the current transformer, and construct a multi-modal input vector based on the time domain features, zero-sequence features, environmental parameters, temperature parameters and historical fault records.

[0046] Specifically, when a current transformer malfunctions or experiences an anomaly, its time-domain characteristics change significantly. Therefore, analyzing the time-domain characteristics is necessary to determine whether the current transformer is malfunctioning. However, a single time-domain characteristic may not be sufficient to accurately determine the type and source of the fault. Therefore, it is necessary to construct a multi-mode input vector, which includes not only the time-domain characteristics of the signal but also the zero-sequence characteristics of the power system, environmental parameters, temperature parameters, and historical fault records.

[0047] The introduction of zero-sequence features helps distinguish between signal anomalies caused by transformer faults and those caused by system faults, reducing the possibility of misjudgment. Environmental parameters include factors in the system's operating environment, such as temperature, humidity, air pressure, and electromagnetic interference. Environmental factors have a significant impact on transformer performance. For example, high temperatures may cause aging of the transformer's coil insulation material, or electromagnetic interference may lead to errors in the measurement signal. Extreme environmental changes (such as excessively high or low temperatures) may cause the equipment to malfunction; rising temperatures may lead to performance degradation or failure of the transformer. Therefore, monitoring temperature changes helps to identify the health status of the equipment in a timely manner and prevent equipment failure due to overheating in high-temperature environments. Historical fault records contain the types of faults that have occurred in the transformer in the past, the time of occurrence, the frequency of occurrence, and maintenance records. Historical fault records can help the system identify long-term fault trends of the transformer. In complex environments, a single feature may be insufficient to cope with various types of faults. By fusing multiple features, the robustness of the model can be enhanced, making it more adaptable to various operating environments and fault types, especially when facing complex or uncertain factors.

[0048] Step S4: Establish a multi-task deep learning model. Input the multimodal input vector into the pre-trained multi-task deep learning model. The deep learning model outputs fault diagnosis results, including the classification results of the transformer fault type, the estimated value of the fault location, and the probability of the fault occurring.

[0049] Specifically, the multimodal input vectors are first fed into a pre-trained multi-task deep learning model for comprehensive processing. The multi-task deep learning model extracts common features from these data through a shared layer, while simultaneously learning for its respective task through multiple dedicated layers, outputting diagnostic results including fault type classification, fault location estimation, and fault occurrence probability. By fusing multiple features, the deep learning model can more comprehensively understand the operating state of the transformer, thereby improving the accuracy of transformer fault diagnosis. Data fusion of different features effectively avoids misjudgments or missed diagnoses caused by a single feature.

[0050] In one specific embodiment, the noise reduction process for the output signal includes the following steps:

[0051] The acquired output signal is initially filtered based on a bandpass filter. The output signal is current or voltage waveform data. The frequency range of the bandpass filter is set based on the operating frequency of the power system. Output signals outside the frequency range are identified, and noise components are determined to be present. The noise components are then removed.

[0052] The output signal after preliminary processing is further denoised using an adaptive filter to remove external noise, environmental noise, and interference noise from within the power system. Wavelet transform is used to convert the denoised signal from the time domain to the frequency domain, extracting key frequency components from the signal. Frequency domain analysis is then used to remove noise generated by the external system and environment.

[0053] Specifically, in practical work, there are various sources of noise in power systems, such as electromagnetic interference, environmental noise, and sensor noise. These noises can interfere with the signals measured by the current transformer, causing signal distortion or inaccuracy, thereby affecting the reliability of fault diagnosis and system monitoring. Therefore, it is necessary to perform noise reduction processing on the output signal.

[0054] This application first uses a bandpass filter to perform preliminary filtering on the acquired current or voltage signals. Power systems typically operate within a certain frequency range, and the frequency of signals is usually concentrated near the operating frequency (e.g., 50Hz or 60Hz). Any signal exceeding this frequency range may be noise or interference. Therefore, the function of the bandpass filter is to remove high-frequency noise and low-frequency interference by setting an appropriate frequency range (e.g., 50Hz ± 5Hz), while retaining the operating frequency components of the power system.

[0055] Secondly, this application employs an adaptive filter to further denoise the initially filtered signal. Adaptive filters, by continuously adjusting their filtering characteristics, can dynamically adapt to different types of noise, exhibiting strong denoising capabilities, especially in environments with constantly changing noise characteristics. Unlike traditional fixed filters, adaptive filters can self-adjust according to changes in the input signal, removing external noise, environmental noise, and interference noise from within the power system.

[0056] Finally, wavelet transform is used to convert the denoised signal from the time domain to the frequency domain, and frequency domain analysis is used to further remove noise generated by the system's external environment. Wavelet transform has good time-frequency localization characteristics, enabling signal decomposition at different scales to extract key frequency components. This series of denoising processes solves the problem of unstable signal quality in the current transformer. Multi-stage filtering effectively removes different types of noise, making the signal cleaner and achieving significant denoising results.

[0057] In one specific embodiment, obtaining the first data feature and the second data feature of the target physical quantity specifically includes the following steps:

[0058] The latest preset amount of time series data is extracted from the time series data. The first data feature is the average value and the first standard deviation of the target physical quantity in the latest time series data. The second data feature includes the time change rate and the second standard deviation of the target physical quantity in the latest time series data. Regression analysis is performed on the latest preset amount of time series data. A linear regression equation is established based on the collection time point. The predicted value of the target physical quantity at each collection time point is calculated. The optimal slope and the optimal intercept are calculated based on the error function. The optimal slope is defined as the time change rate of the target physical quantity.

[0059] Specifically, the first step is to collect data on the target physical quantity, typically current or voltage signals, from the secondary side of the transformer. For example, if the target physical quantity is the secondary current signal, then the data extracted from the time series is the data at different time points. Secondary current data acquired at the site, for example: Similarly, if the target physical quantity is a voltage signal, then the collected data is the voltage value. This application takes the secondary voltage as an example, and the average value is calculated using the following formula: The formula for calculating the first standard deviation of the mean is: Suppose there are 5 current signals in the time series data: 3.5, 3.8, 3.6, 3.7, and 3.9. =3.7, =0.2.

[0060] Traditional fault diagnosis methods typically rely on simple statistical analysis or feature extraction (such as mean and standard deviation) of signals. However, these methods often overlook the dynamic changes of signals over time, especially when signals change drastically, potentially failing to capture subtle fault signals. Calculating the rate of change over time using a linear regression equation is a method that can accurately capture signal trends, particularly suitable for analyzing continuous time series data. Linear regression analysis is performed on the latest n time series data points to establish the relationship between the target physical quantity and the acquisition time points. The linear regression equation is: y = ax + b, where y represents the predicted value of the target physical quantity at time point x, a is the slope of the regression equation, and b is the intercept. The slope of the regression equation represents the rate of change of the target physical quantity over time. Compared to more complex nonlinear models, linear regression models are relatively simple and less prone to overfitting, making them particularly suitable for real-time data analysis. The error function is set as follows: Collect N latest time series data points ,in It is the i-th time point. It is the actual measured value of the target physical quantity at that point in time. By substituting the error function and taking the partial derivatives with respect to a and b, respectively, and setting them to zero, we find the point that minimizes the error function, thus obtaining the optimal slope and intercept. In the linear regression model, the optimal slope represents the rate of change of the target physical quantity over time, and the optimal intercept provides a prediction of the initial state of the equipment. The optimal slope and optimal intercept are obtained by minimizing the error function, ensuring that the regression model better fits the actual data and reduces prediction errors. For dynamic changes in the target physical quantity, the optimized regression model can more accurately reflect the trend of signal changes, helping to improve the accuracy of fault diagnosis.

[0061] In one specific embodiment, performing regression analysis on the latest preset amount of time series data specifically includes the following steps:

[0062] The optimal linear regression equation is determined based on the optimal slope and optimal intercept. The optimal predicted value of the target physical quantity is determined based on the optimal linear regression equation. The sum of squared residuals between the actual measured value and the optimal predicted value is obtained. The sum of the squared differences between each acquisition time point and the average value of the time series is calculated. The sum of squared residuals is divided by the degree of dispersion to define the second standard deviation of the rate of change over time.

[0063] Specifically, the optimal linear regression equation is determined based on the optimal slope and optimal intercept. The predicted value of the target physical quantity is calculated for each acquisition time point. The sum of squared residuals between the actual measured value and the optimal predicted value is obtained. To calculate the degree of dispersion, the difference between the target physical quantity at each acquisition time point and the average value of the time series is calculated. The sum of the squared differences between each acquisition time point and the average value of the time series is obtained. The sum of squared residuals is divided by the degree of dispersion to obtain the second standard deviation of the rate of change over time. The formula for calculating the second standard deviation is as follows: The second standard deviation of the rate of change over time is used to measure the reliability of the slope estimate. When the sum of squared residuals is large or the second standard deviation increases abnormally, it may indicate a fault in the transformer (such as a ground fault, overload, etc.). This method, through regression analysis combined with the actual trend of signal changes, can provide more accurate fault detection, help identify equipment anomalies, avoid the impact of equipment failures on the power system, and improve the accuracy and response speed of fault diagnosis.

[0064] In one specific embodiment, calculating the frequency domain energy distribution and phase shift using Fast Fourier Transform includes the following steps:

[0065] Fast Fourier Transform is performed on the zero-sequence current and zero-sequence voltage to extract the frequency domain signal. The energy of each frequency component is calculated in the frequency domain. Based on the complex number result obtained from the Fast Fourier Transform, the phase of each frequency component is extracted, and the phase difference between the zero-sequence current and zero-sequence voltage is calculated as the phase offset.

[0066] Specifically, the acquired time-domain signal is first converted into a frequency-domain signal using a Fast Fourier Transform (FFT). The FFT decomposes the time-domain signal into different frequency components, revealing the signal's spectral characteristics. Through frequency domain analysis, the energy of each frequency component can be calculated. The magnitude of the energy reflects the importance of each frequency component in the signal, helping to identify the main components and noise in the signal.

[0067] Based on the complex results obtained from FFT, phase information for each frequency component is further extracted. This phase information is used to reveal the phase difference between current and voltage signals. In fault diagnosis, the phase difference between zero-sequence current and zero-sequence voltage is a key feature, reflecting unbalanced loads or fault conditions in the power system. By calculating the phase difference between zero-sequence current and zero-sequence voltage, the phase offset is obtained, thus providing an important basis for fault diagnosis. This technology can effectively identify and locate fault types, especially phase imbalances between current and voltage, ensuring the accuracy and timeliness of diagnostic results and solving the technical challenge of traditional diagnostic methods in accurately identifying complex faults.

[0068] In one specific embodiment, determining whether a current transformer has malfunctioned based on time-domain characteristics includes the following steps:

[0069] If the average value of the target physical quantity in the time series data and the rate of change over time are greater than the preset reference value, and the first standard deviation and the second standard deviation are less than or equal to the preset reference value, then the transformer is determined to be abnormal.

[0070] Specifically, in instrument transformer fault diagnosis, the process of determining whether an instrument transformer is malfunctioning based on time-domain characteristics first requires setting reference values. These reference values ​​are typically based on historical data or statistical data from normal operating conditions. First, the time-series data is analyzed to calculate the average value and rate of change over time of the target physical quantity. If these values ​​exceed the preset reference values, it indicates that the change in the target physical quantity has exceeded the normal fluctuation range, potentially indicating an equipment malfunction. Furthermore, the standard deviation and second standard deviation need to be calculated; these indicators reflect the data volatility. If these two deviations are less than or equal to the preset reference values, it indicates that the signal volatility meets the normal operating conditions of the equipment, and there is no abnormal fluctuation. For example, assuming the average current signal of an instrument transformer is 3.8A and the rate of change over time is 0.02A / s, these values ​​are greater than the preset reference values, and the standard deviation is 0.1A, while the second standard deviation is 0.05A, both less than or equal to the preset reference values. Therefore, the instrument transformer is considered to be currently operating normally. However, if the average value and rate of change of the current exceed the reference values, and the deviation is large, it indicates an equipment malfunction, such as a grounding fault or overload.

[0071] This method effectively determines whether a current transformer is faulty, ensuring the stable operation of the power system, preventing downtime or accidents caused by equipment failure, and improving the accuracy and timeliness of fault diagnosis. By setting appropriate reference values, this technique can accurately identify potential faults, solving the problem of traditional diagnostic methods' inadequacy in handling complex signals and unstable data.

[0072] In one specific embodiment, establishing a multi-task deep learning model includes the following steps:

[0073] The multi-task deep learning model includes at least one shared layer and multiple task-specific layers. Multimodal input vectors are input into the pre-trained multi-task deep learning model. The multi-task deep learning model extracts general features through the shared layer, uses historical fault records as training data, and combines time-domain features, zero-order features, environmental parameters, and temperature parameters for training. The model parameters are optimized through backpropagation, and the model performance is tuned through cross-validation. Fault diagnosis is performed through multiple task-specific layers, and the fault diagnosis results are output.

[0074] Specifically, in the fault diagnosis of instrument transformers, the implementation process of establishing a multi-task deep learning model includes several key steps. First, the multi-task deep learning model comprises at least one shared layer and multiple task-specific layers. The shared layer extracts common features from all input data, including current, voltage, zero-sequence components, and environmental data. The multiple task-specific layers handle different fault diagnosis tasks, such as fault type classification (instrument transformer coil short circuit, aging, insulation damage, etc.), fault location estimation, and fault occurrence probability prediction. In the pre-training phase of the multi-task deep learning model, the model is first trained using historical data. Training data includes historical fault records, time-domain features, zero-sequence features, environmental parameters, and temperature parameters. Through pre-training, the model can learn the basic feature representations for each task from the data. During training, the model uses historical fault records as label data. These records include information such as the time, type, duration, and whether the instrument transformer fault occurred. These records help the model identify the equipment's fault modes and perform appropriate training. During model training, the backpropagation algorithm is used to optimize the model's parameters. The backpropagation algorithm calculates the error for each task and updates the weights of the shared layer and task-specific layers based on the error. This process is iteratively optimized until the model achieves satisfactory performance on each task.

[0075] During training, the multimodal input data is first fed into a pre-trained multi-task deep learning model, which extracts common features from the data through shared layers. Then, combining historical fault records and other environmental data, the model performs backpropagation, continuously adjusting parameters to optimize the network. Cross-validation is used to evaluate the model's performance and fine-tune its parameters to ensure accuracy on different datasets. Through multiple task-specific layers, the model can handle different fault diagnosis tasks separately, ultimately outputting accurate fault diagnosis results, such as identifying the fault type of the instrument transformer, estimating the fault location, and predicting the probability of fault occurrence. By combining multimodal input data, the multi-task deep learning model can perform joint learning on multiple tasks, classifying fault types and estimating fault locations and probabilities of occurrence. This multi-task learning approach helps improve the system's ability to identify and predict faults and comprehensively addresses the problem of diagnosing instrument transformer faults.

[0076] The above describes a current transformer fault diagnosis method based on intelligent algorithms in the embodiments of this application. The following describes a current transformer fault diagnosis system based on intelligent algorithms in the embodiments of this application. Please refer to [link / reference]. Figure 2 One embodiment of a current transformer fault diagnosis system based on intelligent algorithms in this application includes:

[0077] The preprocessing module acquires the output signal from the secondary side of the current transformer at a preset sampling frequency, performs noise reduction processing on the output signal, and converts it into time series data.

[0078] The extraction module performs time-domain feature extraction on the time series data to obtain the first and second data features of the target physical quantity. At the same time, it extracts the zero-sequence current component and the zero-sequence voltage component from the time series data, uses fast Fourier transform to calculate the frequency domain energy distribution and phase shift, and calculates the phase difference between the zero-sequence current and the zero-sequence voltage based on the frequency domain energy distribution as the zero-sequence feature.

[0079] The fusion module determines whether the current transformer is abnormal based on time-domain features. If so, it obtains the environmental and temperature parameters of the current transformer and constructs a multi-modal input vector based on time-domain features, zero-sequence features, environmental parameters, temperature parameters, and historical fault records.

[0080] The diagnostic module establishes a multi-task deep learning model. The multimodal input vector is input into the pre-trained multi-task deep learning model, and the deep learning model outputs fault diagnosis results, including the classification results of the transformer fault type, the estimated value of the fault location, and the probability of the fault occurrence.

[0081] This application also provides a computer-readable storage medium, which can be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium. The computer-readable storage medium stores instructions that, when executed on a computer, cause the computer to perform the steps of a current transformer fault diagnosis method based on an intelligent algorithm.

[0082] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0083] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0084] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A fault diagnosis method for instrument transformers based on intelligent algorithms, characterized in that, The method includes: Step S1: Obtain the output signal of the secondary side of the current transformer at a preset sampling frequency, perform noise reduction processing on the output signal and convert it into time series data; Step S2: Extract time-domain features from the time series data to obtain the first and second data features of the target physical quantity. Simultaneously, extract the zero-sequence current component and the zero-sequence voltage component from the time series data. Use Fast Fourier Transform to calculate the frequency domain energy distribution and phase shift. Calculate the phase difference between the zero-sequence current and the zero-sequence voltage based on the frequency domain energy distribution as the zero-sequence feature. The process of obtaining the first and second data features of the target physical quantity includes: extracting the latest preset amount of time series data from the time series data; the first data feature is the average value and first standard deviation of the target physical quantity in the latest time series data; the second data feature includes the time change rate and second standard deviation of the target physical quantity in the latest time series data; performing regression analysis on the latest preset amount of time series data; establishing a linear regression equation based on the acquisition time points; calculating the predicted value of the target physical quantity at each acquisition time point; calculating the optimal slope and optimal intercept based on the error function; defining the optimal slope as the time change rate of the target physical quantity; and calculating the frequency domain energy distribution and phase shift using Fast Fourier Transform (FFT), including: performing FFT on the zero-sequence current and zero-sequence voltage to extract the frequency domain signal; calculating the energy of each frequency component in the frequency domain; extracting the phase of each frequency component based on the complex result obtained from the FFT calculation; and calculating the phase difference between the zero-sequence current and zero-sequence voltage as the phase shift. Step S3: Determine whether the current transformer has malfunctioned based on the time domain features. If so, obtain the environmental parameters and temperature parameters of the current transformer, and construct a multimodal input vector based on the time domain features, zero-sequence features, environmental parameters, temperature parameters, and historical fault records. Step S4: Establish a multi-task deep learning model. Input the multimodal input vector into the pre-trained multi-task deep learning model. The deep learning model outputs fault diagnosis results, including the classification results of the transformer fault type, the estimated value of the fault location, and the probability of the fault occurring. The process of establishing a multi-task deep learning model includes: the multi-task deep learning model comprising at least one shared layer and multiple task-specific layers; inputting the multimodal input vector into the pre-trained multi-task deep learning model; the multi-task deep learning model extracting general features through the shared layer; using historical fault records as training data; combining time-domain features, zero-order features, environmental parameters, and temperature parameters for training; optimizing model parameters through backpropagation; fine-tuning the model's performance through cross-validation; performing fault diagnosis through multiple task-specific layers; and outputting fault diagnosis results.

2. The method according to claim 1, characterized in that, The output signal is denoised, including: The acquired output signal is initially filtered based on a bandpass filter. The output signal is current and voltage waveform data. The frequency range of the bandpass filter is set based on the operating frequency of the power system. Output signals outside the frequency range are identified, and noise components are determined to be present. The noise components are then removed. The output signal after preliminary processing is further denoised based on an adaptive filter to remove external noise, environmental noise and interference noise from within the power system. Wavelet transform is used to convert the denoised signal from the time domain to the frequency domain, extract key frequency components from the signal, and remove noise generated by the external system and environment through frequency domain analysis.

3. The method according to claim 1, characterized in that, Regression analysis was performed on the latest preset amount of time series data, including: The optimal linear regression equation is determined based on the optimal slope and the optimal intercept. The optimal predicted value of the target physical quantity is determined based on the optimal linear regression equation. The sum of squared residuals between the actual measured value and the optimal predicted value is obtained. The sum of the squared differences between each acquisition time point and the average value of the time series is calculated. The sum of squared residuals is divided by the degree of dispersion to define the second standard deviation of the rate of change over time.

4. The method according to claim 1, characterized in that, Determining whether a current transformer is malfunctioning based on time-domain characteristics includes: If the average value of the first data feature and the second data feature are set, and the time change rate of the target physical quantity in the time series data is greater than the preset reference value, and the first standard deviation and the second standard deviation are less than or equal to the preset reference value, then it is determined that the transformer is abnormal.

5. A fault diagnosis system for instrument transformers based on intelligent algorithms, used to implement the fault diagnosis method for instrument transformers based on intelligent algorithms as described in any one of claims 1-4, characterized in that, The system includes: The preprocessing module acquires the output signal of the secondary side of the current transformer at a preset sampling frequency, performs noise reduction processing on the output signal, and converts it into time series data. The extraction module performs time-domain feature extraction on the time-series data to obtain first and second data features of the target physical quantity. Simultaneously, it extracts the zero-sequence current and zero-sequence voltage components from the time-series data, calculates the frequency-domain energy distribution and phase shift using a Fast Fourier Transform, and calculates the phase difference between the zero-sequence current and zero-sequence voltage as the zero-sequence feature based on the frequency-domain energy distribution. Obtaining the first and second data features of the target physical quantity includes: extracting a preset number of the latest time-series data from the time-series data. The first data feature is the average value and first standard deviation of the target physical quantity in the latest time-series data. The second data feature includes the target physical quantity's value in the latest time-series data. The time change rate and second standard deviation in the time series data are used to perform regression analysis on the latest preset amount of time series data. A linear regression equation is established based on the acquisition time point to calculate the predicted value of the target physical quantity at each acquisition time point. The optimal slope and optimal intercept are calculated based on the error function, and the optimal slope is defined as the time change rate of the target physical quantity. The frequency domain energy distribution and phase offset are calculated using fast Fourier transform, including: performing fast Fourier transform on the zero-sequence current and zero-sequence voltage to extract the frequency domain signal, calculating the energy of each frequency component in the frequency domain, extracting the phase of each frequency component based on the complex number result obtained from the fast Fourier transform, and calculating the phase difference between the zero-sequence current and zero-sequence voltage as the phase offset. The fusion module determines whether the current transformer is abnormal based on time-domain features. If so, it obtains the environmental and temperature parameters of the current transformer and constructs a multi-modal input vector based on time-domain features, zero-sequence features, environmental parameters, temperature parameters, and historical fault records. The diagnostic module establishes a multi-task deep learning model. The multimodal input vector is input into the pre-trained multi-task deep learning model, which outputs fault diagnosis results, including classification results of transformer fault types, estimated fault locations, and the probability of fault occurrence. Establishing the multi-task deep learning model includes: the multi-task deep learning model comprising at least one shared layer and multiple task-specific layers; inputting the multimodal input vector into the pre-trained multi-task deep learning model; the multi-task deep learning model extracts general features through the shared layer; uses historical fault records as training data; combines time-domain features, zero-order features, environmental parameters, and temperature parameters for training; optimizes model parameters through backpropagation; fine-tunes model performance through cross-validation; and performs fault diagnosis through multiple task-specific layers, outputting fault diagnosis results.

6. A computer-readable storage medium storing instructions thereon, characterized in that, When the instruction is executed by the processor, it implements the transformer fault diagnosis method based on intelligent algorithms as described in any one of claims 1-4.

Images