A voltage transformer overvoltage identification and current-limiting harmonic elimination control method

Abstract

This application relates to the field of electrical equipment operation control technology and discloses a method for overvoltage identification and current limiting harmonic elimination control of voltage transformers. The invention first acquires multi-source response data within a preset analysis period corresponding to the abnormal triggering time, performs tagging processing on topology state information and thermal state information to generate context labels, and performs standardized characterization and magnetic state proxy conversion on voltage and current responses to generate pre-control standardized observation results, pre-control magnetic state risk quantities and voltage anomaly characterization. Then, it implements restricted diagnostic control actions, and combines the differences in multi-source responses before and after the control actions with the network real anomaly template responses and measurement distortion template responses in the template library to generate observation confidence results and controlled object risk state results, so as to determine the formal control path label and control level.

Description

Technical Field

[0001] This application relates to the field of electrical equipment operation control technology, and in particular to a method for overvoltage identification and current limiting harmonic elimination control of voltage transformers. Background Technology

[0002] In medium-voltage power distribution networks, power supply and distribution systems in factories and mines, power supply units in rail transit stations, and continuous production scenarios such as chemical and metallurgical industries, electromagnetic voltage transformers typically undertake multiple functions, including bus voltage sampling, insulation monitoring, protection criterion supply, automated interlocking, and abnormal alarms. They are not only important nodes for operational status perception but also crucial data sources for control decision-making. Especially in neutral-point ungrounded systems, systems grounded via arc suppression coils, and networks with a high proportion of cables, frequent switching of reactive power compensation devices, and continuous operation of single-phase grounding line selection, factors such as light line load, changes in switch opening and closing timing, the participation of parallel capacitors, and asymmetry in three-phase parameters can easily induce complex phenomena such as overvoltage, resonant amplification, and waveform distortion. In existing technologies... To suppress the ferroresonance of voltage transformers and the resulting overvoltages, measures such as primary harmonic suppressors, secondary harmonic suppressors, fixed damping resistors, current-limiting branches, or changing system parameters are commonly used. Some solutions employ current-limiting harmonic suppressors based on PTC materials, utilizing the device's low resistance under normal conditions, high resistance transition during abnormal conditions, and self-recovery dynamic impedance characteristics after the abnormality is resolved to suppress the resonant energy. The aforementioned devices and measures can, to some extent, cover the general overvoltage alarm and harmonic suppression requirements. However, their application focuses mostly on the resonance suppression device itself, branch switching logic, or threshold triggering rules. There is a lack of further collaborative processing mechanisms for whether the observation chain is reliable under complex operating conditions and how to determine the recovery state after control.

[0003] In actual operation, voltage transformers are both the protected object and the main measurement source. When primary-side magnetic saturation, core residual magnetism accumulation, temperature rise changes, and secondary-side load disturbances are transmitted to the control system along the measurement chain, deviations can easily occur between the monitoring results and the actual state of the equipment. For example, after switching to a backup feeder during light-load periods at night, some substations may experience abnormal increases in both open delta voltage and phase voltage. However, such anomalies may originate from actual network overvoltages or measurement distortions caused by sudden changes in the voltage transformer's excitation state. Existing solutions typically separate anomaly detection from harmonic suppression control. The former often identifies abnormal states based on the effective value of phase voltage, open delta voltage, zero-sequence voltage, harmonic components, or spectral characteristics, while the latter uses preset thresholds to activate damping branches, switch current-limiting branches, lock out relevant devices, or delay exiting control. This type of scheme assumes that the measured voltage can accurately reflect the network state, but lacks independent verification of the reliability of the observation results. Therefore, when the voltage transformer itself has entered the nonlinear excitation region and the secondary output has been affected by magnetic saturation, it is easy to misjudge the measurement distortion as a system anomaly and further implement control on the erroneous object. More importantly, some control actions will prioritize improving the secondary side sampling waveform to make the monitoring end appear to be in a recovered state, but the primary side magnetic state, heat accumulation level and residual risk are not eliminated at the same time. This leads to the control system prematurely withdrawing the harmonic elimination measures, resulting in repeated oscillation, frequent switching and control oscillation. Therefore, how to simultaneously determine the reliability of the observation results during the voltage transformer overvoltage identification process, and further distinguish between observation recovery and equipment state recovery after formal control, has become a technical problem that urgently needs to be solved in the field of voltage transformer current limiting and harmonic elimination control. Summary of the Invention

[0004] This application proposes a method for overvoltage identification and current limiting harmonic elimination control of voltage transformers to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, this application adopts the following technical solution: a method for overvoltage identification and current limiting harmonic suppression control of a voltage transformer, comprising the following steps:

[0006] S1. Obtain multi-source response data within the preset analysis period corresponding to the abnormal triggering time, perform labeling processing on the topological state information and thermal state information, generate context labels, and perform standardized characterization and magnetic state proxy conversion on the voltage and current response to generate pre-control standardized observation results, pre-control magnetic state risk quantity and voltage anomaly characterization.

[0007] S2. When the risk quantity of the pre-controlled magnetic state is within the diagnostic allowable range, implement restricted diagnostic control actions. Based on the differences in multi-source responses before and after the restricted diagnostic control actions, and combined with the network real abnormal template responses and measurement distortion template responses in the template library, generate observation confidence results and controlled object risk state results to determine the formal control path label and control level.

[0008] S3. Based on the formal control path label and control level, perform formal hierarchical current limiting and harmonic elimination control, and extract recovery characteristic quantities around the electrical recovery process and thermal state recovery process after control to generate observational reliable recovery results, magnetic state fall-off results, thermal state fall-off results and recurrence probability results;

[0009] S4. Perform joint constraint judgment on the observation reliable recovery results, magnetic state fall-off results, thermal state fall-off results and recurrence probability results, generate control withdrawal permission results, and perform the following actions based on the control withdrawal permission results: remove the control action of the damping branch and the current limiting branch, adjust the conduction strength of the damping branch or the control level of the current limiting branch, maintain the current control level or adjust the current control level to the corresponding upgraded control level, and update the template library at the same time.

[0010] Furthermore, the multi-source response data includes the secondary side three-phase phase voltage, open delta voltage, auxiliary winding current, secondary circuit port voltage, secondary circuit current, primary side switch status, housing temperature, and ambient temperature within the preset analysis period corresponding to the abnormal triggering time. Among them, the secondary side three-phase phase voltage, open delta voltage, auxiliary winding current, secondary circuit port voltage, and secondary circuit current constitute voltage and current response, the primary side switch status constitutes topology state information, and the housing temperature and ambient temperature constitute thermal state information.

[0011] Furthermore, the topology status information and thermal status information are tagged, including: generating topology status tags based on the primary-side switch state switching relationship and steady-state conduction relationship within the preset analysis period corresponding to the abnormal trigger time;

[0012] Low-pass filtering is performed based on the temperature difference sequence between the shell temperature and the ambient temperature to obtain a smooth temperature rise. The smooth temperature rise is then converted into a ratio with the reference temperature rise, and a thermal status label is generated according to the preset temperature rise range. The topology status label and the thermal status label are combined to generate a context label.

[0013] Furthermore, the voltage and current responses are standardized and the magnetic state proxy is converted, including: performing frequency band energy extraction on the open delta voltage, performing effective value ratio conversion on the open delta voltage and the secondary three-phase phase voltage, performing phase deviation comparison on the secondary three-phase phase voltage, performing slope normalization on the auxiliary winding current, performing harmonic distortion calculation on the voltage and current responses, and performing conditional standardization processing in combination with the health reference statistical parameters corresponding to the context label to generate pre-control standardized observation results and voltage anomaly characterization.

[0014] Based on the secondary circuit port voltage, secondary circuit current, and calibrated equivalent parameters of the secondary winding, the magnetization branch voltage is reconstructed and the normalized flux linkage proxy is calculated. The pre-controlled magnetic state risk quantity is generated based on the peak value of the normalized flux linkage proxy.

[0015] Furthermore, implement restricted diagnostic control actions, including: when the pre-controlled magnetic state risk quantity is lower than the magnetic state emergency boundary and the voltage anomaly characterization is lower than the voltage emergency boundary, connect the diagnostic damping branch or the diagnostic current limiting branch, and limit the duration of the restricted diagnostic control action to between one power frequency cycle and three power frequency cycles, and limit the peak current of the diagnostic branch to within the preset safety upper limit.

[0016] When the pre-controlled magnetic state risk quantity reaches the magnetic state emergency boundary or the voltage anomaly characterization reaches the voltage emergency boundary, the restricted diagnostic control action is skipped and the formal graded current limiting and harmonic elimination control is directly entered.

[0017] Furthermore, the observation confidence results are generated, including: extracting standardized post-diagnosis observation results around the response period following the restricted diagnostic control action;

[0018] The standardized observation results after diagnosis are differentiated from the standardized observation results before control to generate multi-source response differences. Based on context labels and the duration of restricted diagnostic control actions, the network real anomaly template response and measurement distortion template response are called from the template library.

[0019] The deviation of the multi-source response difference from the network’s true anomalous template response and the measurement distortion template response is calculated respectively, and the observation confidence result is generated based on the comparison results of the two types of deviation.

[0020] Furthermore, the risk status results of the controlled object are generated and the formal control path label and control level are determined, including: generating magnetic state risk components based on the difference results of the normalized magnetic flux proxy quantity in the diagnostic stage; generating thermal state risk components based on the difference results of the smoothed temperature rise quantity in the diagnostic stage.

[0021] Based on the voltage anomaly characterization in the diagnostic phase, a voltage stress risk component is generated. Based on the magnetic state risk component, thermal state risk component, and voltage stress risk component, the risk status result of the controlled object is generated. Based on the observation reliability result, the formal control path label is determined, and the control level is determined based on the risk status result of the controlled object.

[0022] Furthermore, formal hierarchical current limiting and harmonic elimination control is performed based on the formal control path label and control level, including: when the formal control path label corresponds to a real abnormal network path, the damping branch is connected first, and then the control level of the current limiting branch is adjusted.

[0023] When the formal control path label corresponds to the measurement distortion path, first adjust the control level of the current limiting branch, then connect the damping branch, and call the corresponding branch control parameters according to the control level to implement formal graded current limiting and harmonic elimination control. After the formal graded current limiting and harmonic elimination control, establish electrical recovery window and thermal confirmation window respectively to extract recovery characteristic quantities.

[0024] Furthermore, the system generates reliable recovery results for observations, magnetic state fallback results, thermal state fallback results, and recurrence probability results, including: within the electrical recovery window, comparing the deviation of the standardized observation results after formal graded current limiting and harmonic elimination control from the health reference statistical parameters with the deviation before formal graded current limiting and harmonic elimination control, and generating reliable recovery results for observations based on the deviation difference results.

[0025] Based on the difference in normalized flux proxy quantity before and after formal graded current limiting and harmonic elimination control, as well as the difference in the odd-order component ratio of the auxiliary winding, magnetic state fallback results are generated, and residual magnetic state risk quantity is generated.

[0026] Within the thermal confirmation window, the thermal state fallback result is generated based on the difference in smoothed temperature rise before and after formal graded current limiting and harmonic elimination control, and the residual thermal state risk quantity is generated. The recurrence probability result is generated based on the normalized flux proxy variance after formal graded current limiting and harmonic elimination control, the voltage anomaly characterization after formal graded current limiting and harmonic elimination control, and the residual magnetic state risk quantity.

[0027] Furthermore, generating control withdrawal permission results and updating the template library includes: performing joint constraint determination on the observation credible recovery results, magnetic state fallback results, thermal state fallback results and recurrence probability results to generate control withdrawal permission results;

[0028] When the control removal permission result meets the control removal conditions and the residual magnetic state risk and residual thermal state risk respectively meet the corresponding safety boundaries, the control effect of the damping branch and the current limiting branch is removed.

[0029] When the control release permit result meets the conditions for slow withdrawal but does not meet the conditions for control release, adjust the conduction strength of the damping branch or adjust the control level of the current limiting branch according to the preset steps.

[0030] If the withdrawal permission result does not meet the conditions for easing restrictions, the current control level is maintained or the current control level is adjusted to the corresponding upgraded control level. The context label, multi-source response difference, observation credible recovery result, magnetic state fallback result, thermal state fallback result, recurrence probability result and final disposal result corresponding to this disposal are written back to the template library to update the health reference statistical parameters, network real anomaly template response and measurement distortion template response under the corresponding context.

[0031] The beneficial effects of this invention are as follows:

[0032] This invention employs a method of labeling the topological and thermal state information in multi-source response data and forming pre-control standardized observation results, pre-control magnetic state risk quantities, and voltage anomaly characterizations in the corresponding context. This solves the problems of mixed observation baselines under different wiring states and temperature rise states, and inconsistent physical meanings of single anomalies under different operating conditions in the prior art. It plays a role in unifying the state characterization benchmark before anomaly triggering, and also improves the comparability of anomaly characterization and provides stable input for subsequent control judgment.

[0033] This invention employs a method of implementing restricted diagnostic control actions when the pre-controlled magnetic state risk quantity is within the diagnostic allowable range. It combines the differences in multi-source responses before and after the diagnostic action, the real network anomaly template response, and the measurement distortion template response to generate observation confidence results. Then, based on the observation confidence results and the risk state results of the controlled object, the formal control path label and control level are determined. This solves the problems in the prior art where the detection and control links are separated, and the measurement source is still directly used as the control basis after being disturbed, which easily leads to misjudgment and miscontrol. It plays a role in distinguishing between real network anomalies and measurement distortion, and also plays a role in achieving adaptive matching between control path and control intensity.

[0034] This invention employs a method of generating reliable recovery results, magnetic state fallback results, thermal state fallback results, and recurrence probability results after formal graded current limiting and harmonic elimination control, and forming a control withdrawal permission result based on these results. This method further executes control withdrawal, graded withdrawal, maintaining the current control level, or adjusting to the corresponding upgraded control level, while updating the template library. This solves the problems in the prior art where control is withdrawn based solely on apparent waveform recovery, which easily leads to re-oscillation, repeated impacts, and control oscillations. It improves the accuracy of control withdrawal timing, suppresses the risk of recurrence, and enhances the adaptability to subsequent operating conditions. Attached Figure Description

[0035] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort:

[0036] Figure 1 This is a flowchart of the method of the present invention;

[0037] Figure 2 This is a flowchart of the S2 process of the present invention. Detailed Implementation

[0038] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. 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.

[0039] Example

[0040] like Figure 1 and Figure 2 As shown, this invention discloses a method for overvoltage identification and current limiting harmonic suppression control of a voltage transformer, including the following specific steps:

[0041] In this embodiment, the damping branch and the current-limiting branch are used to carry the formal graded current-limiting harmonic elimination control and its subsequent control withdrawal process. It is preferably implemented by a control branch with damping and current-limiting functions, and a composite branch structure containing a PTC core device can be further adopted. For the formal graded current-limiting harmonic elimination control and control withdrawal stage, the PTC core device can use its impedance change characteristics caused by temperature rise to provide support for the damping strength and current-limiting strength during the continuous action stage. For the limited diagnostic control action, the corresponding diagnostic damping branch and diagnostic current-limiting branch are only required to form additional disturbances with limited amplitude, limited duration and repeatable measurement within the diagnostic window to ensure that the differences in multi-source responses before and after the diagnostic action are comparable. Therefore, this invention connects the pre-control standardized observation results, observation reliability results, formal control path labels, control level, observation reliability recovery results and control withdrawal permission results into a unified control chain, so that overvoltage identification, control implementation and control withdrawal can be carried out collaboratively around the same execution carrier.

[0042] In this embodiment, S1 is used to convert the multi-source response data under the corresponding operating condition at the time of abnormal triggering into comparable pre-control standardized observation results before the formal hierarchical current limiting and harmonic elimination control intervention. Simultaneously, it generates the pre-control magnetic state risk quantity and voltage anomaly characterization. After this processing, S2 no longer relies on a single voltage amplitude for judgment, but uses the pre-control standardized observation results, the pre-control magnetic state risk quantity, and the voltage anomaly characterization as joint inputs to carry out restricted diagnostic control actions and generate observation reliability results.

[0043] Specifically, firstly, multi-source response data within a preset analysis period corresponding to the abnormal triggering time is acquired. This multi-source response data includes secondary-side three-phase phase voltage, open delta voltage, auxiliary winding current, secondary circuit port voltage, secondary circuit current, primary-side switch status, casing temperature, and ambient temperature. The preset analysis period is set to 8 to 20 power frequency cycles prior to the abnormal triggering time. At a power frequency of 50 Hz, the corresponding time length is 0.16 to 0.40 seconds. When the value is less than 8 power frequency cycles, the impact of short-time switching jitter and sampling spikes on the descriptive quantities is considered. The impact increases; when the value is greater than 20 power frequency cycles, the operating condition drift before the abnormal trigger will be introduced too much, weakening the ability to characterize the true baseline before the trigger. The synchronous sampling frequency of voltage and current response is set to 3.2 kHz to 12.8 kHz, which corresponds to 64 to 256 times the power frequency. This range can cover the resolution required for subharmonics, low-order harmonics, waveform distortion and phase shift extraction, while not causing excessive sampling load on the field protection and measurement and control terminal. The shell temperature and ambient temperature are obtained using a low-speed sampling channel, and the sampling period is set to 1 second to 5 seconds.

[0044] After acquiring multi-source response data, the topology status information and thermal status information are tagged. For the topology status information, based on the switching relationship and steady-state conduction relationship of the primary side switch status within the preset analysis period, the wiring status of the voltage transformer, the connection status before and after switching, and whether it is in the stable stage after switching before the abnormal trigger are identified, and a topology status tag is generated. This processing can be completed using state machine recognition or rule recognition. For the thermal status information, the difference sequence between the shell temperature and the ambient temperature is first calculated to obtain the original temperature rise sequence; then, low-pass filtering is performed on the original temperature rise sequence to obtain the smoothed temperature rise.

[0045] Low-pass filtering can be performed using first-order digital low-pass filtering or moving average filtering. The filtering time window is set to 30 to 180 seconds. When the window is less than 30 seconds, wind-cooling disturbances and environmental fluctuations have a significant impact on temperature rise judgment. When the window is greater than 180 seconds, the temperature rise change before and after the abnormal trigger is over-averaged. Then, the smoothed temperature rise is converted to the reference temperature rise by performing a ratio conversion to form a dimensionless temperature rise ratio. The reference temperature rise is the type test temperature rise value under rated operating conditions, or the steady-state temperature rise benchmark value obtained from the statistical analysis of the healthy operation phase after the equipment is put into operation. Then, thermal status labels are generated segment by segment based on the temperature rise ratio.

[0046] Preferably, a temperature rise ratio between 0 and 0.30 corresponds to a low heat load state, between 0.30 and 0.60 corresponds to a medium heat load state, between 0.60 and 1.00 corresponds to a high heat load state, and greater than 1.00 corresponds to an overheating edge state. Finally, the topology state label and the thermal state label are combined to form a context label. The purpose of using context labels is to fix subsequent comparisons within the same type of topology state and thermal state, thereby reducing the interference of topology switching differences and temperature rise offsets on anomaly identification.

[0047] After generating context labels, the voltage and current responses are standardized and characterized. First, frequency band energy extraction is performed on the open delta voltage. This step preferably uses wavelet packet decomposition to extract the energy proportions within the subharmonic and low-order harmonic frequency bands. Wavelet packet decomposition is chosen because the open delta voltage before anomaly triggering typically includes power frequency offset, short-term energy accumulation, and low-order harmonic fluctuations simultaneously. When using fixed-frequency raster analysis, local frequency band shifts within a short window are not easily and stably reflected. The frequency bands are set as follows: the subharmonic frequency band covers 10 Hz to 40 Hz, and the low-order harmonic band covers 100 Hz. The range is from Hz to 250 Hz, corresponding to 0.2 to 0.8 times and 2 to 5 times the power frequency at 50 Hz. This range can cover the low-frequency and low-order harmonic characteristics commonly found in ferroresonance and measurement distortion. Then, the effective value ratio of the open delta voltage and the three-phase phase voltage on the secondary side is converted to form the open delta amplitude ratio; the phase deviation of the three-phase phase voltage on the secondary side is compared to form the phase inconsistency quantity; the time scale normalization of the derivative peak value of the auxiliary winding current is performed to form the auxiliary winding slope ratio; and the harmonic distortion calculation is performed on the voltage and current response to form the second-order distortion ratio.

[0048] The aforementioned frequency band energy extraction, RMS conversion, phase extraction, and harmonic distortion transformation can all be implemented using existing signal processing methods. Their function is to construct basic descriptive quantities. Subsequently, combined with the health reference statistical parameters corresponding to the context labels, conditional standardization processing is performed on the basic descriptive quantities to generate pre-control standardized observation results and voltage anomaly characterization. The preferred source of the health reference statistical parameters is the statistical results of healthy operation samples for more than 30 consecutive days after equipment commissioning, or the statistical results of healthy samples collected during type testing and commissioning phases. The statistical content includes the mean, standard deviation, and allowable deviation interval of each basic descriptive quantity in the corresponding context. After conditional standardization processing, the original quantities are converted into dimensionless observations that can be directly compared in the corresponding context. Subsequent steps can directly use this result to perform template matching and deviation comparison.

[0049] After obtaining the pre-control standardized observation results and voltage anomaly characterization, the magnetic state proxy conversion is then performed. This step is the core method of S1, used to convert the measurable port voltage and loop current on the secondary side into a normalized flux proxy quantity that can characterize the trend of magnetic state changes. The calculation relationship is as follows:

[0050] ;

[0051] In the formula, This is the voltage at the secondary circuit port, in volts. This refers to the secondary circuit current, measured in amperes. This is the equivalent resistance of the secondary winding, in ohms. Leakage inductance is the equivalent inductance, measured in Henry. The reference value for safe magnetic flux is expressed in volt-seconds. This is the length of the integration window, in seconds. It is a dimensionless normalized flux linkage proxy quantity.

[0052] In the above formula, the secondary circuit port voltage is itself a voltage quantity; the product of the equivalent resistance of the secondary winding and the secondary circuit current is a voltage quantity; the product of the leakage inductance, equivalent inductance, and the current derivative is also a voltage quantity. Therefore, the three terms in parentheses have the same dimensions and can be directly added. After integrating this voltage quantity over time, the flux linkage is obtained. Dividing this by the safety flux linkage reference yields the dimensionless normalized flux linkage surrogate quantity. The equivalent resistance and leakage inductance of the secondary winding are obtained through factory calibration, power-off injection test, or maintenance calibration, which are existing parameter identification methods. The safety flux linkage reference is determined by the rated value. The constant voltage, rated frequency, and healthy operating flux boundaries are determined. The integration window length is set to 0.01 to 0.02 seconds (i.e., 0.5 to 1 power frequency cycle). This range can balance the sensitivity of magnetic state characterization and integration stability. To avoid integration drift, half-cycle reset integration or full-cycle reset integration is used for integration. If the sampled waveform shows clipping, sample loss, or synchronization error exceeding the preset range, the normalized flux surrogate quantity in the current window is marked as invalid, and subsequent steps combine the odd-order component ratio of the auxiliary winding for compensation and determination.

[0053] After calculating the normalized flux linkage surcharge, the peak value of the normalized flux linkage surcharge within the preset analysis period is extracted, and a pre-controlled magnetic state risk quantity is formed based on the preset risk mapping interval. Preferably, when the peak value of the normalized flux linkage surcharge is between 0 and 0.60, it is mapped to a low magnetic risk interval; when it is between 0.60 and 0.85, it is mapped to a medium magnetic risk interval; when it is between 0.85 and 1.00, it is mapped to a high magnetic risk interval; and when it is greater than 1.00, it is mapped to an emergency risk interval. The above interval settings are based on the following: when the normalized flux linkage surcharge is close to the safety flux linkage reference, the voltage transformer magnetic circuit is close to the nonlinear operating boundary; when it exceeds the safety flux linkage reference, the risk of entering the saturation edge or saturation region increases significantly. This mapping result directly serves to determine whether the restricted diagnostic control action in S2 is allowed to be executed.

[0054] Through the above processing, S1 finally outputs context labels, pre-control standardized observation results, pre-control magnetic state risk quantities, and voltage anomaly characterizations.

[0055] In this embodiment, S2 is used to execute a restricted diagnostic control action based on the context label, pre-control standardized observation results, pre-control magnetic state risk quantity and voltage anomaly characterization already output by S1. Then, it uses the difference in multi-source response before and after the diagnostic action to generate the observation confidence result, and further generates the risk state result of the controlled object, the formal control path label and the control level. After this processing, the formal hierarchical current limiting and harmonic elimination control adopted by S3 is based on the input that has completed the distinction between true and false anomalies and the risk stratification.

[0056] In this embodiment, the restricted diagnostic control action is preferably implemented through a diagnostic damping branch or a diagnostic current limiting branch. Its mechanism is to introduce additional damping or current limiting effects with limited amplitude and duration into the secondary circuit within the diagnostic window without entering the formal graded current limiting and harmonic elimination control intensity. This induces differentiated responses between the actual abnormal path and the measurement distortion path in terms of abnormal frequency band energy ratio, open delta amplitude ratio, auxiliary winding slope ratio, and secondary distortion ratio. In this embodiment, the requirement for the diagnostic damping branch and the diagnostic current limiting branch is that they can form repeatable and measurable restricted disturbances within one to three power frequency cycles. It does not require the branch material to complete a thermally induced resistance jump within the diagnostic window as a prerequisite. Therefore, the difference in multi-source response before and after the diagnostic action mainly comes from the difference in transient response of the restricted disturbance to the controlled object and the measurement chain.

[0057] Specifically, firstly, based on the pre-controlled magnetic state risk quantity and voltage anomaly characterization, it is determined whether restricted diagnostic control actions are permitted. When the pre-controlled magnetic state risk quantity is lower than the magnetic state emergency boundary and the voltage anomaly characterization is lower than the voltage emergency boundary, the diagnostic damping branch or diagnostic current limiting branch is connected. When the pre-controlled magnetic state risk quantity reaches the magnetic state emergency boundary or the voltage anomaly characterization reaches the voltage emergency boundary, the restricted diagnostic control actions are skipped and the formal graded current limiting and harmonic elimination control is directly entered. The duration of the diagnostic action is set to 1 to 3 power frequency cycles, corresponding to 0.02 to 0.06 seconds under 50 Hz conditions. When the value is less than one power frequency cycle, the diagnostic branch has not yet formed a stable and identifiable difference in the voltage and current response; when the value is greater than three power frequency cycles, the additional impact of the diagnostic disturbance on the original state of the controlled object increases. The upper limit of the peak current of the diagnostic branch is set to 0.30 to 0.60 times the upper limit of the peak current of the formal control branch corresponding to the lowest control level. This value ensures that the diagnostic action is observable, while being lower than the minimum action intensity of the formal control, thus avoiding the diagnostic action from prematurely rewriting the true risk state of the formally controlled object. The connection method of the diagnostic damping branch and the diagnostic current limiting branch can adopt the existing branch control method.

[0058] After executing the restricted diagnostic control action, a post-diagnostic analysis window is established around the end time of the diagnostic action. The post-diagnostic analysis window is set to 2 to 6 power frequency cycles after the end of the diagnostic action. When the window is less than 2 power frequency cycles, the proportion of instantaneous spikes during branch switching is too high; when the window is greater than 6 power frequency cycles, the external operating state continues to evolve, and the response differences caused by the diagnostic action and natural drift will be mixed. Then, the descriptive quantity extraction, thermal state label confirmation, and conditional standardization process in S1 is repeated to generate standardized post-diagnostic observation results. Since the restricted diagnostic control action does not change the primary side wiring structure, and the duration of the diagnostic action is much shorter than the thermal inertia time constant of the temperature channel, the context label in S1 is used in the post-diagnostic analysis stage, and thermal state labels and topology state labels are not regenerated. The purpose of this processing is to ensure that the standardized observation results before and after diagnosis can be directly compared under the same health reference statistical parameter system.

[0059] Subsequently, the standardized observation results after diagnosis are differentiated from the standardized observation results of pre-control to generate multi-source response differences. These multi-source response differences are obtained by subtracting each item from the basic descriptive quantities that have already undergone standardization in S1. The template library stores template responses hierarchically according to context labels and diagnostic action durations. The template response of real network anomalies is formed by statistically analyzing the multi-source response differences of verified real network anomaly samples under the same context labels and the same diagnostic action durations. The template response of measurement distortion is formed by statistically analyzing the multi-source response differences of verified measurement distortion samples under the same context labels and the same diagnostic action durations. The template construction preferably uses median statistics or truncated mean statistics to reduce the influence of extreme samples on the template center. The template sample size is preferably not less than 30 groups. When the sample size is less than 30 groups, the template response fluctuates greatly. When the sample size reaches more than 30 groups, the central response of the hierarchical template begins to stabilize.

[0060] Specifically, for each context label, each diagnostic action duration, and each standardized observation descriptor, corresponding multi-source response difference sequences are extracted from the verified set of real network anomaly samples and the set of measurement distortion samples, respectively. Missing value samples and anomaly samples exceeding preset physical boundaries are first removed, and then bilateral truncation is performed according to the magnitude of the difference value, preferably with a truncation ratio set to 5% to 10%, to obtain an effective sample set. The median of the effective sample set is used as the central response of the corresponding template, and the median absolute deviation of the effective sample set relative to the median, after robust conversion, is used as the fluctuation scale of the corresponding template. Using this method, the template central response is used to characterize the typical difference pattern under this type of assumption, and the template fluctuation scale is used to characterize the allowable fluctuation range of each descriptor under this type of assumption, thereby providing a component-level comparison benchmark for subsequent deviation calculations.

[0061] When the number of valid samples under one context label and one diagnostic action duration is less than 30, it is preferable to merge the samples corresponding to adjacent hot state labels upwards, or to use the template corresponding to the diagnostic action duration of the previous level and mark the template as a low confidence template, so as to maintain the continuity of template calling when there are insufficient samples.

[0062] The observation confidence result is generated according to the following formula: ;

[0063] in, To observe the credibility results; The deviation under the assumption of real network anomalies; To measure the deviation under the distortion assumption.

[0064] For any hypothesis , Its deviation is generated according to the following formula:

[0065]

[0066] in, For the first The difference in response of a standardized observation descriptor before and after the diagnostic action; For context labels Duration of diagnostic actions Next, the The class hypothesis corresponding to the first Each template center responds; For context labels Duration of diagnostic actions Next, the The class hypothesis corresponding to the first Each template fluctuation scale, and ; For the first The weights of each descriptor, and ; The numbers up to 6 correspond to the abnormal frequency band energy ratio, open delta amplitude ratio, phase inconsistency, auxiliary winding slope ratio, secondary distortion ratio, and temperature rise ratio, respectively.

[0067] Here is the Huber robust bias function, defined as:

[0068] ;

[0069] The aforementioned multi-source response differences, template central response, and template fluctuation scale are all dimensionless standardized quantities, therefore As a dimensionless quantity, after being mapped by the Huber robust bias function, It is still a dimensionless quantity.

[0070] The inflection threshold of Huber robust bias function The preferred setting is 1.2 to 1.8 standardized units. To adapt the transition threshold to specific operating conditions, in the current context label... Extract the first from the corresponding set of healthy samples. The absolute standardized deviation sequence of each standardized observation descriptor is calculated, and its 90th percentile is determined. ; then according to By determining the corresponding inflection threshold, this method maintains high deviation sensitivity under conditions of small background fluctuations and suppresses excessive amplification of deviation by individual peak points under conditions of frequent topology switching or large thermal state fluctuations.

[0071] The sum of the weights of each descriptive quantity is 1. The weights for the abnormal frequency band energy ratio, the open delta amplitude ratio, the phase inconsistency ratio, the auxiliary winding slope ratio, the second-order distortion ratio, and the temperature rise ratio are all set to 0.05 to 0.15.

[0072] The basis for this weighting configuration is that the restricted diagnostic control action mainly changes the voltage frequency band, open delta amplitude, auxiliary winding response, and secondary distortion, while the temperature rise ratio mainly serves as a slow variable constraint within the cycle-level window. Through the above calculations, the matching degree of the network real anomaly template response and the matching degree of the measurement distortion template response are mapped to a unified observation confidence result. The closer the observation confidence result is to 1, the closer the current observation is to the network real anomaly; the closer the observation confidence result is to 0, the closer the current observation is to the measurement distortion.

[0073] After obtaining the observation confidence results, the risk status results of the controlled object are further generated. These results are composed of magnetic state risk components, thermal state risk components, and voltage stress risk components. The magnetic state risk components are generated based on the difference results of the normalized magnetic flux surrogate quantity in the diagnosis stage. Specifically, the peak value of the normalized magnetic flux surrogate quantity in the analysis window after diagnosis is extracted and compared with the corresponding peak value of the pre-controlled magnetic state risk quantity in S1. Then, the comparison results are mapped to the low magnetic risk interval, medium magnetic risk interval, high magnetic risk interval, and emergency risk interval. This mapping interval is consistent with S1, thereby ensuring that the risk characterization before and after has a unified physical meaning. The thermal state risk components are generated based on the difference results of the smoothed temperature rise in the diagnosis stage.

[0074] Because the diagnostic phase has a short timescale, the thermal state risk component is mainly categorized based on the current smooth temperature rise interval, with the difference in smooth temperature rise before and after diagnosis used as a correction term. This avoids physically amplifying short-term temperature rise fluctuations unreasonably. The voltage stress risk component is generated based on the voltage anomaly characterization during the diagnostic phase. Specifically, the abnormal frequency band energy ratio, the open triangle amplitude ratio, and the second-order distortion ratio are weighted, normalized, and converged to form the voltage stress risk component. The preferred weights for these three components are 0.40, 0.35, and 0.25, respectively. This configuration reflects the actual situation where overvoltage and resonance evolution are more sensitive to abnormal frequency band energy and open triangle amplitude.

[0075] Based on this, the magnetic state risk component, thermal state risk component, and voltage stress risk component are converged to generate the risk state result of the controlled object. The convergence weights of the three types of risk components are preferably set as follows: magnetic state risk component 0.45 to 0.55, voltage stress risk component 0.25 to 0.35, and thermal state risk component 0.15 to 0.25. The engineering basis for this configuration is that the purpose of S2 is to provide the path and strength of formal control for S3. Magnetic state risk has the greatest impact on the safety boundary of the controlled object, followed by voltage stress, and thermal state undertakes the boundary constraint function. Then, the formal control path label is determined based on the observation confidence result, and the control level is determined based on the risk state result of the controlled object.

[0076] The formal control path labeling method is as follows: when the observation confidence result is between 0.60 and 1.00, it corresponds to the network true anomaly path; when the observation confidence result is between 0 and 0.60, it corresponds to the measurement distortion path. The reason for setting the path switching boundary to 0.60 is that the formal hierarchical current limiting and harmonic elimination control will directly change the state of the controlled object. Therefore, the path selection should be based on a confidence interval with clear evidentiary advantage. When the observation confidence result is between 0.50 and 0.60, although the network true anomaly hypothesis has begun to dominate, its advantage over the measurement distortion hypothesis is still insufficient to support the formal path switching. Therefore, a conservative judgment is still performed according to the measurement distortion path. When the observation confidence result reaches 0.60 or above, it indicates that the matching strength corresponding to the network true anomaly hypothesis has formed a clear advantage over the measurement distortion hypothesis. At this time, the path is switched to the network true anomaly path.

[0077] Through the above processing, S2 finally outputs the observation confidence result, the risk status result of the controlled object, the formal control path label, and the control level. Compared with the existing technology, the technical effect of this step is reflected in two aspects. First, the restricted diagnostic control action provides a controlled disturbance basis for the generation of multi-source response differences. The observation confidence result is based on the comparison between the difference before and after the diagnostic action and the template response. Therefore, it can distinguish between real network anomalies and measurement distortions under the condition that the measurement chain is disturbed. Second, the risk status result of the controlled object simultaneously introduces three types of risk components: magnetic state, voltage stress, and thermal state. This makes the formal control path and control level constrained by both the observation confidence result and the actual risk status of the controlled object, thereby providing a stable input for S3.

[0078] In this embodiment, S3 is used to perform formal hierarchical current limiting and harmonic elimination control based on the formal control path label and control level output by S2. After the formal control is implemented, recovery characteristic quantities are extracted from the electrical recovery process and the thermal state recovery process, respectively, and then the observation reliability recovery result, magnetic state fallback result, thermal state fallback result, recurrence probability result, residual magnetic state risk quantity, and residual thermal state risk quantity are generated. The technical problem solved by S3 is that S2 has completed the generation of observation reliability results and controlled object risk state results, but after the formal hierarchical current limiting and harmonic elimination control action is put into action, the secondary side-view measurement may return to the normal range first, while the magnetic state and thermal state of the voltage transformer body are still in the high-risk range. Therefore, it is necessary to perform hierarchical evaluation of the recovery process after formal control to prevent the apparent recovery from being misjudged as the object recovery.

[0079] First, based on the formal control path label and control level, formal hierarchical current limiting and harmonic suppression control is executed. When the formal control path label corresponds to a real abnormal path in the network, the damping branch is connected first, and then the control level of the current limiting branch is adjusted. When the formal control path label corresponds to a measurement distortion path, the control level of the current limiting branch is adjusted first, and then the damping branch is connected. The basis for this setting is that the main problem corresponding to the real abnormal path in the network is the propagation of abnormal oscillation energy and abnormal voltage, and prioritizing the connection of the damping branch can dissipate the oscillation energy first; the main problem corresponding to the measurement distortion path is voltage mutual inductance. To address magnetic state deviation and secondary circuit distortion, prioritizing the adjustment of the current-limiting branch control level can constrain current and flux sway. The control level is implemented in a three-tiered manner: Level 1 corresponds to mild risk conditions, Level 2 to moderate risk conditions, and Level 3 to severe risk conditions. The control parameters for the damping branch and the current-limiting branch corresponding to each level are formed through factory calibration and commissioning calibration. The damping branch control parameters include conduction mode, conduction duty cycle, and duration; the current-limiting branch control parameters include current-limiting channel access status, current-limiting level, and holding time.

[0080] In this embodiment, the formal graded current limiting and harmonic suppression control is preferably implemented through damping branches and current limiting branches. The damping branches and current limiting branches preferably adopt a branch structure containing a PTC core device. As the abnormal current and device temperature rise change, the impedance state of the PTC core device changes in stages, so that different control levels correspond to different damping and current limiting effects. When the formal control path label corresponds to the actual abnormal path of the network, the damping effect is established first to suppress the abnormal oscillation energy. When the formal control path label corresponds to the measurement distortion path, the current limiting effect is established first to constrain the current and flux swing. In this way, the formal control path label and control level can form a correspondence with the dynamic action process of the branch.

[0081] After the formal graded current limiting and harmonic elimination control is started, an electrical recovery window and a thermal confirmation window are established respectively. The electrical recovery window is used to evaluate the recovery of the cycle-level electrical response after the formal control is started. It is preferably set to 4 to 12 power frequency cycles after the formal graded current limiting and harmonic elimination control is started. When the window is less than 4 power frequency cycles, the transient of branch switching has not yet decayed completely, and the proportion of spikes and transition components in the electrical response is relatively high. When the window is greater than 12 power frequency cycles, changes in external operating conditions begin to superimpose, and the specificity of the recovery evaluation decreases. The thermal confirmation window is used to evaluate the decline of the thermal state after the formal control is started. It is preferably set to 30 to 180 seconds after the formal graded current limiting and harmonic elimination control is started. When the window is less than 30 seconds, the difference between the shell temperature and the ambient temperature has not yet reflected the change in thermal inertia. When the window is greater than 180 seconds, the interference of changes in heat dissipation conditions on the temperature rise judgment increases. By separating the electrical recovery window and the thermal confirmation window, it is possible to avoid mixing the judgment of cycle-level electrical recovery and second-level thermal recovery, and ensure that the physical process is consistent with the evaluation time window.

[0082] Within the electrical recovery window, the first step is to generate reliable recovery results of the observations. Specifically, the descriptive quantity extraction and conditional standardization process in S1 is repeated to generate standardized observation results after formal graded current limiting and harmonic elimination control. Since formal graded current limiting and harmonic elimination control does not change the primary side topology and the electrical recovery window is within a short time range, the context labels formed in S1 are reused without regenerating them. Then, the standardized deviations corresponding to the abnormal frequency band energy ratio, open delta amplitude ratio, phase inconsistency, auxiliary winding slope ratio, and secondary distortion ratio are extracted from the standardized observation results after formal graded current limiting and harmonic elimination control.

[0083] The standardized deviation mentioned here refers to the absolute deviation of each descriptive quantity from the center of the health reference statistical parameter in the corresponding context. Subsequently, the standardized deviation after formal graded current limiting and harmonic elimination control is compared with the standardized deviation before formal graded current limiting and harmonic elimination control to obtain the deviation improvement rate. :

[0084] ;

[0085] in, Before formal graded current limiting and harmonic elimination control The absolute deviation of a standardized observational descriptor from the center of health reference statistical parameters; For the first time after formal graded current limiting and harmonic elimination control The absolute deviation of a standardized observational descriptor from the center of health reference statistical parameters; To prevent stable terms with a denominator of zero, a value of 0.01 is preferred. A value greater than 0 indicates that the degree of deviation decreases after formal control is implemented. This indicates that no improvement has occurred after formal control was implemented.

[0086] Then Convert to recovery contribution value in the range of 0 to 1 The preferred approach is the piecewise linear mapping method as follows: ;

[0087] When this piecewise linear mapping method is adopted, the recovery contribution value increases slowly when the deviation improvement rate is small, so as to avoid small fluctuations being amplified and interpreted as significant recovery; after the deviation improvement rate enters the middle range, the recovery contribution value increases approximately linearly to ensure the resolution of the main recovery stage; after the deviation improvement rate reaches the higher range, the recovery contribution value approaches 1 to reflect that the observation has approached the state of complete recovery.

[0088] The various recovery contribution values ​​are then aggregated according to preset weights to form a reliable recovery result. The preferred weight settings are: abnormal frequency band energy ratio 0.25 to 0.30, open delta amplitude ratio 0.20 to 0.25, phase inconsistency 0.10 to 0.15, auxiliary winding slope ratio 0.15 to 0.20, and second-order distortion ratio 0.20 to 0.25. These values ​​are based on the fact that after formal control, electrical recovery is primarily manifested in the convergence of abnormal frequency bands, the regression of open delta amplitude, and the reduction of second-order distortion. The auxiliary winding response and phase consistency serve as auxiliary correction functions. Therefore, the resulting reliable recovery result reflects the degree of recovery of the observation chain relative to the healthy reference state after formal control.

[0089] After generating the observational credible recovery results, the magnetic state fallback results and residual magnetic state risk quantities are generated. The magnetic state fallback results are composed of two types of recovery characteristics. The first type of recovery characteristic is the difference between the normalized flux surrogate quantity before and after the formal graded current limiting and harmonic elimination control. The calculation process of the normalized flux surrogate quantity has been clarified in S1, and this step directly calls this result.

[0090] Specifically, the peak value of the normalized flux linkage surrogate quantity after the formal graded current limiting and harmonic elimination control is extracted and compared with the peak value of the normalized flux linkage surrogate quantity in the corresponding window before the formal graded current limiting and harmonic elimination control to obtain the peak value difference result. The second type of recovery characteristic quantity is the difference result of the odd-order component ratio of the auxiliary winding. The odd-order component ratio of the auxiliary winding is obtained by performing harmonic decomposition on the auxiliary winding current. Specifically, the amplitudes of the 3rd, 5th and 7th harmonics are extracted, the squares of the three are summed, and the square root is taken to obtain the comprehensive amplitude of the odd-order harmonics. Then, the ratio of the comprehensive amplitude of the odd-order harmonics to the fundamental amplitude is converted to obtain the dimensionless odd-order component ratio of the auxiliary winding. The reason for this processing is that after the magnetic state of the voltage transformer enters the nonlinear region, the odd-order harmonic response of the auxiliary winding current will be significantly enhanced. If the magnetic state returns to the relatively stable region after formal control, the odd-order component ratio of the auxiliary winding will fall back synchronously.

[0091] Subsequently, the results of the normalized flux surrogate peak difference and the auxiliary winding odd-order component ratio difference are converted into fallback contribution values ​​in the range of 0 to 1, and then aggregated according to weights to generate magnetic state fallback results. The weight of the normalized flux surrogate peak difference is preferably set to 0.60 to 0.75, and the weight of the auxiliary winding odd-order component ratio difference is preferably set to 0.25 to 0.40. This configuration conforms to physical reality because the normalized flux surrogate directly characterizes the magnetic state boundary position, and the auxiliary winding odd-order component ratio plays a role in nonlinear response correction. The fallback contribution values ​​in the magnetic state fallback results preferably follow the above piecewise linear mapping framework, only replacing the input quantities with the normalized flux surrogate peak improvement rate and the auxiliary winding odd-order component ratio improvement rate.

[0092] After generating the magnetic state fallback result, the residual magnetic state risk quantity is generated based on the range of the peak value of the normalized flux surrogate quantity after formal graded current limiting and harmonic elimination control. This range division is consistent with S1 and S2, that is, when the peak value of the normalized flux surrogate quantity is between 0 and 0.60, it corresponds to the low magnetic risk range; when it is between 0.60 and 0.85, it corresponds to the medium magnetic risk range; when it is between 0.85 and 1.00, it corresponds to the high magnetic risk range; and when it is greater than 1.00, it corresponds to the emergency risk range. This ensures that the physical meaning of the magnetic state risk quantity remains consistent in the preceding and following steps.

[0093] Within the thermal confirmation window, the thermal state fallback result and residual thermal state risk quantity are generated. Specifically, following the thermal state processing flow in S1, the difference sequence between the shell temperature and the ambient temperature is first calculated, and then a low-pass filter is applied to the temperature difference sequence to obtain the smooth temperature rise before and after the formal graded current limiting and harmonic elimination control. The smooth temperature rise is used here because the temperature channel sampling speed is slow, and there are environmental fluctuations and sensor noise in the original temperature difference sequence. Directly using the original temperature difference will amplify the judgment error.

[0094] Then, the smoothed temperature rise after formal graded current limiting and harmonic elimination control is compared with the smoothed temperature rise before formal graded current limiting and harmonic elimination control to obtain the temperature rise difference result. This temperature rise difference result is then converted into a fallback contribution value in the range of 0 to 1. The fallback contribution value in the thermal state fallback result preferably follows the above piecewise linear mapping framework, only replacing the input with the smoothed temperature rise improvement rate to form the thermal state fallback result. After that, the residual thermal state risk quantity is generated based on the temperature rise ratio range corresponding to the smoothed temperature rise after formal graded current limiting and harmonic elimination control. The temperature rise ratio range is consistent with S1, that is, when the temperature rise ratio is between 0 and 0.30, it corresponds to the low heat load range; when it is between 0.30 and 0.60, it corresponds to the medium heat load range; when it is between 0.60 and 1.00, it corresponds to the high heat load range; and when it is greater than 1.00, it corresponds to the overheating edge range. After this processing, the thermal state fallback result is used to represent the degree of thermal state release, and the residual thermal state risk quantity is used to represent the thermal boundary pressure that is still retained after formal control.

[0095] After generating the observational reliable recovery results, magnetic state fallback results, thermal state fallback results, residual magnetic state risk quantity, and residual thermal state risk quantity, the recurrence probability results are then generated. The recurrence probability results reflect the possibility of the anomaly reappearing after formal graded current limiting and harmonic elimination control. Its input quantities are the normalized flux surrogate variance after formal graded current limiting and harmonic elimination control, the voltage anomaly characterization after formal graded current limiting and harmonic elimination control, and the residual magnetic state risk quantity.

[0096] Specifically, the variance of the normalized flux surrogate quantity sequence is first calculated within the electrical recovery window. The normalized flux surrogate quantity has already been normalized in S1 using the safety flux benchmark, so this variance is still a dimensionless quantity and can directly participate in subsequent mappings along with other dimensionless inputs. The larger the variance of the normalized flux surrogate quantity, the more obvious the magnetic state oscillation after formal control. Then, the voltage anomaly characterization after formal graded current limiting and harmonic elimination control is extracted. This characterization has already been formed in S1 through conditional standardization. Here, the corresponding result is directly called. Then, the variance of the normalized flux surrogate quantity, the voltage anomaly characterization, and the residual magnetic state risk quantity are input into the monotonic logic mapping model to generate the recurrence probability result.

[0097] The parameters of the model are obtained through offline calibration using historical samples. The number of historical samples should preferably be no less than 50 sets, covering the result ranges after formal control in the low, medium, and high categories. If the number of historical samples is insufficient, a segmented mapping method can be used instead: when all three inputs are in the low range, the recurrence probability result is mapped to 0 to 0.30; when any input is in the medium range and no input is in the high range, the recurrence probability result is mapped to 0.30 to 0.70; when at least one input is in the high range, the recurrence probability result is mapped to 0.70 to 1.00. This mapping method is consistent with engineering judgment because the recurrence risk is simultaneously affected by residual anomaly characteristics, magnetic state oscillation, and residual magnetic state risk.

[0098] Through the above processing, S3 finally outputs the observation reliable recovery result, magnetic state fallback result, thermal state fallback result, recurrence probability result, residual magnetic state risk quantity, and residual thermal state risk quantity.

[0099] In this embodiment, S4 is used to generate a control withdrawal permission result based on the observation reliable recovery result, magnetic state fallback result, thermal state fallback result, recurrence probability result, residual magnetic state risk quantity, and residual thermal state risk quantity already output by S3. Based on this, it executes exit control, graded exit, maintains the current control level, or adjusts to the corresponding upgraded control level. At the same time, it completes the template library update. The technical problem solved by this step is that after the formal graded current limiting and harmonic elimination control is put into operation, the secondary side waveform may have recovered to the normal range, but the magnetic state and thermal state of the voltage transformer body have not yet returned to the safe range. If the control effect of the damping branch and the current limiting branch is directly removed at this time, it is easy to cause re-oscillation, repeated impact, and control oscillation.

[0100] In this embodiment, the generation of the control withdrawal permission result is preferably determined in conjunction with the impedance recovery process of the PTC core device after the formal graded current limiting and harmonic elimination control. Since the PTC core device needs to undergo a recovery process from a high-resistance state to a low-resistance state after the abnormality is cleared, if the control action is withdrawn based solely on the recovery of the secondary side waveform, it is easy to cause re-oscillation before the device impedance has stabilized. Therefore, the control withdrawal permission result is simultaneously constrained by the observed reliable recovery result, the magnetic state fallback result, the thermal state fallback result, and the recurrence probability result, so that the control withdrawal action and the PTC branch recovery process are coordinated.

[0101] Therefore, S4 needs to unify the constraints of observation layer recovery, object layer fallback, and recurrence risk into the same decision space, and then perform control exit and template update according to the control withdrawal permission result. Specifically, first, joint constraint judgment is performed on the observation reliable recovery result, magnetic state fallback result, thermal state fallback result, and recurrence probability result to generate the control withdrawal permission result. The control withdrawal permission result is calculated according to the following formula:

[0102] ;

[0103] In the formula, The result of the withdrawal of control; To observe reliable recovery results; This is the result of the magnetic state falling back; This is the result of the temperature drop. This represents the recurrence probability result. Let be the weight, and satisfy... The above four input quantities have been converted into dimensionless quantities in the range of 0 to 1 in S3. The physical meaning of this calculation relationship is: first, the comprehensive evaluation of the recovery side is formed by using the observation reliable recovery results, magnetic state fall-off results and thermal state fall-off results, and then the risk reduction is performed on the comprehensive evaluation of the recovery side by using the recurrence probability results to obtain the control withdrawal permission result.

[0104] The optimal weighting is set as follows: 0.35 to 0.50 for the observation confidence recovery result, 0.30 to 0.45 for the magnetic state fallback result, and 0.15 to 0.25 for the thermal state fallback result. This setting is based on the fact that the control withdrawal action first requires the observation link to have recovered to a credible state and the magnetic state to have returned to a relatively stable range. The thermal state plays a role in boundary correction. If the thermal state weight is higher than 0.25, thermal inertia will cause excessive lag in control withdrawal; if the thermal state weight is lower than 0.15, the thermal boundary constraint is insufficient; if the observation confidence recovery result weight is lower than 0.35, the constraint effect of secondary observation recovery on control withdrawal is insufficient; if the magnetic state fallback result weight is lower than 0.30, a higher control withdrawal permission result may still be formed when the magnetic state has not fully fallen back. The recurrence probability result is multiplicatively reduced into the formula. Its function is to suppress scenarios where the overall evaluation of the recovery side is good but the probability of re-oscillation is still high. Therefore, the larger the recurrence probability result, the smaller the control withdrawal permission result.

[0105] After obtaining the control withdrawal permission result, the control withdrawal permission result is then jointly compared with the control withdrawal permission threshold, the easing conditions, and the safety boundaries corresponding to the residual magnetic state risk and residual thermal state risk. The control withdrawal permission threshold is preferably set to 0.75 to 0.90. When the threshold is below 0.75, the control withdrawal conditions are too lenient, and it is easy to misjudge short-term apparent recovery as a control withdrawal state. When the threshold is above 0.90, the control withdrawal conditions are too strict, and the holding time of formal graded current limiting and harmonic elimination control is significantly prolonged. The control withdrawal permission range corresponding to the easing conditions is preferably set to 0.45 to 0.75. This range indicates that the comprehensive evaluation on the recovery side has been formed, but the recurrence probability result or residual risk has not yet met the direct control withdrawal conditions. At this time, it is appropriate to adopt a graded withdrawal method to gradually release the control effect.

[0106] The safety boundary for the residual magnetic state risk is preferably set to 0.55 to 0.65, and the safety boundary for the residual thermal state risk is preferably set to 0.40 to 0.60. The safety boundary for the residual magnetic state risk is set within the range of 0.55 to 0.65 because the division of the magnetic state risk range in S1 and S3 has shown that the area around 0.60 is the boundary between the low magnetic risk range and the medium magnetic risk range. Setting the safety boundary within this range can ensure that the magnetic state is in the low risk range or close to the low risk range when the control is withdrawn.

[0107] The safety boundary for residual thermal state risk is set between 0.40 and 0.60 because the thermal state declines more slowly than the electrical state declines. If the boundary is below 0.40, the control holding time will be significantly too long; if the boundary is above 0.60, the thermal boundary constraint is too loose.

[0108] When the control withdrawal permission result reaches the control withdrawal permission threshold, and the residual magnetic state risk and residual thermal state risk respectively meet the corresponding safety boundaries, the control exit is executed. The specific execution method is as follows: first, the control effect of the damping branch and the current limiting branch is removed, and then the monitoring link continues to operate to detect whether the recurrence probability result rises within a short window after the control exit. The observation period after the control exit is preferably set to 5 to 20 power frequency cycles. When the observation period is less than 5 power frequency cycles, the electrical redistribution process after the branch is removed has not been completed. When the observation period is greater than 20 power frequency cycles, it exceeds the necessary range of the short-term exit verification stage. If the recurrence probability result is always lower than 0.20 within the observation period, and the residual magnetic state risk and residual thermal state risk do not re-enter their respective high-risk ranges, then the control exit is valid.

[0109] When the control withdrawal permission result meets the easing conditions but does not reach the control withdrawal permission threshold, a graded withdrawal is executed. The graded withdrawal adjusts the conduction intensity of the damping branch or the control level of the current limiting branch according to the preset steps. For the damping branch, a 3-level conduction intensity step is preferred, with each level adjusting by 10% to 25% relative to the previous level. The holding time between steps is set to 5 to 20 power frequency cycles. When the amplitude is less than 10%, the change in control effect brought by a single adjustment is too small, and the graded withdrawal effect is not obvious. When the amplitude is greater than 25%, a single adjustment is close to a one-time withdrawal, which is easy to cause a response rebound. For the current limiting branch, it is preferred to pull back step by step according to the control level, one level at a time, and re-enter the recovery feature quantity extraction process corresponding to S3 after the pullback. The reason for adopting graded withdrawal is that the controlled object after formal graded current limiting and harmonic elimination control may still be in the intermediate range where the observed reliable recovery result is relatively high, but the magnetic state fall-off result or thermal state fall-off result has not been completely stable. Direct control withdrawal will bring large fluctuations. Gradually releasing the control effect can reduce control oscillations.

[0110] If the withdrawal of control permission does not meet the conditions for easing restrictions, the current control level will be maintained or adjusted to the corresponding upgraded control level. The specific determination criteria are as follows: if the recurrence probability is between 0.70 and 1.00, or the residual magnetic state risk is in the high magnetic risk range, or the residual thermal state risk is in the overheating edge range, then the current control level will be adjusted to the corresponding upgraded control level; if none of the above situations occur, the current control level will be maintained, and the corresponding upgraded control level will be determined according to the rule of increasing the current control level by 1 level; if the current control level is already level 3, level 3 will be maintained.

[0111] After completing the exit control, graded exit, maintaining the current control level or adjusting to the corresponding upgraded control level, the template library is updated. The template library update includes updating health reference statistics parameters, updating network real anomaly template responses, and updating measurement distortion template responses.

[0112] Specifically, the context labels, multi-source response differences, observed reliable recovery results, magnetic state decline results, thermal state decline results, recurrence probability results, and final treatment results corresponding to this treatment are written back to the template library. For health reference statistical parameters, only samples that meet the following conditions are accepted into the update set: the final treatment result is successful exit from control, the recurrence probability result is kept below 0.20 during the observation period, and the residual magnetic state risk and residual thermal state risk are both in their respective low-risk ranges. This processing can prevent samples with residual risks from entering the health reference statistical parameter set.

[0113] For network real anomaly template responses, only samples corresponding to network real anomaly paths with formal control path labels and subsequently verified as network real anomalies are accepted into the update set. For measurement distortion template responses, only samples corresponding to measurement distortion paths with formal control path labels and subsequently verified as not having network-side anomalies are accepted into the update set. Template updates preferably adopt a sliding update method, that is, after a new sample enters, the corresponding template center response and fluctuation range are corrected according to a preset update weight. The update weight is preferably set to 0.05 to 0.15. When the update weight is lower than 0.05, the template response adapts slowly to the new operating conditions; when the update weight is higher than 0.15, the impact of a single sample on the template center is too large. Through this update mechanism, the template library can absorb effective samples under new operating conditions while maintaining historical stability, thereby improving the adaptability of the observation confidence results generated in S2.

[0114] Through the above processing, S4 finally completes the generation of control withdrawal permission results, exit control or hierarchical exit or control level adjustment, and template library update.

[0115] In this embodiment, the damping branch and current-limiting branch can be implemented using control branches that have both damping and current-limiting functions. Preferably, a composite branch structure containing a PTC core device is adopted. During the formal graded current-limiting and harmonic elimination control and the subsequent control removal stage, the PTC core device can utilize its impedance change characteristics caused by temperature rise to provide support for the damping strength and current-limiting strength during the continuous action stage. The diagnostic damping branch and diagnostic current-limiting branch corresponding to the restricted diagnostic control action are only required to form additional disturbances with constrained amplitude and duration within the diagnostic window to ensure that the differences in multi-source responses before and after the diagnostic action are comparable.

[0116] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for overvoltage identification and current limiting harmonic suppression control of a voltage transformer, characterized in that, Includes the following steps: S1. Obtain multi-source response data within the preset analysis period corresponding to the abnormal triggering time, perform labeling processing on the topological state information and thermal state information, generate context labels, and perform standardized characterization and magnetic state proxy conversion on the voltage and current response to generate pre-control standardized observation results, pre-control magnetic state risk quantity and voltage anomaly characterization. S2. When the risk quantity of the pre-controlled magnetic state is within the diagnostic allowable range, implement restricted diagnostic control actions. Based on the differences in multi-source responses before and after the restricted diagnostic control actions, and combined with the network real abnormal template responses and measurement distortion template responses in the template library, generate observation confidence results and controlled object risk state results to determine the formal control path label and control level. S3. Based on the formal control path label and control level, perform formal hierarchical current limiting and harmonic elimination control, and extract recovery characteristic quantities around the electrical recovery process and thermal state recovery process after control to generate observational reliable recovery results, magnetic state fall-off results, thermal state fall-off results and recurrence probability results; S4. Perform joint constraint judgment on the observation reliable recovery results, magnetic state fall-off results, thermal state fall-off results and recurrence probability results, generate control withdrawal permission results, and perform the following actions based on the control withdrawal permission results: remove the control action of the damping branch and the current limiting branch, adjust the conduction strength of the damping branch or the control level of the current limiting branch, maintain the current control level or adjust the current control level to the corresponding upgraded control level, and update the template library at the same time.

2. The method for overvoltage identification and current limiting harmonic suppression control of a voltage transformer according to claim 1, characterized in that, The multi-source response data includes the secondary three-phase phase voltage, open delta voltage, auxiliary winding current, secondary circuit port voltage, secondary circuit current, primary side switch status, housing temperature, and ambient temperature within the preset analysis period corresponding to the abnormal triggering time. Among them, the secondary three-phase phase voltage, open delta voltage, auxiliary winding current, secondary circuit port voltage, and secondary circuit current constitute the voltage and current response, the primary side switch status constitutes the topology state information, and the housing temperature and ambient temperature constitute the thermal state information.

3. The method for overvoltage identification and current limiting harmonic suppression control of a voltage transformer according to claim 2, characterized in that, Tagging processing is performed on topology status information and thermal status information, including: generating topology status tags based on the primary side switch state switching relationship and steady-state conduction relationship within a preset analysis period corresponding to the abnormal trigger time; Low-pass filtering is performed based on the temperature difference sequence between the shell temperature and the ambient temperature to obtain a smooth temperature rise. The smooth temperature rise is then converted into a ratio with the reference temperature rise, and a thermal status label is generated according to the preset temperature rise range. The topology status label and the thermal status label are combined to generate a context label.

4. The method for overvoltage identification and current limiting harmonic suppression control of a voltage transformer according to claim 3, characterized in that, The voltage and current responses are standardized and the magnetic state proxy is converted, including: performing frequency band energy extraction on the open delta voltage, performing effective value ratio conversion on the open delta voltage and the secondary three-phase phase voltage, performing phase deviation comparison on the secondary three-phase phase voltage, performing slope normalization on the auxiliary winding current, performing harmonic distortion conversion on the voltage and current responses, and performing conditional standardization processing in combination with the health reference statistical parameters corresponding to the context label to generate pre-control standardized observation results and voltage anomaly characterization. Based on the secondary circuit port voltage, secondary circuit current, and calibrated equivalent parameters of the secondary winding, the magnetization branch voltage is reconstructed and the normalized flux linkage proxy is calculated. The pre-controlled magnetic state risk quantity is generated based on the peak value of the normalized flux linkage proxy.

5. The method for overvoltage identification and current limiting harmonic suppression control of a voltage transformer according to claim 4, characterized in that, Implementing restricted diagnostic control actions includes: when the pre-controlled magnetic state risk quantity is lower than the magnetic state emergency boundary and the voltage anomaly characterization is lower than the voltage emergency boundary, connecting the diagnostic damping branch or the diagnostic current limiting branch, and limiting the duration of the restricted diagnostic control action to between one and three power frequency cycles, and limiting the peak current of the diagnostic branch to within the preset safety upper limit. When the pre-controlled magnetic state risk quantity reaches the magnetic state emergency boundary or the voltage anomaly characterization reaches the voltage emergency boundary, the restricted diagnostic control action is skipped and the formal graded current limiting and harmonic elimination control is directly entered.

6. The method for overvoltage identification and current limiting harmonic suppression control of a voltage transformer according to claim 5, characterized in that, Generate observation confidence results, including: extracting standardized post-diagnosis observation results around the response period following restricted diagnostic control actions; The standardized observation results after diagnosis are differentiated from the standardized observation results before control to generate multi-source response differences. Based on context labels and the duration of restricted diagnostic control actions, the network real anomaly template response and measurement distortion template response are called from the template library. The deviation of the multi-source response difference from the network’s true anomalous template response and the measurement distortion template response is calculated respectively, and the observation confidence result is generated based on the comparison results of the two types of deviation.

7. The method for overvoltage identification and current limiting harmonic suppression control of a voltage transformer according to claim 6, characterized in that, Generate risk status results for the controlled object and determine formal control path labels and control levels, including: generating magnetic state risk components based on the difference results of normalized magnetic flux proxy quantities during the diagnostic phase; generating thermal state risk components based on the difference results of smoothed temperature rise quantities during the diagnostic phase. Based on the voltage anomaly characterization in the diagnostic phase, a voltage stress risk component is generated. Based on the magnetic state risk component, thermal state risk component, and voltage stress risk component, the risk status result of the controlled object is generated. Based on the observation reliability result, the formal control path label is determined, and the control level is determined based on the risk status result of the controlled object.

8. The method for overvoltage identification and current limiting harmonic suppression control of a voltage transformer according to claim 7, characterized in that, Based on the formal control path label and control level, perform formal hierarchical current limiting and harmonic elimination control, including: when the formal control path label corresponds to a real abnormal network path, first connect the damping branch, and then adjust the control level of the current limiting branch; When the formal control path label corresponds to the measurement distortion path, first adjust the control level of the current limiting branch, then connect the damping branch, and call the corresponding branch control parameters according to the control level to implement formal graded current limiting and harmonic elimination control. After the formal graded current limiting and harmonic elimination control, establish electrical recovery window and thermal confirmation window respectively to extract recovery characteristic quantities.

9. The method for overvoltage identification and current limiting harmonic suppression control of a voltage transformer according to claim 8, characterized in that, Generate reliable recovery results for observations, magnetic state fallback results, thermal state fallback results, and recurrence probability results, including: within the electrical recovery window, comparing the deviation of the standardized observation results after formal graded current limiting and harmonic elimination control relative to the health reference statistical parameters with the deviation before formal graded current limiting and harmonic elimination control, and generating reliable recovery results for observations based on the deviation difference results; Based on the difference in normalized flux proxy quantity before and after formal graded current limiting and harmonic elimination control, as well as the difference in the odd-order component ratio of the auxiliary winding, magnetic state fallback results are generated, and residual magnetic state risk quantity is generated. Within the thermal confirmation window, the thermal state fallback result is generated based on the difference in smoothed temperature rise before and after formal graded current limiting and harmonic elimination control, and the residual thermal state risk quantity is generated. The recurrence probability result is generated based on the normalized flux proxy variance after formal graded current limiting and harmonic elimination control, the voltage anomaly characterization after formal graded current limiting and harmonic elimination control, and the residual magnetic state risk quantity.

10. The method for overvoltage identification and current limiting harmonic suppression control of a voltage transformer according to claim 9, characterized in that, Generate control withdrawal permission results and update the template library, including: performing joint constraint judgment on the observation credible recovery results, magnetic state fallback results, thermal state fallback results and recurrence probability results to generate control withdrawal permission results; When the control removal permission result meets the control removal conditions and the residual magnetic state risk and residual thermal state risk respectively meet the corresponding safety boundaries, the control effect of the damping branch and the current limiting branch is removed. When the control release permit result meets the conditions for slow withdrawal but does not meet the conditions for control release, adjust the conduction strength of the damping branch or adjust the control level of the current limiting branch according to the preset steps. If the withdrawal permission result does not meet the conditions for easing restrictions, the current control level is maintained or the current control level is adjusted to the corresponding upgraded control level. The context label, multi-source response difference, observation credible recovery result, magnetic state fallback result, thermal state fallback result, recurrence probability result and final disposal result corresponding to this disposal are written back to the template library to update the health reference statistical parameters, network real anomaly template response and measurement distortion template response under the corresponding context.

Images