Insulation deterioration prediction method for high-voltage switch cabinet based on partial discharge characteristics

Abstract

The present application relates to the technical field of high-voltage switch cabinet monitoring, and more particularly to a high-voltage switch cabinet insulation deterioration prediction method based on partial discharge characteristics, which collects a load current effective value sequence, combines a contact resistance to calculate an instantaneous thermal power input, and based on a thermal model and a thermal time constant, calculates an output thermal state index through exponential recursion; discharge pulse events within a power frequency cycle are collected and power frequency voltage phase and apparent discharge amount are recorded. The present application recursively calculates a real thermal state index based on a first-order thermal model, and under the premise of fully considering thermal inertia, performs thermal decoupling to strip thermal disturbance point by point on the measured peak phase; then the purified features are subjected to rate regression, daily accumulation, and life extrapolation and dynamic alarm grading relying on the measured physical failure threshold, effectively solving the problems of heavy load false alarm and light load missed report, and improving the prediction confidence and operation and maintenance decision guidance value of the online monitoring system in the complex power distribution network service environment.

Description

Technical Field

[0001] This invention relates to the field of high-voltage switchgear monitoring technology. More specifically, this invention relates to a method for predicting insulation degradation in high-voltage switchgear based on partial discharge characteristics. Background Technology

[0002] High-voltage switchgear is the core equipment of power distribution networks, and the health of its internal insulation system directly affects the stable operation of the distribution network. In 12kV ring network switchgear, the vacuum circuit breaker arc-extinguishing chamber, the three-position disconnector contacts, the busbars and connecting conductors inside the gas box, and other components together constitute a composite insulation system. Under the combined effects of long-term electric field stress, mechanical fatigue, and localized heat accumulation within the cabinet, the insulation system undergoes progressive degradation, releasing discharge signals in the form of partial discharges at the initial stage of defect formation. Power frequency phase-resolved partial discharge analysis is one of the most mature techniques for assessing insulation condition. This method aligns the phase of the partial discharge pulse with the power frequency voltage waveform, constructing a three-dimensional distribution map of phase, discharge quantity, and discharge frequency. Characteristic parameters such as the peak phase angle are extracted from this map, and the trend over time is used as the basis for judging insulation degradation.

[0003] In actual load operation, the main circuit current continuously flows through each contact node of the high-voltage switchgear. The contact resistance of the main circuit is on the order of microohms, generating Joule heat and conducting it to the surrounding insulating medium. This phenomenon can affect the judgment of insulation defect propagation. Temperature rise test data shows that under overload conditions, the steady-state temperature rise of the contacts and bolted connections can reach tens of Kelvin. This thermal state causes thermal expansion of the gas-solid composite insulation interface inside the gas chamber, resulting in geometric changes in the local electric field distribution and space charge accumulation state at the interface, thereby systematically shifting the discharge peak phase in the PRPD spectrum. This phase shift is a purely load thermodynamic response and is unrelated to insulation defect propagation.

[0004] Existing PRPD analysis methods lack the ability to separate the influence of load thermal state on peak phase from measured data. Under heavy load operation, thermally induced phase drift superimposed on the measured peak phase, a phenomenon that can lead to the model misinterpreting it as accelerated insulation degradation, triggering false alarms. When the load decreases and the temperature drops, the reverse shift of the thermally induced phase may temporarily mask the gradual phase shift of actual early defects, causing the degree of degradation to be underestimated. Both types of misinterpretations result in a severe mismatch between the predicted conclusions and the actual insulation condition, rendering PRPD-based online prediction methods unreliable under varying load conditions. Summary of the Invention

[0005] To address the existing technical problem of partial discharge phase characteristic distortion caused by variable load thermal effect interference, which leads to misjudgment, this application provides a method for predicting insulation degradation of high-voltage switchgear based on partial discharge characteristics.

[0006] This invention provides a method for predicting insulation degradation of high-voltage switchgear based on partial discharge characteristics, comprising: S1: acquiring the effective value sequence of load current, calculating the instantaneous thermal power input by combining contact resistance, and calculating the output thermal state index by exponential recursion based on a thermal model and thermal time constant; S2: acquiring discharge pulse events within the power frequency cycle and recording the power frequency voltage phase and apparent discharge quantity, dividing the phase window to count the cumulative number of discharge pulses, superimposing them to form a time cumulative phase distribution vector, and extracting the positive half-cycle discharge peak phase from the time cumulative phase distribution vector; S3: obtaining the thermal phase mapping coefficient and reference phase determined in advance through thermal phase correlation calibration test, and performing thermal decoupling correction on the positive half-cycle discharge peak phase based on the thermal state index, the thermal phase mapping coefficient, and the reference phase to obtain the thermal decoupling peak phase; S4: performing linear regression calculation on the thermal decoupling peak phase within a fixed analysis period to calculate the phase drift rate, constructing a cumulative drift quantity based on the phase drift rate, and outputting an insulation degradation warning when the extrapolation trend line of the cumulative drift quantity reaches the preset maximum phase offset upper limit.

[0007] By continuously collecting the effective value sequence of load current and calculating the instantaneous thermal power input in conjunction with contact resistance, and then relying on the thermal model and thermal time constant to output an exponentially updated thermal state index, the real-time temperature rise state of the insulation interface is quantitatively characterized. Furthermore, using pre-calibrated thermal phase mapping coefficients and a reference phase, point-by-point thermal decoupling correction is performed on the measured positive half-cycle discharge peak phase, eliminating the phase shift component, which is purely a load thermodynamic response, from the characteristic quantities. This ensures that the remaining thermally decoupled peak phase only reflects the objective physical expansion of insulation defects. Based on this, linear regression within a fixed analysis period is used to obtain the phase drift rate, which is unaffected by load fluctuations. The accumulated drift is then constructed, and an early warning is output when its extrapolated trend reaches the maximum phase shift limit. This significantly reduces the false alarm and missed alarm rates under varying load conditions in practical engineering, restoring the reliability of online insulation degradation prediction in real-world operating scenarios such as heavy loads and overloads. This provides a credible early decision-making basis for condition-based maintenance of core equipment in the distribution network.

[0008] Furthermore, the thermal state index is calculated and output through exponential recursion, including: calculating the steady-state temperature rise based on the instantaneous thermal power input and the pre-calibrated thermal steady-state coefficient; performing exponential weighted fusion of the steady-state temperature rise and the thermal state index at the previous moment to finally obtain the thermal state index and obtain the thermal state index sequence.

[0009] By exponentially weighting and fusing the steady-state temperature rise determined based on instantaneous thermal power with the thermal state index of the previous moment, the system can realistically reproduce the physical process of thermal inertia and temperature rise delay of the insulating medium under step load or drastic load switching scenarios. Compared with the existing simplified approach that only uses instantaneous current to approximate the thermal state, this method effectively avoids phase correction errors caused by the thermal assessment results being ahead or behind, and ensures that the temperature rise data on which the thermal decoupling operation is based is always close to the actual operating conditions of the equipment.

[0010] Furthermore, the superposition of the time-accumulated phase distribution vector includes: dividing the entire power frequency cycle into multiple phase windows, counting the cumulative number of discharge pulses and the average discharge amount in each phase window to form a phase-resolved discharge vector for a single power frequency cycle; continuously acquiring phase-resolved discharge vectors from multiple power frequency cycles and superimposing them in chronological order to form a phase-resolved discharge matrix; summing the phase-resolved discharge matrix along the time axis to finally form the time-accumulated phase distribution vector.

[0011] By dividing the entire power frequency cycle into phase windows, continuously acquiring phase-resolved discharge vectors from multiple cycles, and performing matrix summation and superposition along the time axis, the sporadic high-frequency electromagnetic noise pulses within a single power frequency cycle are strongly suppressed in multi-cycle averaging statistics. Compared with single-cycle feature extraction techniques, this method significantly highlights the phase concentration trend of real discharge activities, endows the peak phase with extremely strong anti-interference capability from the data source, and ensures extremely high stability of feature extraction in complex and strong electromagnetic field environments.

[0012] Further, the step of extracting the positive half-cycle discharge peak phase from the time-cumulative phase distribution vector includes: extracting the center angle of the phase window corresponding to the maximum value of the positive half-cycle from the time-cumulative phase distribution vector as the positive half-cycle discharge peak phase; simultaneously extracting the negative half-cycle discharge peak phase and calculating the asymmetry coefficient of the number of positive and negative half-cycle discharges; and in response to the deviation of the asymmetry coefficient of the number of positive and negative half-cycle discharges from a preset value, simultaneously performing the thermal decoupling correction on the negative half-cycle discharge peak phase.

[0013] Furthermore, the acquisition of the thermal phase mapping coefficient and reference phase determined in advance through thermal phase correlation calibration test includes: before the equipment is put into operation, applying steady-state conditions of no-load, rated current and over-rated current respectively, and measuring the peak phase of positive half-cycle discharge under each steady-state condition; performing linear regression with the thermal state index in the thermal state index sequence as the abscissa and the peak phase of positive half-cycle discharge as the ordinate, and finally obtaining the thermal phase mapping coefficient and the reference phase.

[0014] By calculating the standard deviation of the phase before and after correction to construct the correction quality ratio, the system gains online closed-loop self-testing capability for the effectiveness of thermal decoupling correction. When it is found that the contact resistance has increased due to frequent switching operations of the equipment during long-term service, which in turn causes the original thermal phase mapping relationship to become inaccurate, the recalibration process can be automatically triggered. Compared with the correction strategy of using static parameters throughout the entire life cycle, this method gives the algorithm the ability to adaptively correct deviations and self-evolve across the entire life cycle of the equipment.

[0015] Furthermore, after obtaining the thermal decoupling peak phase, the process further includes: calculating the standard deviation of the positive half-cycle discharge peak phase before correction and the standard deviation of the thermal decoupling peak phase after correction; constructing a corrected mass ratio based on the standard deviation of the corrected thermal decoupling peak phase and the standard deviation of the positive half-cycle discharge peak phase before correction; and triggering a recalibration process to update the thermal phase mapping coefficient in response to the corrected mass ratio being greater than a determination threshold.

[0016] By performing linear regression on the thermally decoupled peak phase sequence after removing thermal disturbances within a fixed analysis period window to calculate the phase drift rate, the discrete and minute daily phase shifts are transformed into a smooth, continuous, and physically oriented defect propagation rate index. Compared with the traditional method of directly relying on single-point jump values ​​to assess degradation, this method provides an extremely stable and resistant quantitative analysis foundation for subsequent degradation trend judgment and cumulative quantity construction.

[0017] Further, the phase drift rate is calculated by: using a fixed analysis period as a window, performing linear regression on the thermally decoupled peak phase sequence of all time steps within the window to calculate the phase drift rate; and arranging the phase drift rates of each analysis window in chronological order to finally form a drift rate time series.

[0018] By extracting historical cumulative values ​​and multiplying the current period's smooth drift rate by the cumulative values, the net phase shift caused by the objective evolution of insulation defects can be recorded continuously in a way that is completely unaffected by load fluctuations. Compared with the monitoring mechanism that performs tomographic calculations independently each time, the monotonic growth curve of the cumulative drift value presented by this method is 100% driven by the defect expansion process, providing a strongly physically interpretable evolutionary trajectory to support the extrapolation of the remaining safe lifetime.

[0019] Furthermore, the step of constructing the cumulative drift amount based on the phase drift rate includes: extracting the historical cumulative drift amount of the previous analysis day; and adding the product of the phase drift rate of the current analysis day and the fixed analysis period to the historical cumulative drift amount to finally obtain the cumulative drift amount.

[0020] Furthermore, the insulation degradation warning is output, including: selecting a control sample with pre-fabricated typical defects in the thermal phase correlation calibration test, tracking the evolution of the thermal decoupling peak phase until the insulation performance reaches the critical failure state; extracting the total offset throughout the process as the upper limit of the maximum phase offset; and outputting the insulation degradation warning and giving the predicted remaining safe operating days in response to the prediction of the extrapolation trend line of the cumulative drift amount reaching the upper limit of the maximum phase offset.

[0021] Furthermore, after the output insulation degradation warning, the method further includes: monitoring the phase drift rate of multiple consecutive analysis windows in the recent period; and synchronously upgrading the alarm level in response to the phase drift rate of multiple consecutive analysis windows showing a monotonically increasing trend.

[0022] By continuously monitoring the monotonically increasing drift rate of multiple subsequent windows to dynamically upgrade the alarm level, the monitoring system can immediately sound a higher-level alarm when it keenly detects that the internal insulation defect has suddenly changed from a stable expansion to an extremely high-risk critical state of accelerated deterioration. Compared with a single static alarm strategy, this method achieves a precise match between the on-site hazard level and the allocation of manual review resources, effectively preventing sudden insulation breakdown accidents.

[0023] Beneficial effects: Based on the first-order thermal model, the true thermal state index is recursively derived. Under the premise of fully considering thermal inertia, thermal decoupling is performed point by point on the measured peak phase to remove thermal disturbances. Then, the purified features are subjected to rate regression and daily accumulation. Based on the measured physical failure threshold of the system, lifetime extrapolation and dynamic alarm classification are completed. This effectively solves the problems of false alarms under heavy load and missed alarms under light load, and improves the prediction confidence and operation and maintenance decision guidance value of the online monitoring system in the service environment of complex power distribution networks. Attached Figure Description

[0024] Figure 1 This is a flowchart illustrating an insulation degradation prediction method according to an embodiment of the present invention;

[0025] Figure 2 This is a schematic diagram illustrating the trend of cumulative phase drift due to insulation degradation and a comparison of the upper limit warning of maximum phase shift according to an embodiment of the present invention. Detailed Implementation

[0026] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0027] This invention discloses a method for predicting insulation degradation of high-voltage switchgear based on partial discharge characteristics, referring to... Figure 1 This includes steps S1-S4: S1: Collect the load current and calculate the output thermal state index through exponential recursion.

[0028] Specifically, with a fixed time step At time intervals, the effective value sequence of the three-phase load current is continuously acquired through current transformers installed in the main circuit. , This allows for the acquisition of complete current timing data during the online operation of the switchgear. This is combined with the factory-calibrated values ​​of the main circuit contact resistance. The instantaneous thermal power input at each sampling time is calculated according to the following formula: Contact resistance At the microohm level, this Joule heat is continuously conducted to the surrounding gas-solid composite insulation interface, becoming the direct heat source driving changes in the insulation's thermal state.

[0029] The thermal state of the main circuit insulation interface is equivalent to a first-order lumped parameter thermal model, and the steady-state temperature rise is... The instantaneous thermal power input is related to the following equation: ;in, The thermal steady-state coefficient, expressed in K / A², is calibrated by substituting the equipment's factory type test data; that is, the ratio of the measured steady-state temperature rise under rated operating conditions to the square of the corresponding current. The value of this value is directly linked to the equipment model, rated current, and factory measured value of contact resistance.

[0030] The dynamic response of the thermal state is expressed by the thermal time constant. Description, steady-state temperature rise thermal state index compared to the previous moment Perform index-weighted fusion and calculate the output thermal state index according to the following formula. : The above The parameter is determined by a specialized thermal calibration test before equipment commissioning. This involves applying a step load at rated current, recording the temperature rise curve of the insulation interface, and fitting the time constant. This parameter is directly related to the equipment model, gas chamber size, and heat dissipation conditions. There are many other ways to describe the thermal inertia process in this field. Those skilled in the art can choose a higher-order thermal network model according to actual needs. The above-described first-order lumped parameter form is only a typical implementation method with strong engineering practicality.

[0031] Through the above exponential recursive calculation, a complete thermal state index sequence is output. The sequence is updated synchronously within each sampling step, objectively reflecting the real-time temperature rise of the insulation interface under variable load conditions, and providing continuous thermal state input for subsequent quantitative correction of thermally induced phase shift.

[0032] Thus, by continuously collecting load current and exponentially recursively applying the first-order thermal model, real-time tracking of the thermal state of the insulation interface is achieved, providing a quantitative basis for subsequent thermal decoupling correction that is updated every moment.

[0033] S2: Superimpose to form the time cumulative phase distribution vector and extract the peak phase of the positive half-cycle discharge.

[0034] Discharge pulse events within each power frequency cycle are acquired using an online partial discharge sensor, and the power frequency voltage phase corresponding to each pulse is recorded synchronously. and apparent discharge quantity Divide the entire power frequency cycle into equal parts. There are 1 phase window, with a window width of 1. , The specific number is determined based on the phase resolution requirements; for example, We can choose 36, which corresponds to a 10° resolution, and count the cumulative number of discharge pulses within each phase window. and average discharge quantity, This forms a phase-resolved discharge vector for a single power frequency cycle. .

[0035] Continuous data acquisition Phase-resolved discharge vector for each power frequency cycle The specific quantity is determined based on the acquisition time window and the power frequency. For example, under a power frequency of 50Hz and an acquisition time window of 20s, We can take 1000, and then superimpose the phase-resolved discharge vectors of each cycle in chronological order to form a dimension of... Phase-resolved discharge matrix Matrix elements Indicates the first Within the first cycle The number of discharges within each phase window is calculated. The phase-resolved discharge matrix is ​​summed along the time axis to form the time-cumulative phase distribution vector. ,in: ;because Completed in the timeline direction The statistical superposition of each cycle effectively averages out the occasional noise pulses within a single cycle, highlighting the phase concentration trend of the discharge activity. The extracted peak phase has a strong ability to resist occasional interference.

[0036] From the time-accumulated phase distribution vector Extracting the peak phase of the positive half-cycle discharge: taking (corresponding to 0° to 180°) The center angle of the phase window corresponding to the maximum value is propagated backward as the peak phase of the positive half-cycle discharge.

[0037] Synchronously extract the peak phase of the negative half-cycle discharge and take Inside The maximum value corresponds to the phase angle, and the asymmetry coefficient of the positive and negative half-cycle discharge times is calculated according to the following formula. : ; This is the ratio of the number of positive and negative half-cycle pulses. The closer this value is to the preset value, the higher the symmetry of the positive and negative half-cycle discharge activities during internal air gap discharge; the greater the deviation of this value from the preset value, the more obvious the phase bias of the positive and negative half-cycle activities during surface discharge. In response to the deviation of the asymmetry coefficient of the number of positive and negative half-cycle discharges from the preset value, the same thermal decoupling correction as in subsequent steps is performed synchronously on the peak phase of the negative half-cycle discharge.

[0038] Thus, by superimposing the phase-resolved discharge vectors of multiple power frequency cycles on the time axis, the interference of occasional noise on the peak phase extraction is effectively suppressed at the statistical level, ensuring the stability of the peak phase of the positive half-cycle discharge as the subsequent correction input.

[0039] S3: Perform thermal decoupling correction on the peak phase of the positive half-cycle discharge based on the thermal phase mapping coefficient and the reference phase.

[0040] Before commissioning the equipment, a thermal phase correlation calibration test was performed to obtain the thermal phase mapping coefficient and reference phase. The equipment was subjected to three steady-state conditions—no-load, rated current, and over-rated current—at 1.2 times the rated voltage at the power frequency. After the thermal state of each steady-state condition stabilized, the peak phase of the positive half-cycle discharge was measured. A linear regression was performed with the thermal state index on the x-axis and the peak phase of the positive half-cycle discharge on the y-axis to obtain the thermal phase mapping coefficient. and reference phase . Physically, it reflects the peak phase shift caused by the change in space charge distribution at the gas-solid interface for every 1K increase in temperature. The value is determined by the actual measurement calibration data of the equipment and is directly bound to the specific insulating material and geometry.

[0041] During the online operation phase, thermal decoupling correction is performed on the positive half-cycle discharge peak phase extracted for each sampling time window, and the thermally decoupled peak phase is obtained according to the following relationship: ;in, This is the corrected thermal decoupling peak phase. This represents the peak phase of the positive half-cycle discharge extracted during the current sampling window. The current thermal state index has been calculated. To calibrate the baseline value of the thermal state index corresponding to the no-load condition in the test, the above correction operation is used to remove the thermally induced phase shift component point by point from the measured positive half-cycle discharge peak phase, thus... Only information related to the physical propagation of insulation defects is retained.

[0042] Furthermore, an online self-check is performed to assess the correction effect. Within the same runtime period, the standard deviation of the positive half-cycle discharge peak phase sequence before correction is calculated. and the standard deviation of the corrected thermal decoupling peak phase sequence. Construct the corrected mass ratio according to the following formula. : ; The smaller the value, the more effective the correction. A larger value indicates that the dispersion of the corrected sequence has not been effectively reduced compared to the uncorrected sequence, and the thermal phase mapping coefficient may have shifted. The judgment threshold is determined during the thermal phase correlation calibration test, i.e., by applying the same correction to the measured positive half-cycle discharge peak phase sequence under each steady-state condition, and statistically analyzing the results under each condition. The mean value is calculated after value distribution and used as an objective upper bound for online operation. In response to a correction mass ratio exceeding a threshold, a recalibration process is triggered to update the thermal phase mapping coefficients. This is to address the long-term shift in thermal phase mapping caused by the increase in equipment contact resistance with the number of operations.

[0043] Thus, through the coordinated operation of point-by-point thermal decoupling correction and online self-checking of correction quality, the pure extraction of insulation defect extension information from the thermal decoupling peak phase is ensured, and the system is endowed with the ability to adaptively track long-term drift of thermal phase mapping relationship.

[0044] S4: Calculate the phase drift rate and construct the cumulative drift amount to output an early warning of insulation degradation.

[0045] With a fixed analysis cycle As a window, the thermally decoupled peak phase of all time steps within the window. Perform linear regression on the sequence to calculate the phase drift rate. . This reflects the daily net shift of the peak phase caused solely by insulation defect propagation after excluding thermal disturbances. A larger value indicates a faster insulation defect propagation rate, while a smaller value indicates a more stable defect propagation process. The phase drift rates from each analysis window are arranged chronologically to form a drift rate time series. ,in For the first Analysis day.

[0046] Extract the historical cumulative drift from the previous analysis day, and add the product of the phase drift rate of the current analysis day and the fixed analysis period to the historical cumulative drift. The cumulative drift is then obtained according to the following formula. : ;in, Reflecting from the start time of monitoring to the [number]th The net accumulated phase offset caused by the propagation of insulation defects. It is after thermal decoupling has been completed. It is accumulated daily on a basic basis, and the growth curve is not affected by load fluctuations, but is driven only by the objective expansion process of defects, making it possible to... Trend extrapolation has clear physical interpretability.

[0047] Maximum phase offset limit In the thermal phase correlation calibration test, it was determined that: a control sample with prefabricated typical defects was selected, and the thermal decoupling peak phase was tracked from the initial defect-free state. The process evolves until the insulation performance reaches a critical failure state, and the total offset throughout the entire process is extracted as... . The measured data on defect evolution derived from the equipment itself are directly linked to the equipment's insulation materials and geometry.

[0048] Response to cumulative drift The extrapolated trendline prediction has reached the maximum phase offset limit. It outputs an early warning of insulation degradation and provides a predicted number of remaining safe operating days. Following the output insulation degradation warning, the phase drift rate is continuously monitored across multiple recent analysis windows. In response to the recent monotonically increasing phase drift rate across multiple analysis windows, indicating that degradation is accelerating, the alarm level is being upgraded to prompt maintenance personnel to schedule on-site reviews at shorter intervals.

[0049] In this way, by accumulating the drift rate of the thermal decoupling peak phase daily and extrapolating the trend, the abstract phase drift process is transformed into a remaining lifetime prediction with clear physical upper limit constraints, enabling maintenance personnel to obtain sufficient early warning time before defects reach the critical failure state.

[0050] like Figure 2 As shown, it can be observed that the existing method is driven by load thermal disturbance, the cumulative drift increases too quickly and approaches the upper limit earlier, and there is a risk of false warnings. The cumulative drift curve of the method of the present invention increases steadily and is only driven by defect expansion. The intersection of the extrapolated trend line and the upper limit of the maximum phase offset corresponds to the predicted remaining number of safe operating days, so that operation and maintenance personnel can obtain a sufficient warning time window in advance.

Claims

1. A method for predicting insulation degradation of high-voltage switchgear based on partial discharge characteristics, characterized in that, include: S1: Collect the effective value sequence of load current, calculate the instantaneous thermal power input by combining the contact resistance, and calculate the output thermal state index by exponential recursion based on the thermal model and thermal time constant. S2: Collect discharge pulse events within the power frequency cycle and record the power frequency voltage phase and apparent discharge quantity. Divide the phase window to count the cumulative number of discharge pulses, and superimpose them to form a time-accumulated phase distribution vector. Extract the positive half-cycle discharge peak phase from the time-accumulated phase distribution vector. S3: Obtain the thermal phase mapping coefficient and reference phase determined in advance through thermal phase correlation calibration test. Based on the thermal state index, the thermal phase mapping coefficient and the reference phase, perform thermal decoupling correction on the positive half-cycle discharge peak phase to obtain the thermal decoupling peak phase and form a thermal decoupling peak phase sequence. S4: Within a fixed analysis period, perform linear regression to calculate the phase drift rate of the thermally decoupled peak phase, construct a cumulative drift amount based on the phase drift rate, and output an insulation degradation warning when the extrapolation trend line of the cumulative drift amount reaches the preset maximum phase offset limit.

2. The method for predicting insulation degradation of high-voltage switchgear based on partial discharge characteristics according to claim 1, characterized in that, The output thermal state index is calculated through exponential recursion, including: The steady-state temperature rise is calculated based on the instantaneous thermal power input and the pre-calibrated thermal steady-state coefficient. The steady-state temperature rise is fused with the thermal state index of the previous moment using an exponential weighting method to obtain the thermal state index and a thermal state index sequence.

3. The method for predicting insulation degradation of high-voltage switchgear based on partial discharge characteristics according to claim 1, characterized in that, The superposition constitutes the time-cumulative phase distribution vector, including: The entire power frequency cycle is divided into multiple phase windows. The cumulative number of discharge pulses and the average discharge amount in each phase window are counted to form a phase-resolved discharge vector for a single power frequency cycle. Phase-resolved discharge vectors from multiple power frequency cycles are continuously acquired and superimposed in chronological order to form a phase-resolved discharge matrix. The phase-resolved discharge matrix is ​​summed along the time axis to form the time-cumulative phase distribution vector.

4. The method for predicting insulation degradation of high-voltage switchgear based on partial discharge characteristics according to claim 1, characterized in that, Extracting the positive half-cycle discharge peak phase from the time-accumulated phase distribution vector includes: The center angle of the phase window corresponding to the maximum value of the positive half-cycle is extracted from the time-accumulated phase distribution vector and used as the peak phase of the positive half-cycle discharge. The peak phase of the negative half-cycle discharge is extracted simultaneously, and the asymmetry coefficient of the number of discharges in the positive and negative half-cycles is calculated. In response to the deviation of the asymmetry coefficient of the positive and negative half-cycle discharge times from the preset value, the thermal decoupling correction is synchronously performed on the peak phase of the negative half-cycle discharge.

5. The method for predicting insulation degradation of high-voltage switchgear based on partial discharge characteristics according to claim 2, characterized in that, The process of obtaining the thermal phase mapping coefficients and reference phase, which were determined in advance through thermal phase correlation calibration experiments, includes: Before the equipment is put into operation, steady-state conditions of no-load, rated current and over-rated current are applied respectively, and the peak phase of positive half-cycle discharge under each steady-state condition is measured. Linear regression is performed using the thermal state index in the thermal state index sequence as the horizontal axis and the peak phase of the positive half-cycle discharge as the vertical axis to finally obtain the thermal phase mapping coefficient and the reference phase.

6. The method for predicting insulation degradation of high-voltage switchgear based on partial discharge characteristics according to claim 1, characterized in that, After obtaining the peak phase of thermal decoupling, the following steps are also included: Calculate the standard deviation of the peak phase of the positive half-cycle discharge before correction, and the standard deviation of the peak phase of thermal decoupling after correction; A corrected mass ratio is constructed based on the standard deviation of the corrected thermal decoupling peak phase and the standard deviation of the positive half-cycle discharge peak phase before correction. In response to the correction mass ratio being greater than the determination threshold, a recalibration process is triggered to update the thermal phase mapping coefficient.

7. The method for predicting insulation degradation of high-voltage switchgear based on partial discharge characteristics according to claim 1, characterized in that, Calculating the phase drift rate includes: Using a fixed analysis period as a window, linear regression is performed on the thermally decoupled peak phase sequence of all time steps within the window to calculate the phase drift rate; The phase drift rates from each analysis window are arranged in chronological order to form a drift rate time series.

8. The method for predicting insulation degradation of high-voltage switchgear based on partial discharge characteristics according to claim 1, characterized in that, The construction of the cumulative drift amount based on the phase drift rate includes: Extract the historical cumulative drift from the previous analysis day; The product of the phase drift rate of the current analysis day and the fixed analysis period is added to the historical cumulative drift amount to finally obtain the cumulative drift amount.

9. The method for predicting insulation degradation of high-voltage switchgear based on partial discharge characteristics according to claim 1, characterized in that, Output insulation degradation warning, including: In the thermal phase correlation calibration test, a control sample with a prefabricated typical defect was selected, and the evolution of the thermal decoupling peak phase was tracked until the insulation performance reached the critical failure state. Extract the total offset throughout the entire process as the upper limit of the maximum phase offset; In response to the extrapolation trend line of the cumulative drift predicting that the maximum phase offset limit has been reached, the insulation degradation warning is output and the predicted remaining number of safe operating days is given.

10. The method for predicting insulation degradation of high-voltage switchgear based on partial discharge characteristics according to claim 1, characterized in that, Following the output insulation degradation warning, the system also includes: Monitor the phase drift rate across multiple consecutive analysis windows in the recent period; In response to the recent monotonically increasing phase drift rate across multiple consecutive analysis windows, the alarm level is upgraded accordingly.

Images