A transformer diagnosis method and system integrating dielectric loss and partial discharge online detection

Abstract

The application discloses a transformer diagnosis method and system integrating dielectric loss and partial discharge online detection, and relates to the technical field of transformer fault prediction. The method comprises the following steps: obtaining dielectric loss parameters and partial discharge parameters in a sliding time window, respectively calculating a corrected dielectric loss index and a partial discharge intensity index; establishing an insulation state curve with the corrected dielectric loss index as the horizontal coordinate and the partial discharge intensity index as the vertical coordinate; calculating the direction angle of the curve and the horizontal coordinate axis, and determining that the overall slow aging is when the direction angle is less than a first threshold value, the insulation aging is obvious when the direction angle is between the first threshold value and a second threshold value, and the local defect mutation is when the direction angle is greater than or equal to the second threshold value. The application can accurately distinguish the two fault modes of overall aging and local defect mutation by comprehensively reflecting the relative rate of dielectric loss and partial discharge change and introducing interference suppression and interphase comparison mechanism, improve the diagnosis accuracy and anti-interference ability, and reduce the false alarm rate.

Description

Technical Field

[0001] This invention relates to the field of transformer fault prediction technology, specifically to a transformer diagnostic method and system that integrates online detection of dielectric loss and partial discharge. Background Technology

[0002] Transformers are critical equipment in power systems, and their insulation condition directly affects the safety and reliability of the power grid. During long-term operation, transformers are affected by factors such as electrical, thermal, mechanical stress, and environmental humidity, which can gradually cause problems such as moisture absorption, aging, cracking, or localized defects in the insulation materials. When insulation defects develop further, they can easily lead to partial discharge, increased dielectric loss, and even serious faults such as insulation breakdown. Therefore, online monitoring and timely diagnosis of transformer insulation conditions are of great significance.

[0003] Currently, online detection methods for transformer insulation status mainly include dielectric loss detection and partial discharge detection. Dielectric loss factor reflects the overall macroscopic deterioration of transformer insulation, such as moisture absorption and aging, and is suitable for evaluating the overall health level of the insulation system. Partial discharge detection, on the other hand, can respond sensitively to local defects such as oil bubble discharge, solid insulation cracks, and point discharge, and is suitable for detecting early local insulation defects.

[0004] While existing digital dielectric loss testers and ultra-high frequency partial discharge online monitoring devices can acquire dielectric loss parameters and partial discharge parameters online, respectively, they are typically used as relatively independent monitoring methods. Comprehensive assessment of transformer insulation status still primarily relies on manual judgment by operation and maintenance personnel based on multiple monitoring indicators, historical trends, and field experience. This approach is not only highly dependent on personnel experience but also difficult to achieve 24 / 7, continuous, and automated diagnostics. Furthermore, in the event of sudden insulation deterioration in a transformer, reliance on manual monitoring and experience comparison can easily lead to response delays, making it difficult to promptly identify the development stage and severity of insulation degradation, thus affecting the timeliness of alarms and the accuracy of operation and maintenance decisions. Summary of the Invention

[0005] To address the technical problems of existing technologies being highly dependent on human experience and having slow response times, this application provides a transformer diagnostic method integrating online detection of dielectric loss and partial discharge, comprising the following steps:

[0006] Obtain dielectric loss parameters and partial discharge parameters within the sliding time window, calculate and correct dielectric loss index based on real-time dielectric loss parameters, and calculate partial discharge intensity index based on real-time partial discharge parameters.

[0007] A dielectric loss-partial discharge correlation curve is established with the modified dielectric loss index as the horizontal axis and the partial discharge intensity index as the vertical axis.

[0008] Calculate the orientation angle between the insulation state curve and the horizontal axis within the current time window. When the orientation angle is less than the first threshold, it is determined to be overall slow aging; when the orientation angle is greater than or equal to the first threshold and less than the second threshold, it is determined to be obvious insulation aging; when the orientation angle is greater than or equal to the second threshold, it is determined to be a sudden change in local defects.

[0009] This invention also provides a transformer diagnostic system integrating online detection of dielectric loss and partial discharge, comprising:

[0010] The data acquisition module is used to acquire dielectric loss parameters and partial discharge parameters within the sliding time window;

[0011] The index calculation module is used to calculate the corrected dielectric loss index based on the real-time dielectric loss parameters, and to calculate the partial discharge intensity index based on the real-time partial discharge parameters.

[0012] The curve establishment module is used to establish a dielectric loss and partial discharge correlation curve with the modified dielectric loss index as the horizontal axis and the partial discharge intensity index as the vertical axis.

[0013] The diagnostic judgment module is used to calculate the direction angle between the dielectric loss partial discharge correlation curve and the horizontal axis within the current time window, and to determine overall slow aging when the direction angle is less than a first threshold; to determine obvious insulation aging when the direction angle is greater than or equal to the first threshold and less than a second threshold; and to determine a sudden change in local defects when the direction angle is greater than or equal to the second threshold.

[0014] The technical effects and advantages of the invention are as follows: By mapping two types of online monitoring quantities, dielectric loss and partial discharge, to the same state plane, and using direction angle, trajectory displacement, persistence, and hysteresis conditions as joint judgment criteria, the automatic identification of transformer insulation status can be achieved without relying on human experience. It can effectively distinguish different deterioration trends such as overall slow aging, obvious insulation aging, and sudden local defects. At the same time, by introducing temperature correction, load correction, partial discharge phase characteristics, smoothing filtering, continuous window confirmation, entry / exit thresholds, and phase difference judgment, the probability of misjudgment caused by external interference and occasional fluctuations can be significantly reduced, improving diagnostic accuracy, real-time performance, and stability. This provides a more accurate basis for transformer operation and maintenance, and reduces equipment operation risks. Attached Figure Description

[0015] Figure 1 This is a schematic diagram of the overall process of the method of the present invention;

[0016] Figure 2 This is a schematic diagram of the overall structure of the system of the present invention;

[0017] Figure 3 This is a schematic diagram of the process for extracting phase distribution characteristics of partial discharge.

[0018] Figure 4 A schematic diagram illustrating the correlation curve between dielectric loss index and partial discharge intensity index, and the determination of the direction angle;

[0019] Figure 5 This is a schematic diagram of the interference suppression, state hysteresis, and continuous window confirmation process.

[0020] Figure 6 This is a schematic diagram of the process for identifying phases in a multiphase fault and suppressing phase differences. Detailed Implementation

[0021] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] The purpose of this invention is to overcome the aforementioned shortcomings in the prior art and provide a transformer diagnostic method and system that integrates online detection of dielectric loss and partial discharge. This method synchronously acquires, extracts features from, normalizes and fuses, and represents the dielectric loss parameters and partial discharge parameters using trajectory mapping. It maps the corrected dielectric loss index, reflecting the overall insulation state of the transformer, and the partial discharge intensity index, reflecting the activity level of partial discharge, onto the same insulation state plane. Furthermore, it automatically judges the transformer insulation state based on trajectory direction angle, trajectory displacement, persistence conditions, and hysteresis conditions, thereby identifying different deterioration stages such as overall gradual aging, significant insulation aging, and abrupt local defect changes.

[0023] Specifically, refer to Figure 1 The method includes the following steps:

[0024] S1. Obtain dielectric loss parameters and partial discharge parameters within the sliding time window, calculate and correct dielectric loss index based on real-time dielectric loss parameters, and calculate partial discharge intensity index based on real-time partial discharge parameters.

[0025] S2. Establish a dielectric loss-partial discharge correlation curve with the modified dielectric loss index as the horizontal axis and the partial discharge intensity index as the vertical axis.

[0026] S3. Calculate the direction angle between the insulation state curve and the horizontal axis within the current time window. When the direction angle is less than the first threshold, it is determined to be overall slow aging; when the direction angle is greater than or equal to the first threshold and less than the second threshold, it is determined to be obvious insulation aging; when the direction angle is greater than or equal to the second threshold, it is determined to be a sudden change in local defects.

[0027] The overall macroscopic degradation process of transformer insulation, such as moisture absorption and aging, is usually relatively slow, manifesting as a long-term, slow increase in the dielectric loss factor. Partial discharge, on the other hand, is more sensitive to local defects in the insulation (such as sharp points, bubbles, and cracks), and its discharge amplitude, repetition rate, and phase distribution characteristics often change rapidly in the middle and later stages of defect development. By simultaneously tracking changes in both dimensions and analyzing their relative rates and coupling relationships, it is possible to effectively distinguish between two distinct fault development modes: "overall slow aging" and "abrupt local defects." This overcomes the technical shortcomings of existing technologies, such as reliance on human experience, delayed response, and difficulty in accurately determining the stage of insulation degradation.

[0028] In the specific implementation of this invention, to achieve the above method, a complete hardware acquisition and data processing system needs to be constructed. (Reference) Figure 2 The system includes a data acquisition module, an indicator calculation module, a curve creation module, an interference suppression module, and a diagnostic judgment module.

[0029] The system comprises several modules: a data acquisition module for synchronously collecting dielectric loss and partial discharge parameters within a sliding time window during transformer operation; an index calculation module for calculating corrected dielectric loss indices based on real-time dielectric loss parameters and partial discharge intensity indices based on real-time partial discharge parameters; a curve establishment module for establishing a dielectric loss-partial discharge correlation curve with the corrected dielectric loss index on the x-axis and the partial discharge intensity index on the y-axis; an interference suppression module for smoothing filtering, minimum displacement threshold, continuous window confirmation, entry / exit thresholds, and phase-to-phase difference judgment; and a diagnostic judgment module for calculating the direction angle between the dielectric loss-partial discharge correlation curve and the x-axis within the current time window. If the direction angle is less than a first threshold, it is judged as overall gradual aging; if the direction angle is greater than or equal to the first threshold but less than the second threshold, it is judged as significant insulation aging; and if the direction angle is greater than or equal to the second threshold, it is judged as a sudden change in local defects.

[0030] In a typical implementation, a wideband current sensor can be installed at the end of the transformer bushing. This sensor has sufficient bandwidth to simultaneously respond to both power frequency current signals (used to calculate the dielectric loss factor) and high-frequency partial discharge pulse signals. The sensor's output signal, after pre-amplification and filtering, is sent to a signal conditioning unit for power frequency and high-frequency separation. The power frequency signal, after amplification and analog-to-digital conversion, is used to calculate the dielectric loss factor for each phase. The high-frequency signal, after being detected, peak held, or sampled at high speed, is used to extract information such as the amplitude, phase, and repetition rate of the partial discharge pulse. The data acquisition module operates continuously using a sliding time window, the length of which can be set according to the actual application scenario, for example, 15 minutes, 30 minutes, or 1 hour. The shorter the window length, the more sensitive the system is to tracking the insulation state, but it may introduce more short-term fluctuations; the longer the window length, the more stable the system's response to trends, but it may lag behind in responding to sudden defects.

[0031] Regarding the calculation of the corrected dielectric loss index, this method does not directly use the original dielectric loss factor obtained from online measurement because the original value is significantly affected by the transformer's operating conditions. The measured value of the transformer's dielectric loss factor is as follows. The dielectric loss factor is closely related to factors such as the temperature of the insulating material, the moisture content in the oil, and the electric field strength it experiences. In particular, changes in the top layer oil temperature and load current can cause significant fluctuations in the dielectric loss factor. Without correction, even a normal temperature fluctuation can cause the dielectric loss factor to exceed the alarm threshold, leading to false alarms. Therefore, this application corrects the measured dielectric loss factor based on temperature and load.

[0032] Includes the following steps:

[0033] S11. Obtain the real-time dielectric loss factor of each phase of the transformer within the sliding time window. Top oil temperature and load current ;

[0034] The top oil temperature can be measured by a platinum resistance temperature sensor or a fiber optic temperature sensor installed on the top of the transformer tank, while the load current can be obtained by sampling from the secondary side of the current transformer.

[0035] S12. Determine the temperature correction term based on the top oil temperature. And determine the load correction term based on the load current. ;

[0036] These two correction functions can be obtained through factory test data or historical operating data. For example, a simple calibration method is to measure the dielectric loss factor at different oil temperatures under laboratory conditions when the transformer leaves the factory, plot the curve of dielectric loss factor versus temperature, and obtain the temperature correction function by fitting the curve. ;

[0037] To address the impact of load current, during the initial operation of the transformer, a period characterized by significant load current fluctuations but relatively stable insulation conditions can be selected. Regression analysis can then be used to extract the influence coefficient of load current on the dielectric loss factor, yielding the load correction function. .

[0038] S13. The real-time dielectric loss factor is corrected using the temperature correction term and the load correction term to obtain the corrected equivalent dielectric loss value under a unified reference state.

[0039] The specific correction formula is as follows:

[0040]

[0041] This formula is modified by subtraction. Its physical meaning is to subtract the additional dielectric loss caused by temperature and load changes from the measured value, thereby reducing the measured value under different operating conditions to the equivalent value under a unified reference temperature (e.g., 20℃) and reference load (e.g., 50% of the rated current).

[0042] S14. Compare the corrected equivalent dielectric loss value within the current sliding time window with the corrected equivalent dielectric loss value under the initial healthy state to obtain the corrected dielectric loss index.

[0043] Initial health status typically refers to the period during which the transformer is initially put into operation or after a major overhaul, and the insulation condition is confirmed to be good. The specific form of the corrected dielectric loss index can be either an absolute difference or a normalized relative value. When using an absolute difference:

[0044]

[0045] This indicator has a definite physical unit and directly reflects the increase in dielectric loss relative to the initial state.

[0046] When using the normalized form:

[0047]

[0048] This dimensionless index is more suitable as the horizontal axis in a two-dimensional trajectory because it eliminates the dimensional influence caused by the different initial dielectric loss values ​​between different transformers, making the subsequent direction angle threshold setting more universal. In a preferred embodiment of the present invention, a normalized modified dielectric loss index is recommended because its dimensionless nature makes the direction angle calculation between the horizontal and vertical axes (partial discharge intensity index is also a dimensionless quantity) more mathematically consistent, and it is beneficial to promote the application of a unified diagnostic threshold among transformers of different voltage levels and capacities.

[0049] Regarding the calculation of the partial discharge intensity index, this invention fully considers three key dimensions of partial discharge: amplitude, activity, and phase distribution, and forms a single dimensionless index through weighted fusion. This enables joint analysis with the modified dielectric loss index in a two-dimensional plane. Specifically, it includes the following steps:

[0050] S15. Acquire partial discharge pulse signals within the sliding time window;

[0051] These signals originate from the high-frequency separation branch of the broadband current sensor and are converted into discrete pulse sequences after high-speed analog-to-digital conversion. The occurrence time of each pulse is recorded. Peak amplitude And information such as polarity.

[0052] S16. Extract the maximum discharge amplitude from the partial discharge pulse signal. Discharge repetition rate and phase distribution characteristic quantities ;

[0053] The maximum discharge amplitude is the maximum amplitude among all pulses within the window, reflecting the extreme intensity of partial discharge; the discharge repetition rate is the ratio of the total number of pulses within the window to the window duration, reflecting the frequency of partial discharge; the phase distribution characteristic describes the distribution pattern of these discharge pulses within the power frequency cycle, and is an important basis for distinguishing partial discharge from external interference.

[0054] S17. Normalize the maximum discharge amplitude, discharge repetition rate and phase distribution characteristics respectively to make them comparable.

[0055] Normalization methods can employ division based on a benchmark value, such as using the maximum permissible discharge amplitude recorded during the transformer's historical operation or the background noise level during factory testing as the benchmark value. , The normalized amplitude ratio and repetition rate ratio are obtained by dividing the maximum discharge amplitude and discharge repetition rate by the reference value, respectively. For phase distribution characteristics, since they are usually between 0 and 1, they can be used directly.

[0056] S18. The normalized maximum discharge amplitude, discharge repetition rate, and phase distribution characteristics are weighted according to preset weights to obtain the dimensionless partial discharge intensity index:

[0057]

[0058] Among them, the weighting coefficient , , satisfy: The specific value can be set according to the site conditions and the sensitivity requirements for different types of defects.

[0059] The partial discharge intensity index constructed in the above manner can characterize the activity and danger of partial discharge within the sliding time window.

[0060] In the calculation of the above partial discharge intensity index, the phase distribution characteristic can quantify the distribution law of the partial discharge pulse on the power frequency phase. ,refer to Figure 3 Steps S161 to S163 describe in detail a method for obtaining phase distribution entropy:

[0061] S161. Within the sliding time window, based on the occurrence time of the partial discharge pulse and the power frequency reference phase, determine the phase angle corresponding to each partial discharge pulse within the power frequency cycle. ;

[0062] The power frequency reference phase can be obtained by zero-crossing detection of the voltage transformer's secondary voltage signal or by phase-locked loop synchronization. Phase angle. The calculation formula is:

[0063]

[0064] In the formula, This is the power frequency, which is 50Hz in China. This is the phase calibration offset, used to compensate for phase delays caused by the sensor and signal transmission path.

[0065] S162. Divide a power frequency cycle into multiple phase intervals, count the number of partial discharge pulses and / or the cumulative discharge amplitude in each phase interval, and obtain the phase distribution vector.

[0066] For example, there could be 36 intervals in 10° intervals, or 24 intervals in 15° intervals, or 12 intervals in 30° intervals. The finer the interval division, the more detailed the description of the phase distribution, but the greater the statistical fluctuations; the coarser the interval division, the better the statistical stability, but subtle phase features may be lost.

[0067] Taking the number of pulses as an example, suppose the number of pulses in the m-th phase interval is... After normalization, we get:

[0068]

[0069] in, This represents the proportion of pulses in the m-th phase interval to the total number of pulses. This represents the total number of phase intervals. Clearly:

[0070]

[0071] S163. Calculate the phase aggregation degree, phase distribution entropy, or similarity with the preset partial discharge phase template based on the phase distribution vector, and use the phase aggregation degree, phase distribution entropy, or similarity as the phase distribution feature quantity.

[0072] The formula for calculating the phase distribution entropy is as follows:

[0073]

[0074] When the discharge pulses are completely randomly distributed throughout the entire power frequency cycle ,at this time Reaching maximum value:

[0075]

[0076] This indicates that the discharge is unrelated to the power frequency phase and most likely originates from external electromagnetic interference. When discharge pulses are concentrated in one or several specific phase intervals... The value is large in these intervals, but very small or zero in other intervals. Significantly smaller than This indicates that the discharge exhibits strong phase concentration, consistent with the characteristics of a typical partial discharge. Therefore, the phase concentration can be defined as:

[0077]

[0078] This value is between 0 and 1. The closer it is to 1, the stronger the phase clustering; the closer it is to 0, the more random the phase distribution.

[0079] As an alternative implementation, the phase distribution characteristics can also be obtained through template matching. In this method, a typical partial discharge phase distribution template library is pre-established. This library may include, but is not limited to, internal discharge templates, surface discharge templates, suspended discharge templates, and corona discharge templates. These templates can be obtained through statistical averaging based on a large amount of historical measured data or laboratory simulated defect data. During online monitoring, the phase distribution vector obtained within the current sliding window is... With the j-th type template in the template library Perform similarity calculations and use the maximum similarity as the phase distribution feature:

[0080]

[0081] Similarity function Cosine similarity can be used, and its calculation formula is as follows:

[0082]

[0083] The cosine similarity value ranges from -1 to 1. However, in practical applications, since all components of the phase distribution vector are non-negative, the similarity is usually between 0 and 1. The closer to 1, the more similar the current phase distribution is to the template. The advantage of template matching is that it not only provides the strength of phase clustering but also indicates which known typical discharge type the current discharge belongs to, providing maintenance personnel with richer fault type information. However, this method requires a sufficiently comprehensive and representative template library to be established in advance, and the template library needs to be updated regularly to cover more actual discharge modes in the field. Therefore, in specific implementations, an appropriate method for obtaining phase distribution features can be selected based on the complexity of the application scenario. For most field monitoring scenarios, the clustering calculation method based on phase distribution entropy has the advantages of simple calculation, no need for prior templates, and wide adaptability.

[0084] After obtaining a continuous sequence of corrected dielectric loss indicators and partial emission intensity index sequence Afterwards, the system enters the stage of establishing the state curve and calculating the direction angle. In step S2, the state points corresponding to each sliding time window are plotted sequentially on a two-dimensional plane with the modified dielectric loss index as the abscissa and the partial discharge intensity index as the ordinate. These points are then connected in chronological order to form the dielectric loss and partial discharge correlation curve. The geometric shape and trend of this curve contain rich information about the evolution of the insulation state. Specifically, if the insulation is in the overall slow aging stage, that is, the dielectric loss slowly increases due to the overall moisture or thermal aging of the insulation material, while partial discharge has not yet occurred in large quantities or is only caused by weak discharge by aging products, then as time goes by, the abscissa X will gradually increase, while the change in the ordinate Y is relatively small, and the state trajectory will slowly move to the right along a direction close to the abscissa. If local defects appear inside the insulation (such as solid insulation cracks or bubbles in oil), and the defects are gradually developing, then when the defects develop to a certain extent, the partial discharge will be significantly enhanced, and the ordinate Y will rise rapidly. At this time, the state trajectory will show that while the abscissa increases, the ordinate increases sharply, and the trajectory direction deviates significantly. To quantitatively characterize this trajectory direction deflection, in step S3, the orientation angle between the insulation state curve and the horizontal axis within the current time window is calculated. The orientation angle is defined as follows: the difference between the current state point and the previous state point is used to obtain the state change vector. Then calculate the angle between the vector and the positive direction of the x-axis, i.e.:

[0085]

[0086] In actual monitoring, the changes in the corrected dielectric loss index and partial discharge intensity index are usually non-negative, meaning that insulation degradation is a unidirectional deterioration process. Therefore, the directional angle... Typically located between 0° and 90°. When the direction angle When the value is relatively small, it indicates that the dielectric loss increases relatively significantly while the local discharge increases relatively slowly, corresponding to overall gradual aging; when When the value gradually increases, it indicates that the rate of increase in partial discharge gradually exceeds the rate of increase in dielectric loss, corresponding to significant insulation aging or the development of localized defects; when When the azimuth angle increases to nearly 90°, it indicates a sharp increase in partial discharge while the dielectric loss changes relatively little, corresponding to a sudden change in local defects. The physical significance of the azimuth angle lies in its reflection of the relative rate of change between the two types of monitoring indicators. If only the absolute value of a single indicator is observed, it is difficult to determine whether the current state change is a normal operational fluctuation or a true insulation degradation. For example, a small increase in the amplitude of a partial discharge pulse may simply be a fluctuation in background noise, but if the dielectric loss indicator does not increase synchronously with the increase in partial discharge, and the azimuth angle is already close to 90°, then the increase in partial discharge is more likely a true discharge caused by a local defect. Therefore, as a ratio indicator of changes in two dimensions, the azimuth angle naturally has the advantage of suppressing common interference and highlighting differential changes.

[0087] In actual measurement processes, since the change in dielectric loss is not significant in most cases, the calculated... Typically, the variation in dielectric loss of a well-functioning transformer is only on the order of 0.01 over several days or even months. However, the fluctuation in partial discharge intensity caused by partial discharge phenomena within the same time period is on the order of magnitude higher than that of the corrected dielectric loss index. This results in the direction angle calculated within the sliding time window often being close to 90°. Therefore, a scale equalization coefficient K is introduced to ensure that the direction angle of the scatter plot under healthy conditions is at a relatively small angle. The direction angle calculation formula is then modified as follows:

[0088]

[0089] The scale equalization coefficient K is obtained by adjusting the corrected dielectric loss index and partial discharge intensity index of the transformer under healthy conditions obtained on a daily basis.

[0090] In order to map the value of the orientation angle to the specific diagnostic status, this method sets a first threshold. Second threshold .

[0091] refer to Figure 4 ,when When the insulation reaches a certain level, it is determined to be in a state of overall slow aging. Under this state, the dielectric loss factor of the insulation shows a slow increasing trend, but the partial discharge activity is not active, or the increase in the partial discharge level is synchronized with the increase in dielectric loss. The overall insulation is in a uniform deterioration process, which usually corresponds to the normal thermal aging or uniform moisture absorption of the insulation material. At this time, no immediate maintenance measures are required, but monitoring should continue.

[0092] when At this point, the insulation is considered to be in a state of significant aging. In this state, the directional angle has exceeded the first threshold, indicating that the rate of increase in partial discharge intensity has begun to exceed the rate of increase in dielectric loss. There may be some degree of localized defects or uneven aging areas within the insulation, and partial discharge is accelerating. This state is a critical early warning stage, prompting maintenance personnel to pay attention to the partial discharge trend and, if necessary, arrange further power outage tests or online monitoring and analysis.

[0093] when When this occurs, it is determined to be a local defect abrupt change state. In this state, the directional angle is close to 90°, indicating that the partial discharge intensity has increased explosively in a short period of time, while the dielectric loss index has not changed much. This usually corresponds to a precursor to local breakdown inside the insulation, such as bubble discharge developing into surface discharge, or the appearance of penetrating cracks inside the solid insulation. This state is characterized by its suddenness and rapid development. If not handled in time, it may develop into an insulation breakdown accident within hours to days. Therefore, the system should immediately issue a high-priority alarm signal, and it is recommended that maintenance personnel take measures as soon as possible.

[0094] First threshold Second threshold The setting method directly affects the accuracy and reliability of the diagnosis. In this invention, it is recommended to use an adaptive threshold determination method based on the device's own historical data, rather than a fixed universal threshold, because different transformers have different insulation structures, operating years, historical operating conditions, and sensor installation methods, and a fixed threshold is difficult to adapt to diverse field requirements.

[0095] Specifically, for the first threshold It is used to characterize the boundary between the overall gradual aging of equipment and the stage of simultaneous growth of dielectric loss and partial discharge. During the initial commissioning of a transformer or for a period after a major overhaul, the insulation is in a healthy operating phase, and the calculated azimuth sequence reflects the fluctuation range under normal conditions. Azimuth data can be collected during this period, and its mean can be calculated. and standard deviation Then the first threshold is set as follows:

[0096]

[0097] in, 2 to 3 are acceptable.

[0098] This setting method, based on the mean plus a certain number of standard deviations, ensures that the azimuth angle will not easily exceed the first threshold within the normal fluctuation range, thus avoiding false alarms. As another implementation method, a quantile method can also be used, that is, selecting the high quantile of the azimuth angle distribution during the healthy operation phase as the first threshold, for example:

[0099]

[0100] in, The first direction angle of the historical healthy operation stage Quantiles, such as the 85th quantile, 90th quantile, or 95th quantile.

[0101] The advantage of the quantile method is that it does not require the assumption that the orientation angle follows a normal distribution, and it is more adaptable to asymmetric data.

[0102] For the second threshold It is used to characterize the boundary where the partial discharge intensity grows abnormally rapidly relative to the change in dielectric loss. This boundary usually corresponds to the critical point where partial discharge transitions from the development stage to the abrupt change stage. The second threshold can be determined based on the statistical distribution of the azimuth angle in historical partial discharge anomaly samples or field-simulated defect samples. For example, for a specific transformer, if partial discharge anomaly events have occurred in its operating history, and these events have been confirmed by maintenance personnel through other means (such as ultrasonic detection or oil chromatography analysis), then the azimuth angle data for a period of time before and after these anomaly events can be extracted, their distribution statistically analyzed, and then the lower quantile of this distribution can be taken as the second threshold, for example:

[0103]

[0104] in, This is the set of orientation angles corresponding to samples with local defect mutations. You can use the 5th percentile, 10th percentile, or 15th percentile.

[0105] The essence of this approach is to ensure that the second threshold can cover most known anomalous events, so that when the orientation angle reaches the threshold, there is a high degree of confidence in determining it as a local defect mutation.

[0106] In actual operation, the azimuth sequence may be affected by various factors, resulting in short-term fluctuations. For example, an external transient interference pulse may cause a brief increase in the partial discharge intensity index within a certain window, thereby temporarily increasing the azimuth angle of that window and exceeding the threshold. However, if the system immediately changes its diagnostic state at this time, false alarms and frequent state jumps will occur, seriously affecting user experience and alarm reliability. To suppress such occasional interference and short-term fluctuations, the method provided by this invention also includes interference suppression, as described above. Figure 5 Specifically, it includes the following steps:

[0107] S4. Direction angles corresponding to multiple consecutive sliding time windows Perform smoothing filtering or median filtering to obtain the smoothed direction angle. ;

[0108] Smoothing filtering can be achieved using an exponentially weighted moving average, i.e.:

[0109]

[0110] In the formula, Let be the smooth direction angle of the j-th sliding window. The original orientation angle calculated for the j-th sliding window. The smoothing coefficient can be set to 0.6 to 0.9.

[0111] Median filtering takes the median of the direction angles of the most recent q windows as the current smoothing value, that is:

[0112]

[0113] Median filtering has an excellent suppression effect on isolated impulse interference. For example, if a window's orientation angle suddenly jumps to an abnormally high value due to electromagnetic interference, median filtering can exclude this abnormal value from the decision-making process without introducing additional delay.

[0114] S5. Calculate the state trajectory displacement based on the corrected dielectric loss index and partial discharge intensity index corresponding to multiple consecutive sliding time windows:

[0115]

[0116] in, and These represent the changes in the modified dielectric loss index and the partial discharge intensity index between two adjacent windows, respectively.

[0117] Only when the state trajectory is displaced Only when the preset minimum displacement threshold is reached is the current trajectory change considered sufficient to support the update of the diagnostic state. Otherwise, it is considered that the current state point has hardly moved substantially. Even if the calculated direction angle changes significantly, it is entirely possible that it is a false change caused by a small numerical fluctuation. In this case, the system should maintain the previous diagnostic state. Minimum displacement threshold The value of can be determined based on the noise levels of the corrected dielectric loss index and the partial discharge intensity index. It can usually be taken as 2 to 3 times the square root of the sum of the squares of the standard deviations of the noise measurements of the two.

[0118] Furthermore, the interference suppression mechanism also includes a continuous window confirmation rule. Even if the smoothing direction angle has reached a threshold condition for a certain diagnostic state, the system will not immediately switch states, but will require that the threshold condition to remain true for a preset number of consecutive sliding time windows. For example, it can be set that when the smoothing direction angles of three consecutive windows meet the condition... Only when the smooth direction angle of three consecutive windows meets the condition is it confirmed to have entered the "significant insulation aging" state. Only when this condition is met is it confirmed that the "local defect mutation" state has been entered. This continuous confirmation rule makes full use of the continuous characteristic of actual insulation degradation and effectively eliminates short-lived, single, or occasional interference.

[0119] To further reduce the risk of frequent switching of diagnostic states near the threshold, this invention also introduces a threshold comparison mechanism with hysteresis characteristics. Specifically, a first entry threshold is set for the first threshold. and the first exit threshold ,and Set a second entry threshold for the second threshold. Second exit threshold ,and .

[0120] When the smoothing direction angle is greater than or equal to the first threshold Then, the state transition determination process begins. This occurs when the smoothing direction angle gradually increases from a low value and crosses the first entry threshold. Only then does the system enter a state of significant insulation aging; thereafter, even if the orientation angle drops slightly near the threshold, as long as it remains above the first exit threshold... The system will remain in a state of significant insulation aging without reverting to normal operation. It will only resume operation if the azimuth angle falls further below the first exit threshold. The system exits the obvious insulation aging state only when preset degradation conditions are continuously met (e.g., multiple consecutive windows below the exit threshold). Similarly, a second entry threshold is required to transition from the obvious insulation aging state to the local defect abrupt change state. However, to revert from a state of abrupt local defects to a state of significant insulation aging, a second exit threshold must be used. Setting the entry threshold higher than the exit threshold ensures that state transitions are directionally dependent. That is, upward transitions (intensified degradation) and downward transitions (mitigated degradation or recovery from false alarms) occur at different threshold levels. This prevents the system from repeatedly changing its state when the directional angle repeatedly crosses around the threshold, significantly improving the stability and reliability of diagnostic conclusions. The difference between the entry and exit thresholds needs to be adjusted according to the actual application. A difference that is too small will result in insignificant hysteresis, while a difference that is too large may lead to a delayed response to real rapid degradation. As a preferred implementation, the first entry threshold can be... and the first exit threshold The first entry threshold is set symmetrically around the first threshold. Compared to the first exit threshold Increase by 3° to 5°. Similarly, set the second entry threshold... Second exit threshold The second entry threshold is set symmetrically around the second threshold. Compared to the second exit threshold 3° to 5° higher.

[0121] Furthermore, in the process of interference suppression based on the entry threshold and exit threshold, preset confirmation conditions and preset degradation conditions are set for the state switching to be satisfied. The state switching action can only be performed when the smoothing direction angle not only satisfies the comparison relationship between the entry threshold and exit threshold, but also satisfies the preset confirmation condition or preset degradation condition.

[0122] The specific rules are as follows:

[0123] When the smoothing direction angle reaches the first entry threshold and continues to meet the preset confirmation conditions, the system enters the obvious insulation aging state; when the smoothing direction angle is lower than the first exit threshold and continues to meet the preset degradation conditions, the system exits the obvious insulation aging state.

[0124] When the smoothing direction angle reaches the second entry threshold and continues to meet the preset confirmation conditions, the system enters the local defect mutation state; when the smoothing direction angle is lower than the second exit threshold and continues to meet the preset degradation conditions, the system exits the local defect mutation state.

[0125] Specifically, the preset confirmation condition is defined as the number of sliding time windows with a smoothing direction angle greater than the entry threshold being greater than the minimum number of windows. The preset degradation condition is defined similarly to the confirmation condition, but in the opposite direction, i.e., the number of sliding time windows with a smoothing direction angle less than the exit threshold being greater than the minimum number of windows.

[0126] In the logical judgment of entering or exiting a specific diagnostic state, the present invention sets different combinations of conditions for obvious insulation aging state and local defect sudden change state, so as to make full use of their respective characteristic information.

[0127] For entry into a state of significant insulation aging, the smoothing direction angle corresponding to multiple consecutive sliding time windows must be greater than or equal to the first entry threshold, and the state trajectory displacement must be greater than the minimum displacement threshold. The purpose of requiring a displacement threshold here is to exclude cases where the direction angle accidentally exceeds the limit due to data fluctuations. For example, when... and When the values ​​are very small, the calculation of the azimuth angle is inherently unstable. Even if the azimuth angle exceeds the limit, it has no practical physical meaning. Therefore, it must be excluded by a displacement threshold. The condition for exiting the state of obvious insulation aging is relatively simple. It only requires that the smoothed azimuth angle corresponding to multiple consecutive sliding time windows be less than the first exit threshold. This is because when returning from a more severe state to a less severe state, more caution is required, but no additional displacement condition is needed, because the azimuth angle being continuously lower than the exit threshold itself already indicates a significant change in the state trend.

[0128] For the entry into a local defect abrupt state, the conditions are more stringent. In addition to requiring the smoothing direction angle to be greater than or equal to the second entry threshold, the change in the partial discharge intensity index is also required. Greater than the preset partial discharge change threshold Simultaneously, the displacement of the state trajectory exceeds the minimum displacement threshold. Increasing the threshold for the change in partial discharge intensity is intended to confirm that the partial discharge has indeed experienced a substantial and rapid increase, rather than being solely due to minor perturbations in the azimuth angle calculation. The essential characteristic of a sudden change in local defects is a sharp increase in partial discharge intensity; therefore, this increase must be directly detected for confirmation. Partial discharge change threshold. The setting can be based on the background noise level of the partial discharge intensity index, typically 2 to 3 times the upper limit of the normal fluctuation range. The conditions for exiting the abrupt change state of local defects are also more stringent, requiring the smoothing direction angle to be less than the second exit threshold and the partial discharge intensity index to be less than the partial discharge safety threshold. The purpose of adding a partial discharge safety threshold condition here is to ensure that the partial discharge level has fallen back to a safe range before allowing the sudden change alarm to be lifted, thus avoiding premature cancellation of the warning when the partial discharge is still at a high level. The partial discharge safety threshold can be set according to the transformer's historical partial discharge level or the limit recommended by relevant standards. For example, it can be set to 5 or 10 times the average partial discharge intensity index under the initial healthy state.

[0129] The diagnostic logic described above for single-phase data works effectively in most cases, but in certain special scenarios, there is a risk of misjudgment due to systemic interference. For example, when a power system experiences lightning strikes, switching operations, or significant load fluctuations, the high-voltage side of the transformer may simultaneously experience overvoltage or high-current surges. These surges may simultaneously affect the dielectric loss and partial discharge measurements of all three phases, causing a simultaneous increase in the corrected dielectric loss and partial discharge intensity indices for both phases. If only single-phase data is used for judgment in this situation, such systemic interference may be misjudged as a sudden change in local defects. To eliminate the impact of such systemic interference, this invention further introduces an inter-phase comparison verification mechanism to jointly analyze three-phase (or two-phase) data. (Reference) Figure 6 Specifically, it includes the following steps:

[0130] S6. When two or three phases simultaneously meet the criteria for obvious insulation aging or sudden change in local defects, select any one phase as the reference phase and calculate the abnormal scores of the remaining phases.

[0131] The anomaly score is obtained by weighting the ratios of the corrected dielectric loss index, partial discharge intensity index, state trajectory displacement and orientation angle of the corresponding phase to those of the reference phase.

[0132] The formula for calculating the anomaly score is:

[0133]

[0134] In the formula, , , , The relevant parameters of the current phase line, , , , Relevant parameters for the reference phase; , , , These are the weighting coefficients, and .

[0135] For the reference phase itself, its anomaly score is 1.

[0136] S7. If the abnormal scores of the remaining phases are less than the inter-phase difference threshold, it is determined to be a systematic disturbance, and the current determination status is not changed.

[0137] If the anomaly scores of the other two phases are close to 1 (e.g., between 0.8 and 1.2), it indicates that the state changes of each phase are highly consistent. In this case, it is judged as a systemic disturbance or a global common influence. The current judgment state is not changed, and the system continues to maintain the original diagnostic conclusion or is marked as "three-phase common anomaly" or "suspected external interference". Conversely, if the anomaly score difference between a certain phase and the reference phase is large (e.g., greater than the preset inter-phase difference threshold), If the value is 0.3, it indicates that the state change of the phase has significant specificity, which may be due to the existence of local defects within the phase. In this case, the system should maintain the judgment of the abrupt change of local defects in the phase.

[0138] The phase-to-phase comparison mechanism effectively utilizes the electrical and structural symmetry of the three-phase transformer, maintaining diagnostic robustness under external interference and avoiding unnecessary false alarms caused by abnormal fluctuations in single-phase data.

[0139] As can be seen from the detailed description of the above embodiments, the transformer diagnostic method and system integrating online detection of dielectric loss and partial discharge proposed in this invention effectively solves the technical problems of existing technologies, such as strong reliance on human experience, slow response, and susceptibility to interference and false alarms, by constructing a two-dimensional state space, introducing direction angle discrimination, setting multi-level adaptive thresholds, and combining multi-level interference suppression and inter-phase comparison verification. This method can not only accurately distinguish between two different fault development modes—overall gradual aging and sudden local defect changes—but also maintain the stability and reliability of diagnostic conclusions in complex on-site electromagnetic environments. It provides a practical technical solution for the intelligent online assessment of transformer insulation status and has good prospects for widespread application.

[0140] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A transformer diagnostic method integrating online detection of dielectric loss and partial discharge, characterized in that, Includes the following steps: Obtain dielectric loss parameters and partial discharge parameters within the sliding time window, calculate and correct dielectric loss index based on real-time dielectric loss parameters, and calculate partial discharge intensity index based on real-time partial discharge parameters. A dielectric loss-partial discharge correlation curve is established with the modified dielectric loss index as the horizontal axis and the partial discharge intensity index as the vertical axis. Calculate the orientation angle between the insulation state curve and the horizontal axis within the current time window. When the orientation angle is less than the first threshold, it is determined to be overall slow aging; when the orientation angle is greater than or equal to the first threshold and less than the second threshold, it is determined to be obvious insulation aging; when the orientation angle is greater than or equal to the second threshold, it is determined to be a sudden change in local defects.

2. The method according to claim 1, characterized in that, The calculation and correction of the dielectric loss index based on real-time dielectric loss parameters includes: The real-time dielectric loss factor, top oil temperature, and load current of each phase of the transformer are obtained within the sliding time window. The temperature correction term is determined based on the top oil temperature, and the load correction term is determined based on the load current. The real-time dielectric loss factor is corrected using the temperature correction term and the load correction term to obtain the corrected equivalent dielectric loss value under a unified reference state. The corrected equivalent dielectric loss value within the current sliding time window is compared with the corrected equivalent dielectric loss value under the initial healthy state to obtain the corrected dielectric loss index.

3. The method according to claim 1, characterized in that, The calculation of the partial discharge intensity index based on real-time partial discharge parameters includes: Acquire partial discharge pulse signals within a sliding time window; Extract the maximum discharge amplitude, discharge repetition rate, and phase distribution characteristics from the partial discharge pulse signal; The maximum discharge amplitude, discharge repetition rate, and phase distribution characteristics are normalized respectively. The normalized maximum discharge amplitude, discharge repetition rate, and phase distribution characteristics are weighted according to preset weights to obtain a dimensionless partial discharge intensity index. The partial discharge intensity index is used to characterize the activity and danger of partial discharge within the sliding time window.

4. The method according to claim 3, characterized in that, The phase distribution characteristic quantity is obtained in the following manner: Within the sliding time window, the phase angle of each partial discharge pulse within the power frequency cycle is determined based on the occurrence time of the partial discharge pulse and the power frequency reference phase. A power frequency cycle is divided into multiple phase intervals, and the number of partial discharge pulses and / or the cumulative discharge amplitude in each phase interval are counted to obtain the phase distribution vector. The phase clustering degree, phase distribution entropy, or similarity to a preset partial discharge phase template are calculated based on the phase distribution vector, and the phase clustering degree, phase distribution entropy, or similarity are used as the phase distribution feature quantity.

5. The method according to claim 1, characterized in that, The first threshold is taken as the high quantile of the directional angle distribution during the healthy operation stage, or as the mean of the directional angle plus a preset multiple of the standard deviation, which is used to characterize the boundary of the equipment from the overall slow aging stage to the stage of joint growth of dielectric loss and partial discharge; The second threshold is determined based on the statistical distribution of the directional angle in historical partial discharge anomaly samples or on-site simulated defect samples, and is used to characterize the boundary where the partial discharge intensity grows abnormally rapidly relative to the change in dielectric loss.

6. The method according to claim 1, characterized in that, The method further includes an interference suppression step, which includes: Smoothing or median filtering is applied to the direction angles corresponding to multiple consecutive sliding time windows to obtain smoothed direction angles. The displacement of the state trajectory is calculated based on the modified dielectric loss index and partial discharge intensity index corresponding to multiple consecutive sliding time windows. When the displacement of the state trajectory is less than the preset minimum displacement threshold, the previous diagnostic state remains unchanged. The current diagnostic state is switched to the corresponding diagnostic state only when the smoothing direction angle meets the threshold condition of the corresponding diagnostic state, and the threshold condition is continuously met within a preset number of consecutive sliding time windows.

7. The method according to claim 6, characterized in that, A first entry threshold and a first exit threshold are set for the first threshold, and a second entry threshold and a second exit threshold are set for the second threshold. Wherein, the first entry threshold is greater than the first exit threshold, and the second entry threshold is greater than the second exit threshold; When the smoothing direction angle reaches the first entry threshold and continues to meet the preset confirmation conditions, the system enters the obvious insulation aging state; when the smoothing direction angle is lower than the first exit threshold and continues to meet the preset degradation conditions, the system exits the obvious insulation aging state. When the smoothing direction angle reaches the second entry threshold and continues to meet the preset confirmation conditions, the system enters the local defect mutation state; when the smoothing direction angle is lower than the second exit threshold and continues to meet the preset degradation conditions, the system exits the local defect mutation state.

8. The method according to claim 7, characterized in that, The preset confirmation condition is that the number of sliding time windows with a smooth direction angle greater than the entry threshold is greater than the minimum number of windows. The preset degradation condition is that the number of sliding time windows with a smooth direction angle less than the exit threshold is greater than the minimum number of windows.

9. The method according to claim 7, characterized in that, The interference suppression step further includes: When two or three phases simultaneously meet the criteria for obvious insulation aging or sudden change in local defects, any one phase is selected as the reference phase, and the abnormality scores of the remaining phases are calculated. The anomaly score is obtained by weighting the ratios of the corrected dielectric loss index, partial discharge intensity index, state trajectory displacement and orientation angle of the corresponding phase to those of the reference phase. If the abnormal scores of the remaining phases are less than the inter-phase difference threshold, it is determined to be a systematic disturbance, and the current determination status is not changed.

10. A transformer diagnostic system integrating online detection of dielectric loss and partial discharge, characterized in that, include: The data acquisition module is used to acquire dielectric loss parameters and partial discharge parameters within the sliding time window; The index calculation module is used to calculate the corrected dielectric loss index based on the real-time dielectric loss parameters, and to calculate the partial discharge intensity index based on the real-time partial discharge parameters. The curve establishment module is used to establish an insulation state curve with the modified dielectric loss index as the abscissa and the partial discharge intensity index as the ordinate. The diagnostic judgment module is used to calculate the direction angle between the insulation state curve and the horizontal axis within the current time window, and to determine that the overall slow aging is occurring when the direction angle is less than a first threshold; and to determine that the insulation aging is obvious when the direction angle is greater than or equal to the first threshold and less than a second threshold. When the direction angle is greater than or equal to the second threshold, it is determined to be a local defect mutation.

Images