Current transformer saturation detection method and system

Abstract

The invention provides a current transformer saturation detection method and system, and relates to the technical field of state detection and fault diagnosis of current transformers, and the method comprises the steps: synchronously collecting the secondary side current, bus voltage, cable insulation and partial discharge data of a current transformer, and carrying out the normalization according to classes, and then completing the time sequence alignment; secondly, a fixed sliding window is set, current distortion rate, voltage over-limit frequency, insulation parameter mean value and partial discharge pulse quantity characteristics are extracted, and whether current transformer saturation, insulation abnormity and partial discharge abnormity events exist in the window or not is judged according to the current distortion rate, the voltage over-limit frequency, the insulation parameter mean value and the partial discharge pulse quantity characteristics; counting the saturation probability under each node state based on a Bayesian causal network, calculating and normalizing the causal contribution degree of the insulation abnormity and the partial discharge abnormity to the saturation event, and determining a dominant influence factor; and finally, a health index is calculated in combination with the normalized contribution degree and the abnormal state, the running health condition of the current transformer is evaluated through trend analysis in combination with a dominant factor, saturation inducements are positioned, and misjudgment is avoided.

Description

Technical Field

[0001] This invention relates to the technical field of current transformer condition detection and fault diagnosis, specifically to a current transformer saturation detection method and system. Background Technology

[0002] With the advancement of smart grid construction and the continuous expansion of power grid scale, the number of current transformers installed has surged and they are widely distributed in substations, distribution rooms and outdoor lines. Accurate diagnosis and cause location of saturation faults have become one of the core needs in the field of power operation and maintenance.

[0003] Currently, the industry has developed several mature solutions for detecting saturation in current transformers. One approach, based on waveform feature analysis, involves collecting secondary current data from the current transformer and using algorithms such as Fourier transform and wavelet analysis to extract waveform distortion features, thereby identifying the saturation state. Another approach is based on equipment characteristic testing. This method utilizes specialized equipment such as volt-ampere characteristic testers and transformer ratio testers to offline test parameters such as the core permeability and excitation current of the current transformer, determining whether there are defects in the current transformer itself. This provides a reliable basis for troubleshooting equipment defects. However, existing technologies still face limitations in practical application. There are still significant shortcomings in the current transformer: On the one hand, existing methods focus on whether there are problems with the current transformer itself, but ignore the coupling relationship between current transformer saturation and cable. Problems such as cable insulation deterioration and abnormal partial discharge can also lead to an increase in harmonic components of the primary current and aggravated voltage fluctuations, which in turn induces current transformer saturation. Maintenance personnel cannot distinguish whether the saturation is caused by defects in the current transformer itself, such as core aging or air gap in the magnetic circuit, or by abnormal cable correlation. They often misjudge and attribute cable-induced saturation to the current transformer itself, resulting in repeated replacement of current transformers without solving the problem.

[0004] The information disclosed in the background section is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0005] The purpose of this invention is to provide a method and system for detecting saturation of current transformers, so as to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: A method for detecting saturation of a current transformer, comprising the following steps: Step 1: Synchronously collect the current data of the secondary side of the target current transformer and the voltage data of the bus connected to it, as well as the online insulation monitoring data and partial discharge detection data of the cable connected to the output end of the current transformer. Normalize the collected data according to the categories, and perform time-series alignment of all normalized data based on a unified timestamp. Step 2: Set a fixed-length sliding window for all detection data. For each sliding window, extract features from the normalized current, voltage, insulation online monitoring and partial discharge detection data. The features include current distortion rate, voltage over-limit frequency, average insulation parameter and number of partial discharge pulses. Based on the feature extraction results, determine whether there are current transformer saturation events, insulation abnormal events and partial discharge abnormal events in the window. Step 3: Within each sliding window, based on the constructed Bayesian causal network model, the current transformer saturation state, insulation abnormality, and partial discharge abnormality are taken as causal network nodes. The probability of the current transformer saturation event occurring under each node state is statistically analyzed. The causal contribution of each influencing factor to the saturation event is calculated and all causal contributions are normalized. The dominant influencing factor within the window is determined based on the normalization result. The influencing factors include insulation abnormality and partial discharge abnormality. Step 4: Calculate the health index for each window based on the normalized causal contribution and abnormal characteristic state of the sliding window, perform trend analysis on the health index and combine it with the statistical results of the dominant influencing factors to assess the operational health status of the current transformer.

[0007] Furthermore, in the online insulation monitoring data and partial discharge detection data of the cable connected to the output terminal of the current transformer, the online insulation monitoring data is the dielectric loss factor, and the partial discharge detection data is the partial discharge charge.

[0008] Furthermore, the method for normalizing the collected data according to different classes is as follows: Using the rated current on the secondary side of the current transformer as a benchmark, the minimum value is set at 5% of the rated current, and the maximum value is set at 20 times the rated current. Min-max normalization is used to process the current data. Using the rated bus voltage as a benchmark, the minimum value is set at 80% of the rated bus voltage, and the maximum value is set at 120% of the rated bus voltage. Min-max normalization is used to process the voltage data. When the real-time monitored voltage is lower than 80% of the rated voltage, it is normalized to 0; when it is higher than 120% of the rated voltage, it is normalized to 1. Insulation online monitoring data is normalized to... Based on the allowable dielectric loss factor, if the dielectric loss factor is less than 10% of the rated allowable value, it is calculated at 10% of the rated allowable value; if the dielectric loss factor is greater than the rated allowable dielectric loss factor, it is calculated at the rated allowable dielectric loss factor. The min-max normalization method is used to complete the online insulation monitoring data processing. The maximum and minimum values ​​of the partial discharge charge amplitude threshold are preset, and the min-max normalization method is used to complete the partial discharge detection data processing. When the partial discharge charge is less than the minimum value, it is normalized to 0; when the partial discharge charge is greater than the maximum value, it is normalized to 1.

[0009] Furthermore, the method for time-series alignment of all normalized data based on a unified timestamp is as follows: Using the sampling frequencies of current and voltage data as reference frequencies, the interval between adjacent time markers is calculated to generate a continuous unified timestamp sequence. Each time marker in this sequence is defined as a target time point, i.e., a unified time reference point for all data to be aligned. The time point for partial discharge detection is defined as the partial discharge sampling time point. Current and voltage data are directly matched to the target time points. For dielectric loss factor data and partial discharge charge data, time alignment is achieved by covering all target time points in the unified timestamp sequence. Specifically, for the dielectric loss factor, a linear interpolation method is used to calculate the completion value, i.e., based on two adjacent dielectric loss factors... The target time point is calculated by combining the time interval ratio between the target time point and the two dielectric loss factors, and then obtaining the dielectric loss factor filler value through linear calculation. For the partial discharge charge, the time difference between each target time point and all original partial discharge sampling time points is calculated. If the time difference between a certain original partial discharge sampling time point and the target time point does not exceed half of the reference frequency time interval, then the partial discharge charge corresponding to that original partial discharge sampling time point is assigned to the target time point. If the time difference between all original partial discharge sampling time points and the target time point exceeds half of the reference frequency time interval, then 0 is filled in at the target time point.

[0010] Furthermore, for each sliding window, the method for feature extraction of normalized current, voltage, insulation online monitoring, and partial discharge detection data is as follows: Extract the normalized current data for all time steps within the sliding window, perform a Fourier transform on the continuous current data segment, and separate the effective value of the fundamental current and the effective values ​​of each harmonic current. The ratio of the sum of the root mean square values ​​of the effective values ​​of each harmonic current to the effective value of the fundamental current is used as the current distortion rate of the sliding window. Extract the normalized voltage data for all time steps within the sliding window, and count the number of data points with a normalized voltage value of 0 or 1 within the window. This number is the frequency of voltage over-limit in the sliding window. Extract the normalized dielectric loss factor data for all time steps within the sliding window, perform an arithmetic mean operation on the normalized dielectric loss factor data within the window, and use the result as the mean value of the insulation parameters for the sliding window. Extract the normalized partial discharge charge data for all time steps within the sliding window, and count the number of data points with non-zero normalized partial discharge charge within the window. This number represents the number of partial discharge pulses in the sliding window.

[0011] Furthermore, the method for determining whether current transformer saturation events, insulation abnormalities, and partial discharge abnormalities exist within the window based on the feature extraction results is as follows: The system has preset current distortion threshold, dielectric loss factor threshold, and partial discharge pulse threshold. When the current distortion rate of the sliding window exceeds the preset current distortion threshold, a current transformer saturation event is determined to exist within the window. When the average insulation parameter of the sliding window exceeds the preset dielectric loss factor threshold and the voltage over-limit frequency is 0, an insulation abnormality event is determined to exist within the window. When the number of partial discharge pulses in the sliding window exceeds the preset partial discharge pulse threshold and the voltage over-limit frequency is 0, a partial discharge abnormality event is determined to exist within the window.

[0012] Furthermore, the method for calculating the probability of current transformer saturation events occurring under each node state is as follows: After statistically analyzing the insulation abnormality state, partial discharge abnormality state, and saturation event state of all sliding windows, based on the preset insulation abnormality node and partial discharge abnormality node, all sliding window data are divided into four non-overlapping groups: windows with neither insulation abnormality nor partial discharge abnormality, windows with partial discharge abnormality but no insulation abnormality, windows with insulation abnormality but no partial discharge abnormality, and windows with both insulation abnormality and partial discharge abnormality. The total number of windows in each group is counted and recorded as the total number of windows. The number of sliding windows judged to have a saturation event is counted and recorded as the number of saturated windows. The ratio of the number of saturated windows in each group to the total number of windows in that group is the probability of the current transformer in that group experiencing a saturation event, recorded as the saturation probability.

[0013] Furthermore, the method for calculating the causal contribution of each influencing factor to the saturation event and normalizing all causal contributions is as follows: The influencing factors are defined as insulation anomaly and partial discharge anomaly. When there is no partial discharge anomaly, the difference in saturation probability between the presence and absence of insulation anomaly is calculated. When there is a partial discharge anomaly, the difference in saturation probability between the presence and absence of insulation anomaly is calculated. The arithmetic mean of these two differences is taken as the causal contribution of insulation anomaly. Similarly, when there is no insulation anomaly, the difference in saturation probability between the presence and absence of partial discharge anomaly is calculated. When there is an insulation anomaly, the difference in saturation probability between the presence and absence of partial discharge anomaly is calculated. The arithmetic mean of these two differences is taken as the causal contribution of partial discharge anomaly. The sum of the absolute values ​​of the causal contribution of insulation anomaly and the absolute values ​​of the causal contribution of partial discharge anomaly is taken as the total contribution. The ratio of the absolute value of the insulation anomaly contribution to the total contribution is taken as the normalized causal contribution of insulation anomaly. The ratio of the absolute value of the partial discharge anomaly contribution to the total contribution is taken as the normalized causal contribution of partial discharge anomaly.

[0014] Furthermore, the method for calculating the health index of each window based on the normalized causal contribution and abnormal characteristic states of the sliding window, performing trend analysis on the health index, and combining the statistical results of the dominant influencing factors to assess the operational health status of the current transformer is as follows: Under each sliding window, the value of the insulation abnormality state is designed as follows: 0 for no insulation abnormality event and 1 for the presence of insulation abnormality event; the value of the partial discharge abnormality state is designed as follows: 0 for no partial discharge abnormality event and 1 for the presence of partial discharge abnormality event; the product of the normalized causal contribution of the insulation abnormality and the insulation abnormality state value, and the product of the normalized causal contribution of the partial discharge abnormality and the partial discharge abnormality state value are summed to obtain the single-window health index of the sliding window. Pre-set warning and emergency thresholds, arrange all single-window health indices sequentially, and set the sliding average window length to continuous. A sliding window, and For odd numbers greater than 5, take itself and its adjacent numbers before and after it. The health index of each window is calculated, and the arithmetic mean is used as the health index of that window. For windows with insufficient beginnings or endings... The positions of each adjacent window are filled in with the health index of the neighboring windows, and then the average value is calculated. when If the health index of one or more windows increases sequentially, and the health index of each window is all above the warning threshold and all below the emergency threshold, or if the health index of any window is above the emergency threshold, then proceed to the next step of judgment. It is a positive integer greater than 3; when the dominant influencing factor is insulation abnormality, the saturation risk of the output current transformer is caused by insulation abnormality; when the dominant influencing factor is partial discharge abnormality, the saturation risk of the output current transformer is caused by partial discharge abnormality. when The health index of one or more windows is less than the warning threshold, and in this... There are no current transformer saturation events in one or more windows, and there are no abnormal factors in the output that could induce current transformer saturation events. when The health index of one or more windows is less than the warning threshold, and in this... If there are current transformer saturation events in one or more windows, the risk of output current transformer saturation is caused by non-insulation abnormalities and partial discharge abnormalities.

[0015] Additionally, a current transformer saturation detection system is provided, characterized in that: the system is used to perform the aforementioned current transformer saturation detection method, including: The timing alignment module is used to synchronously collect current data on the secondary side of the target current transformer and voltage data on the bus connected to it, as well as online insulation monitoring data and partial discharge detection data of the cable connected to the output end of the current transformer. The collected data are normalized according to class, and all normalized data are time-aligned based on a unified timestamp. The anomaly detection module is used to set a fixed-length sliding window for all detection data. For each sliding window, features are extracted from the normalized current, voltage, insulation online monitoring and partial discharge detection data. The features include current distortion rate, voltage over-limit frequency, average insulation parameter and number of partial discharge pulses. Based on the feature extraction results, it is determined whether there are current transformer saturation events, insulation abnormal events and partial discharge abnormal events in the window. The influence quantification module is used to, within each sliding window, based on the constructed Bayesian causal network model, take the current transformer saturation state, insulation abnormality, and partial discharge abnormality as causal network nodes, count the probability of the current transformer saturation event occurring under each node state, calculate the causal contribution of each influencing factor to the saturation event, normalize all causal contributions, and determine the dominant influencing factor within the window based on the normalization result. The influencing factors include insulation abnormality and partial discharge abnormality. The health status assessment module is used to calculate the health index for each window based on the normalized causal contribution and abnormal characteristic status of the sliding window, perform trend analysis on the health index and combine it with the statistical results of the dominant influencing factors to assess the operational health status of the current transformer.

[0016] Compared with the prior art, the beneficial effects of the present invention are: This invention collects current data from the secondary side of the target current transformer, voltage data from its connected bus, online insulation monitoring data, and partial discharge detection data to determine whether there are abnormal events such as current transformer saturation, cable insulation, or partial discharge. It then constructs a Bayesian causal network to calculate the normalized causal contribution of insulation and partial discharge abnormalities to saturation, thereby obtaining the health index of the current transformer and accurately assessing its operational health. Finally, it combines the dominant influencing factors to determine whether poor current transformer operational health is caused by insulation and partial discharge abnormalities. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the overall method flow of the present invention; Figure 2 This is a graph showing the normalized current and voltage data of the present invention; Figure 3 This is a graph showing the normalized dielectric loss and partial discharge data of this invention; Figure 4 This is a schematic diagram of the package and probability of the present invention; Figure 5 This is a causal contribution statistics chart for the present invention; Figure 6 This is a schematic diagram of the overall system structure of the present invention. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0019] It should be noted that, unless otherwise defined, the technical or scientific terms used in this invention should have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms "first," "second," and similar terms used in this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0020] Example: Please see Figures 1 to 5 The present invention provides a technical solution: A method for detecting saturation of a current transformer, comprising the following steps: Step 1: Synchronously collect the current data of the secondary side of the target current transformer and the voltage data of the bus connected to it, as well as the online insulation monitoring data and partial discharge detection data of the cable connected to the output end of the current transformer. Normalize the collected data according to the categories, and perform time-series alignment of all normalized data based on a unified timestamp. The current data of the secondary side of the target current transformer and the voltage data of the bus connected to it are collected simultaneously, as well as the online insulation monitoring data and partial discharge detection data of the cable connected to the output end of the current transformer. While collecting the current data of the secondary side of the target current transformer and the voltage data of the bus connected to it, the timestamp of each time is recorded. The voltage data of the bus is the target phase voltage of the cable where the current transformer is located, so as to distinguish whether the current transformer saturation is caused by inherent defects in the cable itself or voltage fluctuation interference. In the online insulation monitoring data and partial discharge detection data of the cable connected to the output terminal of the current transformer, the online insulation monitoring data is the dielectric loss factor, and the partial discharge detection data is the partial discharge charge. The dielectric loss factor directly reflects the overall deterioration of the insulation material. It is suitable for long-term online tracking without power outages. When the value is high, it reflects the deterioration of the cable insulation and the increase of insulation leakage current. This is superimposed on the main current of the cable, causing the magnetic flux density of the current transformer core to exceed the saturation magnetic flux density and triggering saturation. The partial discharge charge can quantify the local breakdown intensity of insulation defects. The larger the air gap and the more impurity particles inside the insulation, the larger the value of the partial discharge charge. When the partial discharge value is high, it may cause insulation breakdown, triggering a cable short circuit and causing the current transformer to saturate. During normal operation, the secondary current of a current transformer fluctuates around the rated current. Therefore, using the rated current of the secondary side of the current transformer as a reference, the minimum value is 5% of the rated current. Ordinary measuring current transformers can maintain an acceptable error level at 5% of the rated current. Furthermore, when the real-time current is lower than 5% of the rated current, it may indicate problems such as an open circuit in the secondary circuit or abnormal light load. The maximum value is 20 times the rated current. The short-circuit current multiple is usually 10-30 times the rated current; using the upper limit of 20 times can cover most fault scenarios. Min-max normalization is used for current data processing. Using the rated bus voltage as a reference, the minimum value is 80% of the rated bus voltage, and the maximum value is 120% of the rated bus voltage. 80%-120% covers normal fluctuation scenarios. Min-max normalization is used to complete the current... For voltage data processing, when the real-time monitored voltage is below 80% of the rated voltage, it is normalized to 0; when it is above 120% of the rated voltage, it is normalized to 1. For insulation online monitoring data, the rated allowable dielectric loss factor is used as the benchmark. When the dielectric loss factor is below 10% of the rated allowable value, it is calculated as 10% of the rated allowable value; when the dielectric loss factor is above the rated allowable dielectric loss factor, it is calculated as the rated allowable dielectric loss factor. The min-max normalization method is used to complete the insulation online monitoring data processing. The maximum and minimum values ​​of the partial discharge charge amplitude threshold are preset. The allowable limits for partial discharge are different for different cables, which can avoid the reduction in accuracy caused by setting thresholds based on experience. The min-max normalization method is used to complete the partial discharge detection data processing. When the local discharge charge is below the minimum value, it is normalized to 0; when the local discharge charge is above the maximum value, it is normalized to 1. Using the sampling frequencies of current and voltage data as reference frequencies, the interval between adjacent time markers is calculated to generate a continuous unified timestamp sequence. Each time marker in this sequence is defined as a target time point, i.e., a unified time reference point for all data to be aligned. The time point for partial discharge detection is defined as the partial discharge sampling time point. Current and voltage data are directly matched to the target time points. Since current and voltage are high-frequency continuous monitoring signals, the dielectric loss factor is a low-frequency slowly changing signal, and partial discharge is a high-frequency pulse discrete signal, the dielectric loss factor data and partial discharge charge data are time-aligned by covering all target time points in the unified timestamp sequence. Specifically, for the dielectric loss factor, a linear interpolation method is used to calculate the supplementary value. The dielectric loss factor has continuous and slowly changing characteristics, and is approximately linearly changing. Therefore, based on two adjacent dielectric loss factors, combined with the time interval ratio between the target time point and the two dielectric loss factors, the dielectric loss factor supplementary value for the target time point is obtained through linear calculation. This is then applied to the insulation. The physical laws of degradation; for partial discharge charge, due to the characteristics of its discrete pulse signal, the time difference between each target time point and all original partial discharge sampling time points is calculated. When the time difference between a certain original partial discharge sampling time point and the target time point does not exceed half of the reference frequency time interval, the partial discharge charge corresponding to the original partial discharge sampling time point is assigned to the target time point to ensure that each pulse corresponds to only one target time point and avoid the same pulse being counted repeatedly by multiple target time points. When the time difference between all original partial discharge sampling time points and the target time point exceeds half of the reference frequency time interval, 0 is filled in at the target time point to represent the state of no discharge at that time point, which fits the discrete existence of partial discharge. Table 1 shows the basic data statistics table generated after the secondary side current, bus voltage, insulation online monitoring data and partial discharge detection data collected by 40 sets of target current transformers in actual operation are filled and normalized according to the above rules.

[0021] like Figures 2-3 As shown, when the coverage time is 0.495 seconds and the sampling interval is 5ms, the collected normalized current, normalized voltage, normalized dielectric loss factor, and normalized partial discharge data range from 0.0169 to 0.7870, which does not exceed the range of 0.05-20 times the rated current. For voltage data, there are 3 data points that are 120% higher than the rated voltage. Among the dielectric loss factor data, there are 10 data points that are higher than the rated allowable dielectric loss factor. In number 5, the partial discharge charge is the largest, which can help to judge the operating health status of the current transformer in the future.

[0022] Step 2: Set a fixed-length sliding window for all detection data. For each sliding window, extract features from the normalized current, voltage, insulation online monitoring and partial discharge detection data. The features include current distortion rate, voltage over-limit frequency, average insulation parameter and number of partial discharge pulses. Based on the feature extraction results, determine whether there are current transformer saturation events, insulation abnormal events and partial discharge abnormal events in the window. A fixed-length sliding window is set, for example, 200ms. When the power system frequency is 50Hz, this includes 10 complete fundamental cycles, covering the periodic characteristics of current and voltage signals. The window sliding step size is consistent with the window length, forming a non-overlapping continuous sliding window sequence. Normalized current data for all time steps within the sliding window is extracted. A Fast Fourier Transform (FFT) is performed on this continuous current data segment. Specifically, the current data is padded with zeros to powers of 2. For example, when the sampling frequency is 5000Hz and the sliding window length is 200ms, zeros are padded to 1024 points. The forward DFT result of the zero-padded current sequence is then converted to a complex sequence.

[0023] This is represented as a sequence of current data padded with zeros to the nearest power of 2. Represented as the imaginary unit, It is represented as a complex exponential basis function, where, Represents discrete frequency point indices. This represents the length of the power of the current data sequence padded with zeros to the nearest power of 2. Indicates the index of the sampling point in discrete time. This represents the first current data sequence before zero-padding within the sliding window. The FFT process takes discrete-time sampling points, then uses a butterfly operation to output a complex number sequence, consisting of real and imaginary parts. The real part of each complex number is squared, and then the imaginary part is also squared. The two squared results are added together, and the square root of the sum is taken to obtain the amplitude. Since the spectrum obtained after FFT processing is symmetrical, the analysis only needs to focus on the first half of the frequency points. For the first half of the frequency points, except for the first point representing the DC component, all other frequency points need to be multiplied by 2 and divided by the total length after zero-padding to reconstruct the actual signal strength. This is done using the index of each point. Multiply by the sampling frequency and divide by the total length after zero padding to obtain the indexes of the closest frequency points for the fundamental, 3rd, 5th, 7th, and 9th harmonics. The corresponding amplitudes represent the effective current values ​​of the corresponding frequency components. First, square the effective values ​​of the 3rd, 5th, 7th, and 9th harmonics respectively, then add these squared results together and take the square root of the sum to obtain the total effective harmonic value. Then, divide this total effective harmonic value by the effective value of the fundamental frequency. The result is the current distortion rate within the sliding window, which reflects the degree to which the current waveform deviates from a sine wave. Extract the normalized voltage data for all time steps within the sliding window, and count the number of data points with a normalized voltage value of 0 or 1 within the window. This number is the frequency of voltage over-limit in the sliding window. Extract the normalized dielectric loss factor data for all time steps within the sliding window, perform an arithmetic mean operation on the normalized dielectric loss factor data within the window, and use the result as the mean value of the insulation parameters for the sliding window. Extract the normalized partial discharge charge data for all time steps within the sliding window, and count the number of data points with non-zero normalized partial discharge charge within the window. This number represents the number of partial discharge pulses in the sliding window.

[0024] The preset current distortion threshold, dielectric loss factor threshold, and partial discharge pulse threshold are configured. The specific values ​​of the thresholds need to be adjusted in conjunction with industry standards and the adaptability of the equipment itself. For example, industry standards stipulate that the limit for the total current distortion rate of industrial equipment is 0.05, so the current distortion threshold can be preset to 0.05. Industry standards also stipulate that the dielectric loss factor of 10kV and below XLPE cables should be less than 0.005, and the dielectric loss factor of 35kV and above XLPE cables should be less than 0.003. Based on the rated allowable value when normalizing the dielectric loss factor, the dielectric loss factor threshold can be set to 0.6, corresponding to an actual dielectric loss of 0.003. 10kV cables have thinner insulation and are more prone to partial discharge, so the threshold can be set to 8 times. 220kV cables have thicker insulation, so the threshold can be set to 12 times.

[0025] When the current distortion rate of the sliding window exceeds the preset current distortion threshold, a current transformer saturation event is determined to exist within that window, because an excessively high distortion rate indicates severe distortion of the current waveform, which conforms to saturation characteristics. When the average insulation parameter of the sliding window exceeds the preset dielectric loss factor threshold, and the voltage over-limit frequency is 0, an insulation abnormality event is determined to exist within that window. The fact that the dielectric loss factor still exceeds the set threshold when the voltage is normal indicates that the insulation itself is deteriorating, rather than due to voltage interference. When the number of partial discharge pulses in the sliding window exceeds the preset partial discharge pulse threshold, and the voltage over-limit frequency is 0, a partial discharge abnormality event is determined to exist within that window. If partial discharge is still frequent even when the voltage is normal, it indicates that the discharge is caused by insulation defects rather than voltage interference. Table 2 shows the results of the judgment of whether there are current transformer saturation events, cable insulation abnormal events and partial discharge abnormal events in each window, based on the current distortion rate, voltage over-limit frequency, average insulation parameter and partial discharge pulse number characteristics extracted in each sliding window. The threshold values ​​for current distortion rate are set to 0.05, voltage over-limit frequency is 1 (voltage over-limit frequency is not 0), dielectric loss factor is 0.6 and partial discharge pulse is 3.

[0026] Step 3: Within each sliding window, based on the constructed Bayesian causal network model, the current transformer saturation state, insulation abnormality, and partial discharge abnormality are taken as causal network nodes. The probability of the current transformer saturation event occurring under each node state is statistically analyzed. The causal contribution of each influencing factor to the saturation event is calculated, and all causal contributions are normalized. The dominant influencing factor within the window is determined based on the normalization result. The influencing factors include insulation abnormality and partial discharge abnormality. After statistically analyzing the insulation anomaly status, partial discharge anomaly status, and saturation event status of all sliding windows, each sliding window is marked as follows: 1 for the presence of an insulation anomaly event, 0 for its absence; 1 for the presence of an insulation anomaly event, 0 for its absence; 1 for the presence of a partial discharge anomaly event, 0 for its absence; and 1 for the presence of a saturation event, 0 for its absence. Insulation anomaly nodes and partial discharge anomaly nodes are preset. All sliding window data is divided into four non-overlapping groups: insulation anomaly node status 0 and partial discharge anomaly node status 0; insulation anomaly node status 0 and partial discharge anomaly node status 1; insulation anomaly node status 1 and partial discharge anomaly node status 0; and insulation anomaly node status 1 and partial discharge anomaly node status 1. These correspond to windows with neither insulation nor partial discharge anomalies, windows with no insulation anomalies but with partial discharge anomalies, windows with insulation anomalies but no partial discharge anomalies, and windows with both insulation and partial discharge anomalies, respectively, and are labeled as group 1, 2, 3, and 4.

[0027] The total number of windows in each group is counted and recorded as the total number of windows. The number of sliding windows that are determined to have a saturation event is counted and recorded as the number of saturated windows. The ratio of the number of saturated windows in each group to the total number of windows in that group is the probability of the current transformer in that group having a saturation event, recorded as the saturation probability. There are a total of 4 groups of saturation probabilities. At the same time, when the total number of windows in a group is 0, the saturation probability is directly set to 0. These represent the saturation probabilities of having neither insulation abnormality nor partial discharge abnormality, having no insulation abnormality but having partial discharge abnormality, having insulation abnormality but no partial discharge abnormality, and having both insulation abnormality and partial discharge abnormality.

[0028] The influencing factors are defined as insulation anomalies and partial discharge anomalies. For the insulation anomaly state, partial discharge anomaly state, and saturation event state of each sliding window, four groups are divided and saturation probability is calculated. Then, the normalized causal contribution of insulation anomalies and partial discharge anomalies is calculated. When there is no partial discharge anomaly, the difference in saturation probability between the presence and absence of insulation anomaly is calculated. When there is a partial discharge anomaly, the difference in saturation probability between the presence and absence of insulation anomaly is calculated again. The arithmetic mean of these two differences is taken as the causal contribution of the insulation anomaly. The sum of the absolute values ​​of the causal contributions of insulation anomalies and partial discharge anomalies is taken as the total contribution:

[0029] In the formula, Indicates the total contribution. Indicates the causal contribution of insulation abnormalities. Indicates the causal contribution of partial discharge anomalies; The ratio of the absolute value of the insulation anomaly contribution to the total contribution is used as the normalized causal contribution of the insulation anomaly.

[0030] In the formula, This represents the normalized causal contribution of the insulation anomaly. The ratio of the absolute value of the contribution of partial discharge anomalies to the total contribution is used as the normalized causal contribution of partial discharge anomalies:

[0031] In the formula, The normalized causal contribution of partial discharge anomalies, such as Figure 5As shown, in the statistics of the normalized causal contribution of insulation abnormality and the normalized causal contribution of partial discharge abnormality of 40 groups of current transformers, the fluctuation of the normalized causal contribution of insulation and partial discharge of different current transformers is significantly different, so the main influencing factors affecting the operation of current transformers can be intuitively analyzed. when The insulation anomaly is determined to be the dominant influencing factor of the saturation event within the current window. The study determined that partial discharge anomalies are the dominant influencing factor for saturation events within the current window. In this study, insulation abnormalities and partial discharge abnormalities are the dominant influencing factors of saturation events within the current window.

[0032] Step 4: Calculate the health index for each window based on the normalized causal contribution and abnormal characteristic states of the sliding window, perform trend analysis on the health index, and combine it with the statistical results of the dominant influencing factors to assess the operational health status of the current transformer; In each sliding window, the values ​​for insulation anomalies are designed as follows: 0 for no insulation anomaly and 1 for an insulation anomaly. Similarly, the values ​​for partial discharge anomalies are designed as follows: 0 for no partial discharge anomalies and 1 for an partial discharge anomaly. The sum of the product of the normalized causal contribution of the insulation anomaly and the insulation anomaly value, and the product of the normalized causal contribution of the partial discharge anomaly and the partial discharge anomaly value, yields the single-window health index for that sliding window.

[0033] In the formula, Indicates the health index of a single window. Indicates the value of abnormal insulation condition. This indicates the value of the partial discharge abnormal state, due to , The value can be either 0 or 1, therefore, The value ranges from 0 to 1, and a warning threshold is preset. and emergency threshold ,and The value is adjusted according to the equipment characteristics of the current transformer and cable; Arrange all single-window health indices sequentially, and set a sliding average window with a continuous length. A sliding window ensures effective noise filtering, and For odd numbers higher than 5, the average value should precisely correspond to the current window. The number of windows needs to be set in conjunction with the time characteristics and data volatility of the sliding window. The health index of a single window may be affected by occasional interference, such as misjudgment of a single partial discharge or instantaneous noise in the current signal. Covering more windows can make the sliding average health index closer to the actual state of the device. For example, when the sliding window duration is 200ms... It can quickly capture health trend changes within 1-2 seconds, taking the value of itself and adjacent values ​​before and after it. The health index of each window is calculated, and the arithmetic mean is used as the health index of that window. For windows with insufficient beginnings or endings... The positions of each adjacent window are filled in with the health index of the neighboring windows, and then the average value is calculated. when If the health index of one or more windows increases sequentially, and the health index of each window is above the warning threshold and below the emergency threshold, or if the health index of any window is above the emergency threshold, proceed to the next step of judgment. If the health index increases sequentially and the health index of each window is above the warning threshold and below the emergency threshold, it indicates that the insulation or partial discharge abnormality is continuously deteriorating. If the health index of any window is above the emergency threshold, it indicates that the risk of inducing current transformer saturation is sufficiently high. It is a positive integer greater than 3. If it is less than this value, it may be misjudged as an abnormality of the current transformer due to random fluctuations. If it is greater than 3, the influence of short-term fluctuations can be eliminated. when When the health indices of one or more windows increase sequentially, and the health indices of each window are all above the warning threshold and all below the emergency threshold, or the health index of any window is above the emergency threshold, then when the dominant influencing factor is insulation abnormality, the saturation risk of the output current transformer is caused by insulation abnormality; when the dominant influencing factor is partial discharge abnormality, the saturation risk of the output current transformer is caused by partial discharge abnormality; when the dominant influencing factors are both insulation abnormality and partial discharge abnormality, the saturation risk of the output current transformer is caused by both insulation abnormality and partial discharge abnormality. The dominant influencing factor is determined by normalized causal contribution. The higher the contribution, the stronger the causal effect of the abnormality on the increase of saturation risk. Outputting the dominant factor can directly guide maintenance personnel to focus on core abnormalities and avoid blind investigation.

[0034] when The health index of one or more windows is less than the warning threshold, and in this... There are no current transformer saturation events in one or more windows, and there are no abnormal factors that would induce current transformer saturation events at the output. That is, the effects of insulation and partial discharge abnormalities on current transformer saturation are negligible and do not have the ability to induce saturation. when The health index of one or more windows is less than the warning threshold, and in this... If there are current transformer saturation events in one or more windows, the risk of output current transformer saturation is not caused by insulation abnormalities or partial discharge abnormalities. In this case, the impact of insulation and partial discharge abnormalities on current transformer saturation is negligible and they do not have the ability to induce saturation. There are other causes.

[0035] Please see Figure 6The present invention also provides a current transformer saturation detection system for performing the above-described current transformer saturation detection method, comprising: The timing alignment module is used to synchronously collect current data on the secondary side of the target current transformer and voltage data on the bus connected to it, as well as online insulation monitoring data and partial discharge detection data of the cable connected to the output end of the current transformer. The collected data are normalized according to class, and all normalized data are time-aligned based on a unified timestamp. The anomaly detection module is used to set a fixed-length sliding window for all detection data. For each sliding window, features are extracted from the normalized current, voltage, insulation online monitoring and partial discharge detection data. The features include current distortion rate, voltage over-limit frequency, average insulation parameter and number of partial discharge pulses. Based on the feature extraction results, it is determined whether there are current transformer saturation events, insulation abnormal events and partial discharge abnormal events in the window. The influence quantification module is used to, within each sliding window, based on the constructed Bayesian causal network model, take the current transformer saturation state, insulation abnormality, and partial discharge abnormality as causal network nodes, count the probability of the current transformer saturation event occurring under each node state, calculate the causal contribution of each influencing factor to the saturation event, normalize all causal contributions, and determine the dominant influencing factor within the window based on the normalization result. The influencing factors include insulation abnormality and partial discharge abnormality. The health status assessment module is used to calculate the health index for each window based on the normalized causal contribution and abnormal characteristic status of the sliding window, perform trend analysis on the health index and combine it with the statistical results of the dominant influencing factors to assess the operational health status of the current transformer.

[0036] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0037] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented in software, the above embodiments can be implemented, in whole or in part, as a computer program product. Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution.

[0038] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0039] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A method for detecting saturation of a current transformer, characterized in that, The specific steps include: Step 1: Synchronously collect the current data of the secondary side of the target current transformer and the voltage data of the bus connected to it, as well as the online insulation monitoring data and partial discharge detection data of the cable connected to the output end of the current transformer. Normalize the collected data according to the categories, and perform time-series alignment of all normalized data based on a unified timestamp. Step 2: Set a fixed-length sliding window for all detection data. For each sliding window, extract features from the normalized current, voltage, insulation online monitoring and partial discharge detection data. The features include current distortion rate, voltage over-limit frequency, average insulation parameter and number of partial discharge pulses. Based on the feature extraction results, determine whether there are current transformer saturation events, insulation abnormal events and partial discharge abnormal events in the window. Step 3: Within each sliding window, based on the constructed Bayesian causal network model, the current transformer saturation state, insulation abnormality, and partial discharge abnormality are taken as causal network nodes. The probability of the current transformer saturation event occurring under each node state is statistically analyzed. The causal contribution of each influencing factor to the saturation event is calculated and all causal contributions are normalized. The dominant influencing factor within the window is determined based on the normalization result. The influencing factors include insulation abnormality and partial discharge abnormality. Step 4: Calculate the health index for each window based on the normalized causal contribution and abnormal characteristic state of the sliding window, perform trend analysis on the health index and combine it with the statistical results of the dominant influencing factors to assess the operational health status of the current transformer.

2. The current transformer saturation detection method according to claim 1, characterized in that: In the online insulation monitoring data and partial discharge detection data of the cable connected to the output terminal of the current transformer, the online insulation monitoring data is the dielectric loss factor, and the partial discharge detection data is the partial discharge charge.

3. The current transformer saturation detection method according to claim 2, characterized in that: The method for normalizing the collected data according to class is as follows: Using the rated current on the secondary side of the current transformer as a benchmark, the minimum value is set at 5% of the rated current, and the maximum value is set at 20 times the rated current. Min-max normalization is used to process the current data. Using the rated bus voltage as a benchmark, the minimum value is set at 80% of the rated bus voltage, and the maximum value is set at 120% of the rated bus voltage. Min-max normalization is used to process the voltage data. When the real-time monitored voltage is lower than 80% of the rated voltage, it is normalized to 0; when it is higher than 120% of the rated voltage, it is normalized to 1. Insulation online monitoring data is normalized to... Based on the allowable dielectric loss factor, if the dielectric loss factor is less than 10% of the rated allowable value, it is calculated at 10% of the rated allowable value; if the dielectric loss factor is greater than the rated allowable dielectric loss factor, it is calculated at the rated allowable dielectric loss factor. The min-max normalization method is used to complete the online insulation monitoring data processing. The maximum and minimum values ​​of the partial discharge charge amplitude threshold are preset, and the min-max normalization method is used to complete the partial discharge detection data processing. When the partial discharge charge is less than the minimum value, it is normalized to 0; when the partial discharge charge is greater than the maximum value, it is normalized to 1.

4. The current transformer saturation detection method according to claim 3, characterized in that: The method for time-series alignment of all normalized data based on a unified timestamp is as follows: Using the sampling frequencies of current and voltage data as reference frequencies, the interval between adjacent time markers is calculated to generate a continuous unified timestamp sequence. Each time marker in this sequence is defined as a target time point, i.e., a unified time reference point for all data to be aligned. The time point for partial discharge detection is defined as the partial discharge sampling time point. Current and voltage data are directly matched to the target time points. For dielectric loss factor data and partial discharge charge data, time alignment is achieved by covering all target time points in the unified timestamp sequence. Specifically, for the dielectric loss factor, a linear interpolation method is used to calculate the completion value, i.e., based on two adjacent dielectric loss factors... The target time point is calculated by combining the time interval ratio between the target time point and the two dielectric loss factors, and then obtaining the dielectric loss factor filler value through linear calculation. For the partial discharge charge, the time difference between each target time point and all original partial discharge sampling time points is calculated. If the time difference between a certain original partial discharge sampling time point and the target time point does not exceed half of the reference frequency time interval, then the partial discharge charge corresponding to that original partial discharge sampling time point is assigned to the target time point. If the time difference between all original partial discharge sampling time points and the target time point exceeds half of the reference frequency time interval, then 0 is filled in at the target time point.

5. The current transformer saturation detection method according to claim 4, characterized in that: For each sliding window, the method for feature extraction of normalized current, voltage, insulation online monitoring, and partial discharge detection data is as follows: Extract the normalized current data for all time steps within the sliding window, perform a Fourier transform on the continuous current data segment, and separate the effective value of the fundamental current and the effective values ​​of each harmonic current. The ratio of the sum of the root mean square values ​​of the effective values ​​of each harmonic current to the effective value of the fundamental current is used as the current distortion rate of the sliding window. Extract the normalized voltage data for all time steps within the sliding window, and count the number of data points with a normalized voltage value of 0 or 1 within the window. This number is the frequency of voltage over-limit in the sliding window. Extract the normalized dielectric loss factor data for all time steps within the sliding window, perform an arithmetic mean operation on the normalized dielectric loss factor data within the window, and use the result as the mean value of the insulation parameters for the sliding window. Extract the normalized partial discharge charge data for all time steps within the sliding window, and count the number of data points with non-zero normalized partial discharge charge within the window. This number represents the number of partial discharge pulses in the sliding window.

6. The current transformer saturation detection method according to claim 5, characterized in that: The method for determining whether current transformer saturation events, insulation abnormalities, and partial discharge abnormalities exist within a window based on feature extraction results is as follows: The current distortion threshold, dielectric loss factor threshold, and partial discharge pulse threshold are preset. When the current distortion rate of the sliding window exceeds the preset current distortion threshold, it is determined that there is a current transformer saturation event in the window. When the average insulation parameter of the sliding window exceeds the preset dielectric loss factor threshold and the voltage over-limit frequency is 0, it is determined that there is an insulation abnormality event in the window; when the number of partial discharge pulses in the sliding window exceeds the preset partial discharge pulse threshold and the voltage over-limit frequency is 0, it is determined that there is a partial discharge abnormality event in the window.

7. The current transformer saturation detection method according to claim 6, characterized in that: The method for calculating the probability of current transformer saturation events at each node state is as follows: After statistically analyzing the insulation abnormality state, partial discharge abnormality state, and saturation event state of all sliding windows, based on the preset insulation abnormality node and partial discharge abnormality node, all sliding window data are divided into four non-overlapping groups: windows with neither insulation abnormality nor partial discharge abnormality, windows with partial discharge abnormality but no insulation abnormality, windows with insulation abnormality but no partial discharge abnormality, and windows with both insulation abnormality and partial discharge abnormality. The total number of windows in each group is counted and recorded as the total number of windows. The number of sliding windows judged to have a saturation event is counted and recorded as the number of saturated windows. The ratio of the number of saturated windows in each group to the total number of windows in that group is the probability of the current transformer in that group experiencing a saturation event, recorded as the saturation probability.

8. The current transformer saturation detection method according to claim 7, characterized in that: The method for calculating the causal contribution of each influencing factor to the saturation event and normalizing all causal contributions is as follows: The influencing factors are defined as insulation anomaly and partial discharge anomaly. When there is no partial discharge anomaly, the difference in saturation probability between the presence and absence of insulation anomaly is calculated. When there is a partial discharge anomaly, the difference in saturation probability between the presence and absence of insulation anomaly is calculated. The arithmetic mean of these two differences is taken as the causal contribution of insulation anomaly. Similarly, when there is no insulation anomaly, the difference in saturation probability between the presence and absence of partial discharge anomaly is calculated. When there is an insulation anomaly, the difference in saturation probability between the presence and absence of partial discharge anomaly is calculated. The arithmetic mean of these two differences is taken as the causal contribution of partial discharge anomaly. The sum of the absolute values ​​of the causal contribution of insulation anomaly and the absolute values ​​of the causal contribution of partial discharge anomaly is taken as the total contribution. The ratio of the absolute value of the insulation anomaly contribution to the total contribution is taken as the normalized causal contribution of insulation anomaly. The ratio of the absolute value of the partial discharge anomaly contribution to the total contribution is taken as the normalized causal contribution of partial discharge anomaly.

9. A method for detecting saturation of a current transformer according to claim 8, characterized in that: The method for assessing the operational health status of current transformers is as follows: This method involves calculating the health index for each window based on the normalized causal contribution and abnormal characteristic states using a sliding window approach, performing trend analysis on the health index, and combining this with the statistical results of the dominant influencing factors. Under each sliding window, the value of the insulation abnormality state is designed as follows: 0 for no insulation abnormality event and 1 for the presence of insulation abnormality event; the value of the partial discharge abnormality state is designed as follows: 0 for no partial discharge abnormality event and 1 for the presence of partial discharge abnormality event; the product of the normalized causal contribution of the insulation abnormality and the insulation abnormality state value, and the product of the normalized causal contribution of the partial discharge abnormality and the partial discharge abnormality state value are summed to obtain the single-window health index of the sliding window. Pre-set warning and emergency thresholds, arrange all single-window health indices sequentially, and set the sliding average window length to continuous. A sliding window, and For odd numbers greater than 5, take itself and each of its two adjacent numbers. The health index of each window is calculated, and the arithmetic mean is used as the health index of that window. For windows with insufficient beginnings or endings... The positions of each adjacent window are filled in with the health index of the neighboring windows, and then the average value is calculated. when If the health index of one or more windows increases sequentially, and the health index of each window is all above the warning threshold and all below the emergency threshold, or if the health index of any window is above the emergency threshold, then proceed to the next step of judgment. It is a positive integer greater than 3; when the dominant influencing factor is insulation abnormality, the saturation risk of the output current transformer is caused by insulation abnormality; when the dominant influencing factor is partial discharge abnormality, the saturation risk of the output current transformer is caused by partial discharge abnormality. when The health index of one or more windows is less than the warning threshold, and in this... There are no current transformer saturation events in one or more windows, and there are no abnormal factors in the output that could induce current transformer saturation events. when The health index of one or more windows is less than the warning threshold, and in this... If there are current transformer saturation events in one or more windows, the risk of output current transformer saturation is caused by non-insulation abnormalities and partial discharge abnormalities.

10. A current transformer saturation detection system, characterized in that: The system is used to perform a current transformer saturation method according to any one of claims 1-9: The timing alignment module is used to synchronously collect current data on the secondary side of the target current transformer and voltage data on the bus connected to it, as well as online insulation monitoring data and partial discharge detection data of the cable connected to the output end of the current transformer. The collected data are normalized according to class, and all normalized data are time-aligned based on a unified timestamp. The anomaly detection module is used to set a fixed-length sliding window for all detection data. For each sliding window, features are extracted from the normalized current, voltage, insulation online monitoring and partial discharge detection data. The features include current distortion rate, voltage over-limit frequency, average insulation parameter and number of partial discharge pulses. Based on the feature extraction results, it is determined whether there are current transformer saturation events, insulation abnormal events and partial discharge abnormal events in the window. The influence quantification module is used to, within each sliding window, based on the constructed Bayesian causal network model, take the current transformer saturation state, insulation abnormality, and partial discharge abnormality as causal network nodes, count the probability of the current transformer saturation event occurring under each node state, calculate the causal contribution of each influencing factor to the saturation event, normalize all causal contributions, and determine the dominant influencing factor within the window based on the normalization result. The influencing factors include insulation abnormality and partial discharge abnormality. The health status assessment module is used to calculate the health index for each window based on the normalized causal contribution and abnormal characteristic status of the sliding window, perform trend analysis on the health index and combine it with the statistical results of the dominant influencing factors to assess the operational health status of the current transformer.

Images