A current transformer intelligent control system

Abstract

The present invention discloses an intelligent control system for current transformers, which specifically relates to the technical field of power system monitoring. The control system includes five major units: fluctuation identification, redundant acquisition, error fitting, protection delay and fault response. The control system realizes real-time identification and classification of current fluctuations, has the ability to construct high-speed sampling and fusion signals of alternative paths, judges fluctuation trends and dynamically sets protection strategies through error regression, and delays responses to avoid false operations. Ultimately, when the offset exceeds the limit or alignment fails, a high-level protection instruction is triggered and a fault event is output, thereby ensuring data accuracy and system response stability during power switching. The present invention realizes accurate identification of current fluctuations through multi-dimensional feature extraction, enhances data reliability through redundant sampling and error fitting, and combines delay control with a hierarchical response mechanism to effectively reduce the risk of false operations, improve the accuracy of fault judgment and system protection stability, and is suitable for complex power switching scenarios.

Description

Technical Field

[0001] The present invention relates to the technical field of power system monitoring, and more particularly to a current transformer intelligent control system. Background Art

[0002] As a key measurement and protection component in power systems, current transformers are widely used for real-time current detection, state sensing, and protection control. Under normal operating conditions, traditional current transformers, combined with protection units, continuously monitor steady-state current parameters and trigger protection responses when thresholds are exceeded or anomalies occur.

[0003] However, during the backup power switching phase in data centers or precision industrial equipment, problems inherent in the power switching logic, such as relay intermittent switching, phase drift, and frequency disturbances, can cause slight sudden changes or fluctuations in the current signal within a very short period of time. These fluctuations typically have the following characteristics: short duration (e.g., 5-15 milliseconds); moderate amplitude (tolerable); and unpredictable trends, potentially indicating a true fault. In these scenarios, existing technologies present the following main issues:

[0004] It is impossible to effectively distinguish between tolerable switching fluctuations and real fault signals. Most traditional protection strategies adopt a fixed threshold + fixed response method. When facing transient disturbances, there is a lack of trend judgment and hierarchical processing mechanism, which makes it very easy for false operations or premature triggering of protection responses to occur.

[0005] There is a lack of dynamic tracking and error fitting mechanism for current fluctuation trends. Some current technologies introduce oversampling or redundant acquisition methods, but no stable fitting judgment model has been established, making it difficult to dynamically adjust the protection strategy, resulting in a "rigid response" of the system.

[0006] The protection response behavior lacks a buffer and linkage mechanism, and the execution of control instructions is too rigid. The protection response in the existing technology is basically "judgment established → immediate action", lacking a "state buffer - delayed confirmation - trend linkage" mechanism, and insufficient fault tolerance in complex fluctuation scenarios.

[0007] There are still deficiencies in the collaborative judgment of multi-source data, and feedforward control and event structured management cannot be achieved. Therefore, the present invention proposes an intelligent control system for current transformers in order to solve the above problems. Summary of the Invention

[0008] To achieve the above object, the present invention provides the following technical solutions:

[0009] A current transformer intelligent control system, comprising:

[0010] The fluctuation identification unit is used to obtain the current signal during the switching period through real-time sampling. Based on predefined instantaneous change identification criteria, it extracts the continuous change amplitude, rising edge slope, phase difference change, and frequency mutation value from the current signal, and completes the feature identification and classification of abnormal fluctuation behavior according to data statistical rules;

[0011] The redundant acquisition unit, upon identifying the tolerable switching fluctuation type, immediately triggers the backup sampling path and enables high-speed sampling instructions to obtain redundant current information flow. It then performs a time axis comparison with the main path data and, while ensuring timing continuity, constructs a fused signal result to replace the distorted or drifted segments in the main signal.

[0012] The error fitting unit continuously compares the fused signal with the preset reference data model to perform fitting regression on the error offset trend of the current fluctuation state. Based on the fitting results, the protection trigger threshold and response time are set to make real-time corrections to the main control strategy.

[0013] The protection delay unit automatically delays the execution of protection when it determines that the fluctuation state is within an acceptable range, and outputs a fusion signal as an effective monitoring benchmark. At the same time, it suspends high-priority protection response instructions to prevent false operation due to transient errors and improve power continuity during switching.

[0014] When the offset between the fusion signal and the reference data model exceeds the preset limit, or the redundant sampling path cannot complete effective time alignment, the fault response unit immediately executes the emergency response process, generates a high-level protection instruction, interrupts the signal output process, and sends a fault event identification to the external system for protection processing and log recording of power switching control.

[0015] In a preferred embodiment, the fluctuation identification unit includes an abnormal current fluctuation screening structure, which is used to set a multi-threshold condition judgment mechanism based on the extraction of continuous change amplitude, rising edge slope, phase difference change and frequency mutation value. Different characteristic parameters are combined and calculated in a time-aligned manner to construct a high-dimensional characteristic behavior discrimination space. Through the set three-value hierarchical decision logic, the detection results are marked as stable fluctuation, critical fluctuation or severe abnormality.

[0016] In the time window slip judgment process, the first judgment factor is whether the fitting coefficient between the average amplitude change and the rising rate of three adjacent sampling periods is less than 0.1, and the second judgment factor is whether the instantaneous frequency change is within 5 Hz. When both conditions are met, the current state is determined to be a critical fluctuation.

[0017] The critical fluctuation type is identified as the tolerable switching fluctuation type, which is used as the basis for starting the subsequent backup sampling path, so that the working behavior of the redundant acquisition unit has a traceable judgment basis based on the feature recognition results, thereby ensuring the continuity of the technical chain between fluctuation judgment and data compensation.

[0018] In a preferred embodiment, the redundant acquisition unit establishes a data mirror relationship by operating the dual-path current signal channels in parallel, wherein the set error between the primary path sampling period and the backup path sampling period does not exceed ten microseconds to ensure that the time alignment accuracy meets the data fusion requirements;

[0019] The backup path sampling values must undergo a preliminary noise removal process. The removal method includes one of three methods: maximum difference value method, root mean square limit method, and adaptive judgment based on sliding window filtering. The specific method is automatically switched based on the current data fluctuation type and is not fixed.

[0020] Under the condition that the sampling values meet the redundant timing structure, an interleaved data stream is established by delaying one cycle to ensure the continuity of the redundant data and that it does not cause logical interference with the main path. At the same time, all redundant data will be brought into the phase reconstruction operation process to participate in the construction of the error fitting unit behavior.

[0021] In a preferred embodiment, the error fitting unit includes a dynamic regression correction structure, wherein the regression behavior is analyzed based on the mean deviation trend of the fused signal from the reference model within the current five sampling periods, while taking into account the consistency of the deviation direction and the stability of the duration;

[0022] The regression strategy adopts one of two methods: interval minimum variance fitting method and exponential weighted moving average method. The former is used when the current fluctuation intensity is greater than 10%, and the latter is used when the fluctuation duration is less than 10 milliseconds.

[0023] The offset is converted into a trigger adjustment amplitude factor, whose value range is 0.05 to 0.15, and is used to correct the protection trigger threshold;

[0024] When the fusion signal has a trend-reverse fluctuation behavior, the fitting strategy automatically enters the hysteresis weight adjustment logic, so that the protection response maintains adjustable continuity within the stable response range and avoids frequent oscillating operations.

[0025] In a preferred embodiment, the error fitting unit further includes a trend segmentation tracking structure for isolating the influence of high-frequency disturbances. The behavior trend is analyzed by fitting residuals in each detection cycle. When the residual is less than a reference value of 0.03 for three consecutive cycles, the fitting trend conservative correction phase is entered. In this phase, no rapid threshold regression adjustment is performed, and the protection strategy is updated only based on the historical cumulative average error.

[0026] When the residual reverses direction twice within five cycles, the automatic output signal does not participate in the threshold correction and only maintains the current delay behavior.

[0027] In a preferred embodiment, the protection delay unit includes a behavior priority conversion structure. When the fused signal is classified as a critical fluctuation type, the protection execution time interval is actively delayed. The delay value is not less than five milliseconds and not more than twenty milliseconds. The specific delay time is obtained by linear mapping based on the adjustment amplitude factor output by the error fitting unit.

[0028] During the delay period, all signal outputs are marked as medium risk, and high-level interrupt control commands are prohibited from being sent; after the delay, if the regression value of the error fitting unit remains stable, the delay will continue; otherwise, the protection action call behavior in the fault response unit will be executed; the priority conversion structure ensures that before the abnormal state is accurately verified, the system response has a clear waiting mechanism to minimize the probability of false operation.

[0029] In a preferred embodiment, the protection delay unit generates a delay behavior state label after entering the delay state. The label includes the delay time interval, the current fusion signal fitting residual value, the amplitude change trend parameter and the fitting regression level information, and is embedded in the output signal stream in real time and transmitted to the fault response unit;

[0030] The delayed behavior status tag is automatically updated every two sampling cycles to assist the fault response unit in executing the response level determination process. When the fitted regression level in the status tag shows a deteriorating trend or the residual value continues to expand, the fault response unit loads the response determination conditions in advance and enters the candidate preparation state.

[0031] The delay behavior status tag also serves as one of the input fields for generating the fault event identification tag. Its content is written into the event identification structure after the fault response behavior is triggered, supporting subsequent external fault trend analysis and multi-tag clustering processes. Through the generation and dynamic update of the delay behavior status tag, a data sharing mechanism is established between the protection delay unit and the fault response unit, realizing feedforward judgment of the response path and cross-unit linkage.

[0032] In a preferred embodiment, the fault response unit introduces a delayed behavior status label as a preprocessing information source for response determination before executing the emergency response process. The fitting regression level and residual trend in the status label are used as one of the judgment factors, and the offset degree, duration, and fitting trend change between the current fusion signal and the reference model are combined to jointly determine the response behavior level.

[0033] The response behaviors are divided into three categories: interrupt output, signal bypass, and data freeze. If the residual value in the delay behavior state label increases continuously for five cycles, and the fusion signal deviation degree is greater than 15% for three consecutive cycles, an interrupt output is triggered.

[0034] If the tag trend does not deteriorate significantly but the fluctuation lasts less than ten milliseconds, the data freeze strategy is executed;

[0035] If the fitting trend shows nonlinear growth and the status label level is medium risk, the signal bypass logic is executed to redirect the current data flow to the redundant path output channel; this response mechanism introduces the delayed behavior status label as a linkage condition to improve the stability and reliability of the response trigger judgment.

[0036] In a preferred embodiment, the fault response unit generates a fault event identification tag when triggering a protection behavior. The tag structure includes a fault type code, a fluctuation behavior tag, a protection action record, fusion signal fitting trajectory information, and a delay behavior status tag field. It is encapsulated and output synchronously in a standard communication data frame format. Each event identification tag has a unique number for fault tracing and historical archiving. When similar fitting trends, similar delay behavior status fields, and different fluctuation starting points exist in multiple event identification tags, the fault trend clustering logic is triggered, and an early warning report is sent to the superior platform.

[0037] Technical effects and advantages of the present invention:

[0038] The present invention uses a fluctuation recognition unit to achieve real-time sampling and high-dimensional feature extraction of current signals during power switching. The extracted indicators, such as continuous change amplitude, rising edge slope, phase difference change, and frequency mutation value, are combined with predefined transient change recognition standards and data statistical rules to identify and classify abnormal behavior. This allows the system to identify the onset of fluctuations and distinguish between normal disturbances and potential anomalies. Compared to traditional judgment methods based solely on current amplitude changes, the present invention can more accurately perceive short-term disturbance states and accurately identify tolerable fluctuations, thus avoiding the misinterpretation of non-fault fluctuations during power switching as faults.

[0039] The present invention uses the backup sampling path startup mechanism of the redundant acquisition unit to compare, compensate and fuse the main path signal, ensuring that a time-continuous and highly reliable fusion signal can be constructed under abnormal interference conditions such as transient drift and inductive distortion of the main channel signal; at the same time, the error fitting unit performs dynamic regression analysis on the error offset trend between the fusion signal and the reference data model, and sets the protection trigger threshold and response time according to the fitting result, thereby effectively improving the data stability and judgment accuracy of the system before protection execution, and providing a true and reliable current state basis for subsequent control logic.

[0040] This invention utilizes a linkage mechanism between the protection delay unit and the fault response unit. When fluctuations are detected as acceptable, the system automatically delays the execution of protection instructions and temporarily suspends high-priority protection responses. Only when the fused signal fit trend deteriorates or the offset exceeds a preset limit does the fault response unit generate a high-level protection instruction and initiate a disconnection or interruption. This response process incorporates clear delay buffering, trend verification, and hierarchical protection logic, effectively reducing the probability of false triggering due to transient disturbances or sampling anomalies while ensuring timely fault response and stable system protection. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to facilitate understanding by those skilled in the art, the present invention will be further described below with reference to the accompanying drawings;

[0042] Figure 1 This is a schematic diagram of a current transformer intelligent control system in the present invention;

[0043] Figure 2 This is a working principle diagram of the fluctuation identification unit in the present invention;

[0044] Figure 3 This is a working principle diagram of the fault response unit in the present invention. DETAILED DESCRIPTION

[0045] The following will provide a clear and complete description of the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0046] Reference Figure 1-3 The following examples were obtained:

[0047] Example

[0048] A current transformer intelligent control system, comprising:

[0049] The fluctuation identification unit is used to acquire the current signal during the switching period through real-time sampling. Based on predefined instantaneous change identification criteria, it extracts the continuous change amplitude, rising edge slope, phase difference change, and frequency mutation value from the current signal. It then uses statistical data rules to complete the feature identification and classification of abnormal fluctuation behavior. This unit serves as the "sensing portal" for the entire system, and its goal is to accurately classify and identify abnormal current fluctuations that may occur during the backup power supply switching period. Traditional current transformer systems use only current amplitude as a judgment criterion, which is prone to false alarms or omissions during power switching periods. The fluctuation identification unit extracts richer signal features (such as slope, phase difference, and frequency mutation) to construct a multi-dimensional feature space, thereby accurately determining whether it is a normal switching disturbance or a true fault anomaly, providing a classification basis for all subsequent behaviors.

[0050] The redundant acquisition unit, after identifying the result as belonging to the tolerable switching fluctuation type, immediately triggers the backup sampling path and enables high-speed sampling instructions to obtain redundant current information flow, performs time axis comparison on the main path data, and constructs a fusion signal result on the premise of meeting the timing continuity to replace the distortion or drift segment in the main signal; after the fluctuation identification confirms that the current fluctuation is of the tolerable type, the system enters the data fault tolerance stage. The redundant acquisition unit is responsible for verifying and compensating the sampling signal of the main path. By starting the backup sampling path and performing high-speed data acquisition and timing comparison to construct a fusion signal, the signal distortion problem caused by phase drift, inductive saturation, external interference, etc. in the main channel is solved, ensuring that the system judgment is based on real and reliable data and improving the overall judgment robustness.

[0051] The error fitting unit continuously compares the fused signal with a preset reference data model to perform a regression fit of the error offset trend of the current fluctuation state. Based on the fitting results, it sets the protection trigger threshold and response time, and makes real-time corrections to the main control strategy. The dynamic changes in the fused signal reflect the evolution of the system state. The error fitting unit compares the results with the reference model to track the fluctuation offset trend in real time and perform mathematical fitting modeling. The fitting results are not only used to adjust the trigger conditions (thresholds, delays) of the protection action, but also play a key role in determining the stability trend of the system and whether it is deteriorating. They are the core logic hub that transforms data processing into control decisions.

[0052] When the protection delay unit determines that the fluctuation state is within an acceptable range, it automatically delays the execution of protection and outputs a fusion signal as an effective monitoring benchmark. At the same time, it suspends high-priority protection response instructions to prevent false operations due to transient errors and improve power continuity during switching. To avoid short-term fluctuations triggering unnecessary protection actions, the protection delay unit provides a protection response buffer mechanism based on intelligent judgment. Based on the output trend of the error fitting, the protection execution timing and strategy are dynamically adjusted, and the control response is delayed when the system can still tolerate it. The fusion signal is continuously used as a monitoring benchmark. This design significantly reduces false operations caused by transient disturbances and improves the system's anti-interference ability and control flexibility in actual operation.

[0053] The fault response unit immediately executes the emergency response process, generates a high-level protection instruction, interrupts the signal output process, and issues a fault event identifier to the external system for protection processing and logging of power switching control when the offset between the fusion signal and the reference data model exceeds the preset limit, or the redundant sampling path cannot complete effective time alignment. It is used for the protection of power switching control. When the system determines that the fluctuation deviates seriously from the trend or the compensation fails, it must be dealt with quickly. The fault response unit introduces a high-priority judgment mechanism to make protection decisions immediately when redundant data fails or the offset exceeds the limit, including output interruption, fault recording, external reporting, etc. This unit ensures that the current transformer can quickly protect the safety of key equipment and systems under abnormal circumstances, and is the "execution terminal" of the entire protection chain.

[0054] The fluctuation identification unit includes an abnormal current fluctuation screening structure. It is used to set a multi-threshold condition judgment mechanism based on the extraction of continuous change amplitude, rising edge slope, phase difference change and frequency mutation value. Different characteristic parameters are combined and calculated in a time-aligned manner to construct a high-dimensional characteristic behavior discrimination space. Through the set three-value hierarchical decision logic, the detection results are marked as stable fluctuation, critical fluctuation or severe abnormality.

[0055] In the time window slip judgment process, the first judgment factor is whether the fitting coefficient between the average amplitude change and the rising rate of three adjacent sampling periods is less than 0.1, and the second judgment factor is whether the instantaneous frequency change is within 5 Hz. When both conditions are met, the current state is determined to be a critical fluctuation.

[0056] The critical fluctuation type is identified as the tolerable switching fluctuation type, which is used as the basis for starting the subsequent backup sampling path, so that the working behavior of the redundant acquisition unit has a traceable judgment basis based on the feature recognition results, thereby ensuring the continuity of the technical chain between fluctuation judgment and data compensation.

[0057] The fluctuation identification unit is used to continuously sample the input current signal in real time during the backup power supply switching period and extract multiple fluctuation characteristic parameters, including: continuous change amplitude (i.e., the absolute mean of the difference in current values between adjacent sampling points, used to measure the intensity of the fluctuation), rising edge slope (indicates the growth rate of the current signal in the rising section, reflecting the instantaneous transition trend), phase difference change (indicates the change amplitude of the phase difference between the current signal and the reference clock or reference power supply within a certain period of time, used to determine whether the synchronization is unstable), and frequency mutation value (by analyzing the time interval between zero crossing points to reversely infer the instantaneous frequency change, used to identify whether the system has short-term switching or harmonic disturbances). After the above characteristic parameters are collected, the feature identification and classification attribution process of abnormal fluctuation behavior begins. This process makes judgments based on pre-set data statistical rules. The statistical rules may include multi-factor condition combination judgment, interval classification threshold setting, and feature trend fitting within a sliding window.

[0058] To improve recognition accuracy, a time window sliding judgment mechanism is introduced into the recognition process. This involves analyzing several characteristic parameters of the current signal within a dynamically sliding time interval. For example, within three consecutive sampling periods (each group can contain a fixed number of sampling points, such as 20 points), the average amplitude change is calculated, that is, the average amplitude change between the maximum and minimum current values within each sampling group. The rise rate is then calculated, that is, the slope of the fitted line between the current value and the time point during the rising phase of the signal. A fitting coefficient is then calculated for these two factors. The "fitting coefficient" here refers to the slope fit calculated using the least squares method and serves as a quantitative indicator of the degree of linear correlation between the two factors. If the fitting coefficient is less than 0.1 (indicating a weak linear correlation and no apparent growth trend in the fluctuation behavior), the first judgment factor is considered to have met the criteria.

[0059] The system then evaluates the instantaneous frequency change and determines the frequency change trend by measuring the timing difference of the zero-crossing points of the sampled signal. If the fluctuation frequency remains within a range of five hertz (for example, between 48.5Hz and 51.5Hz), the system's main frequency is considered to have not been significantly disturbed, and the second judgment factor is met. If both the first and second judgment factors are met, the system classifies the period as a critical fluctuation.

[0060] Critical fluctuation is defined as a current signal exhibiting a certain amplitude and fluctuation trend, but without exceeding the master control threshold or exhibiting signs of instability. This is a typical tolerable disturbance phenomenon, typically occurring when power is transferred from the main power source to the UPS backup path. This is caused by short-term signal disturbances such as relay activation and transient electromagnetic interference. Therefore, the system identifies critical fluctuations as tolerable switching fluctuations, which serve as the basis for determining whether to proceed with the subsequent data compensation and sampling switching processes.

[0061] When the fluctuation recognition unit outputs a critical fluctuation type, it immediately triggers the redundant acquisition unit's backup sampling path activation instruction, which then initiates high-speed sampling, ensuring that data compensation can be performed based on accurate judgment. This approach ensures that the redundant acquisition unit's working behavior has a clear and traceable feature recognition source, establishing a technical closed-loop chain from "feature acquisition → judgment and classification → data compensation", ensuring that there are no logical jumps or ambiguous judgments between different processing logics, and improving the reliability and robustness of the system's judgment.

[0062] The redundant acquisition unit establishes a data mirror relationship by running dual-path current signal channels in parallel. The error between the sampling period of the primary path and the sampling period of the backup path is set to no more than ten microseconds to ensure that the time alignment accuracy meets the data fusion requirements.

[0063] The backup path sampling values must undergo a preliminary noise removal process. The removal method includes one of three methods: maximum difference value method, root mean square limit method, and adaptive judgment based on sliding window filtering. The specific method is automatically switched based on the current data fluctuation type and is not fixed.

[0064] Under the condition that the sampling values meet the redundant timing structure, an interleaved data stream is established by delaying one cycle to ensure the continuity of the redundant data and that it does not cause logical interference with the main path. At the same time, all redundant data will be brought into the phase reconstruction operation process to participate in the construction of the error fitting unit behavior.

[0065] In order to ensure the stability and reliability of current monitoring data during power switching, the redundant acquisition unit uses a parallel operation mode of dual-path current signal channels to build a data mirror structure.

[0066] Definition and purpose of dual-path current signal channel: The so-called "dual-path channel" means that the main path and the backup path independently receive the same current source output signal, and sample, store and process it through their respective sampling units. Among them:

[0067] The main path is responsible for routine sampling, that is, the system standard sampling process, with a moderate response speed, and is used for normal operation data output.

[0068] The backup path is activated after identifying the "tolerable switching fluctuation type" and enters the high-speed sampling state. Its task is to perform higher frequency and higher precision compensation acquisition of the detailed behavior during the fluctuation period.

[0069] To ensure that the sampled data between the primary and backup paths can be fused, strict time alignment accuracy requirements must be met. In this invention, the error between the primary and backup path sampling periods is controlled to no more than ten microseconds. That is, the time sampling interval difference between the two channels must not exceed 10 μs. This setting ensures that during the subsequent fusion phase, the redundant data can be compared one-to-one with the primary data on the time axis, preventing data misalignment, interleaving, or redundancy overflow, and ensuring the accuracy of the fused signal construction.

[0070] Definition of noise rejection mechanisms and methods: To improve the validity of the backup path sampling data, the sampled values must undergo preliminary noise rejection before entering the data fusion process. Noise rejection refers to eliminating non-real current change data introduced by sensors, electromagnetic interference, or sampling bias. Noise rejection methods include the following three processing mechanisms:

[0071] Maximum difference value method: Based on the difference between adjacent sampling points, a rejection threshold is set. Any sample with a difference greater than the threshold is considered an outlier and is rejected.

[0072] Root mean square limiting method: Calculate the root mean square value (RMS) of the sampling sequence, then set the upper and lower limit amplitude ranges. Sampling points outside the range are regarded as noise.

[0073] Adaptive judgment based on sliding window filtering: Within the sliding time window (such as 5-10 o'clock), the sample mean and standard deviation are dynamically calculated, and an adaptive algorithm is used to determine whether the sampling point deviates from the statistical central value, and then determine whether to remove it.

[0074] The three methods are not fixed in one particular case. Instead, the filtering method is automatically switched based on the current data fluctuation type (i.e., the level of fluctuation characteristics output by the fluctuation identification unit, such as critical fluctuation or stable fluctuation). For example, in scenarios with high fluctuation intensity, a sliding window filtering method may be preferred; whereas in scenarios with periodic disturbances, a maximum difference method may be more appropriate. This mechanism enhances the flexibility and adaptability of noise processing to different scenarios, avoiding "over-removal" or "incorrect retention."

[0075] Interleaved Data Stream Construction and Fusion Mechanism: After the backup path sampling values have been noise-removed and deemed valid, the next step is the redundant timing structure comparison phase. In this phase, a one-cycle delay is used to establish the interleaved data structure. Specifically, the backup path sampling points lag slightly behind the primary path by one cycle (for example, if the sampling period is 10μs, the backup path is delayed by a 10μs period), forming a cross-timing structure.

[0076] This method has two effects:

[0077] Avoid complete overlap of processing logic between the primary and backup paths, which could lead to resource conflicts.

[0078] When signal distortion occurs, the backup path can provide front and back redundant points for interpolation compensation, thereby improving the stability of the fusion result.

[0079] The output structure of the interleaved data stream is fed into the next step of the fusion signal construction process. During this process, the sampling sequence in the backup path not only serves as data compensation but also serves as a reference for comparing the sampling accuracy and integrity of the primary path, ensuring the temporal continuity and reliability of the data fusion.

[0080] Phase reconstruction process description: Redundant data is also uniformly incorporated into the phase reconstruction process. The core of this process is to analyze the relative phase drift of the backup sampling values, perform fitting corrections using a reference phase model (which can be derived from a historical stable period or reference frequency), and output a precise reference signal that can be used in subsequent error fitting units. Phase reconstruction is designed to address data errors in the primary path caused by power supply phase drift during switching, ensuring that the final fused signal is consistent in time, amplitude, and phase, providing a reliable and reliable current state input for subsequent protection actions.

[0081] The error fitting unit includes a dynamic regression correction structure, in which the regression behavior is analyzed based on the mean deviation trend of the fusion signal from the reference model within the current five sampling periods, while considering the consistency of the deviation direction and the stability of the duration;

[0082] The regression strategy adopts one of two methods: interval minimum variance fitting method and exponential weighted moving average method. The former is used when the current fluctuation intensity is greater than 10%, and the latter is used when the fluctuation duration is less than 10 milliseconds.

[0083] The offset is converted into a trigger adjustment amplitude factor, whose value range is 0.05 to 0.15, and is used to correct the protection trigger threshold;

[0084] When the fusion signal has a trend-reverse fluctuation behavior, the fitting strategy automatically enters the hysteresis weight adjustment logic, so that the protection response maintains adjustable continuity within the stable response range and avoids frequent oscillating operations.

[0085] The error fitting unit is responsible for comparing the fused signals generated by the redundant acquisition units with the system's preset reference data model in real time. Using trend analysis and dynamic regression techniques, it determines the degree of fluctuation deviation and adaptively adjusts the protection strategy. To achieve this, the error fitting unit incorporates a dynamic regression correction structure. Its core function is to calculate, through regression modeling, whether the current fluctuation exhibits a continuous deviation and adjust the protection control parameters accordingly.

[0086] Definition and purpose of the dynamic regression correction structure: This structure uses the fused signal as the input data source. Within the current five sampling periods (i.e., the time window consisting of the last five equally spaced sampling points), it compares each of the five expected current values in the system reference model and calculates their mean shift trend to identify whether the current signal is experiencing a systematic shift. Determining the shift trend depends not only on numerical differences but also on satisfying two trend conditions:

[0087] Offset direction consistency: that is, the offset values within multiple sampling periods should have the same change direction (continuously rising or falling) to avoid occasional fluctuations being misinterpreted as continuous offsets;

[0088] Duration stability: This means the trend must persist for more than two full sampling periods (for example, at least three consecutive periods out of five must show the same direction of change) to confirm statistical persistence. Once these conditions are met, the specific regression strategy selection process begins.

[0089] Regression strategy selection logic and algorithm description: After the error fitting unit passes the trend judgment, it will select one of the following two regression strategies for calculation based on the specific characteristics of the current fluctuation:

[0090] Interval minimum variance fitting method: This method is suitable for situations where current fluctuations are greater than 10 percent, meaning that the maximum deviation of the fused signal from the reference model exceeds 10% of its rated value. By performing a least-squares fit on the current five-cycle sampling values, a trend regression line is generated to reflect the current fluctuation direction and rate, enabling rapid response to large fluctuations.

[0091] Exponentially Weighted Moving Average: This method is used to smooth fluctuations that occur quickly but briefly, often less than ten milliseconds. This method weights recent samples more heavily to smooth out sudden changes, preventing misjudgments caused by short-term data anomalies.

[0092] Once the regression strategy is selected, the current fitted offset will be output. Interval minimum variance fitting is a trend modeling method based on the minimum residual sum of squares. Its core goal is to find a straight line or smooth curve within a fixed time window that minimizes the sum of squared errors between the curve and the fused signal. The data processing flow is as follows: the fused signal values of the current five consecutive sampling periods are arranged in chronological order; the corresponding reference model values are used as a comparison baseline; with time as the independent variable and the current difference as the dependent variable, a trend line is found using the principle of minimum sum of squared errors; the slope of this trend line represents the direction and speed of the current signal offset.

[0093] The exponentially weighted moving average (EWMA) is a strategy that emphasizes the influence of recent data and smoothes sudden disturbances. The basic idea is that newer data points are weighted more heavily in the average calculation, thereby enhancing the system's real-time sensitivity to the current state while suppressing occasional anomalies. The data processing process involves extracting the fused signal values within the current five sampling periods; setting an initial weighting rule (for example, the last period has the highest weight, decreasing progressively); summing all data points according to their weights and subtracting this from the reference model average; and finally calculating the smoothing offset for the current window.

[0094] The offset is converted into a control parameter, the trigger adjustment amplitude factor. The offset in the fitting result is the mean difference between the current fused signal and the reference model. After the trend is determined to be established, the offset is converted into a control parameter, for example, through a preset mapping relationship, called the trigger adjustment amplitude factor. The value range of this factor is limited to 0.05 to 0.15. The larger the value, the closer the current system is to the critical value requiring response.

[0095] This factor is used to adjust the protection trigger threshold in the protection delay unit. For example, when the fitted offset is positive and exceeds the upper limit, the original protection current threshold is adjusted downward by a certain percentage to speed up system response. If the offset is negative, the protection response is delayed to improve system tolerance. This dynamic control method allows the system to better adapt to the stability requirements of different loads or switching conditions.

[0096] Trend reversal and hysteresis weight logic: In actual operation, the fusion signal fitting trend may reverse direction, that is, the fitting curve changes from rising to falling or vice versa, indicating that the system state has changed unstablely. To avoid frequent switching and control oscillation, the error fitting unit will automatically start the hysteresis weight adjustment logic in this case:

[0097] In the weight adjustment logic, a hysteresis factor within an additional time window is applied to the fitting algorithm; the values of recent cycles are assigned lower weights, and the historical stable cycle values dominate; through the preset hysteresis smoothing strategy, the protection response is kept within the stable response range, which can effectively avoid the oscillation behavior caused by frequent switching protection logic and improve the continuity and stability of the current transformer system response in complex transient scenarios.

[0098] The error fitting unit also includes a trend segmentation tracking structure to isolate the impact of high-frequency disturbances. It analyzes behavioral trends through fitting residuals within each detection cycle. When the residual is less than the baseline value of 0.03 for three consecutive cycles, it enters the conservative correction phase of the fitting trend. In this phase, no rapid threshold regression adjustment is performed, and the protection strategy is updated based solely on the historical cumulative average error.

[0099] When the residual reverses direction twice within five cycles, the automatic output signal does not participate in the threshold correction and only maintains the current delay behavior.

[0100] In this embodiment of the present invention, the error fitting unit includes not only a dynamic regression correction structure but also a segmented trend tracking structure. This structure is used to enhance the ability to distinguish short-term disturbances from true trends under complex or unstable current conditions, and in particular to suppress and adjust strategies for residual fluctuations caused by high-frequency disturbances.

[0101] "Trend segmented tracking structure" means: within a fixed monitoring window, the fitting residuals between the fusion signal and the reference data model are statistically analyzed on a cycle-by-cycle basis, and the stability of the signal fluctuation behavior is judged based on the continuity and change direction of the residuals, and then a decision is made whether to perform regression adjustment or enter the conservative stage of the protection strategy.

[0102] Fitting residual: refers to the difference between the fusion signal and the regression fitting curve at each sampling point, which is used to measure the degree to which the current data point deviates from the trend line;

[0103] High-frequency disturbances: These are small oscillation signals that change rapidly over a short period of time. These disturbances are usually caused by electromagnetic interference, relay bounce, power switching edge states, etc. They are unrelated to the actual changes in system current.

[0104] Detection cycle: It is the sliding analysis cycle defined by the system, usually a number of consecutive sampling cycles (for example, each cycle contains 5-10 groups of sampling points);

[0105] Behavioral trend: refers to the overall deviation direction determined based on the change in the direction of continuous residuals.

[0106] During each test cycle, the system normalizes the fitting residuals at all sampling points, calculates their absolute value, and compares it with a fixed reference value. This reference value is set to 0.03, indicating that residuals within ±0.03 current units (such as amperes) are considered "stable."

[0107] If the fitting residual is less than this baseline value for three consecutive periods, the current fusion signal fluctuates minimally. The system then determines that the current state has a stable trend and no sudden anomalies exist. The system automatically enters the conservative correction phase for the fitting trend. This conservative correction phase is characterized by the fact that the error fitting unit no longer performs fast fitting regression, meaning it does not recalculate the trend line slope or reevaluate the threshold response factor. Instead, it uses only the historical cumulative average error for control decisions.

[0108] "Historical cumulative average error" means taking a weighted average of the residual means over multiple previous normal cycles and using it as a control reference. This approach is more conservative and more tolerant than fast fitting responses, making it suitable for improving system robustness in stable scenarios.

[0109] In another case, the trend segment tracking structure also focuses on the direction of change of the fitting residual, that is, whether the residual is increasing positively or decreasing negatively. The system identifies whether the trend change has reversed by comparing period by period: when the fitting residual reverses direction more than twice within five consecutive detection periods (for example, the residual changes from positive to negative, and then from negative to positive); the system determines that: the current fluctuation trend is uncertain, oscillatory, or affected by high-frequency interference; in order to avoid triggering erroneous protection actions when the trend is unclear, the system does not perform any threshold correction behavior. In this case, the system will automatically maintain the current protection delay behavior and will not increase or decrease the protection level. This method can prevent frequent switching of protection strategies and avoid protection misoperation or protection strategy fatigue caused by jitter.

[0110] Suppose, during a backup power supply switchover, the system's fused signal fluctuates slightly for five consecutive cycles, with the residual remaining less than 0.02. In this case, the system enters a conservative correction phase, relying solely on historical averages and eliminating the need for recalculation. In another scenario, the system's residual fluctuates repeatedly from "rising, falling, and rising" over five cycles. Although the residual amplitude remains within the specified range, the system halts threshold updates due to two reversals. Instead, it maintains the original delay, waiting for a clearer trend before deciding on a response.

[0111] The protection delay unit includes a behavior priority conversion structure. When the fused signal is classified as a critical fluctuation type, it actively delays the protection execution time interval. The delay value is not less than 5 milliseconds and not more than 20 milliseconds. The specific delay time is obtained by linear mapping based on the adjustment amplitude factor output by the error fitting unit.

[0112] During the delay period, all signal outputs are marked as medium risk, and high-level interrupt control commands are prohibited from being sent; after the delay, if the regression value of the error fitting unit remains stable, the delay will continue; otherwise, the protection action call behavior in the fault response unit will be executed; the priority conversion structure ensures that before the abnormal state is accurately verified, the system response has a clear waiting mechanism to minimize the probability of false operation.

[0113] In an embodiment of the present invention, the protection delay unit is provided with a behavior priority conversion structure, which is used to perform intelligent judgment and hierarchical control on the execution timing and response level of the protection action according to the severity of the current current fluctuation state and the system identification level, thereby avoiding unnecessary protection responses due to misjudgment or transient disturbances.

[0114] The "behavior priority conversion structure" is a control mechanism within the current transformer control logic that prioritizes protection responses based on the system's current identification results. This priority logic then determines whether and when protection actions should be executed. Its design aims to introduce a response buffer strategy, enabling the system to proactively delay response when judgments are insufficient or the state is not yet stable, thereby preventing false trips caused by brief anomalies or false signals.

[0115] The fused signal is classified as a critical fluctuation response logic. In actual operation, the fused signal is the output of the current data of the primary and backup paths after timing comparison, error compensation, and phase calibration. This signal is the most reliable reflection of the current status of the system.

[0116] If the fusion signal is judged by the fluctuation recognition unit to belong to the "critical fluctuation type", that is, the fluctuation amplitude is within the tolerance range, the frequency change has no drastic mutation, and the fitting trend has not deteriorated, but there is still potential for change, the system will activate the behavior priority conversion structure and enter the response delay process.

[0117] Delay mechanism and control logic for protection execution time: When the behavior priority conversion structure is activated, the system will immediately and proactively delay the protection execution time interval, that is, delay the execution of the protection instructions originally planned to be executed in the current cycle for a period of time to observe whether there is a trend of deterioration.

[0118] The time delay range is set: for example, the delay time should be no less than five milliseconds and no more than twenty milliseconds. The specific delay time is linearly calculated based on the adjustment amplitude factor output by the error fitting unit. The so-called "adjustment amplitude factor" is a dynamic coefficient generated by the error fitting unit based on the deviation between the current fused signal and the reference model and the trend fitting results. The range is usually set between 0.05 and 0.15. The delay time is proportional to the factor: the smaller the factor, the shorter the delay; the larger the factor, the higher the trend risk and the longer the delay time. This mapping method ensures the system's adaptive response to fluctuations of varying severity.

[0119] Medium-risk marking and command suppression mechanism: During the delay phase, to ensure that the system's external control devices and upper-level systems do not misunderstand the current status, the behavioral priority conversion structure temporarily marks the output status: all signal outputs are marked as "medium risk", which informs the external system that the current status has not been confirmed but requires attention; during this phase, the system is prohibited from sending high-level interrupt control commands, such as circuit breaker tripping and master device switching; this suppression measure prevents external execution of erroneous actions, while preserving the system's window for continued observation and correction.

[0120] The judgment behavior after the delay expires is connected with the fault response: After the delay time interval ends, the system makes a judgment based on the latest output result of the error fitting unit: If the fitting regression value remains stable, that is, the residual change has not significantly expanded, and the fluctuation direction has not continued to deteriorate; the delay state is automatically renewed, the delay time is reset, and the observation state is entered again; if the fitting value shows a trend of deterioration or the offset exceeds the limit, it is considered that the abnormality has been confirmed, the system exits the delay state, and immediately enters the fault response unit execution stage; the fault response unit will call the corresponding protection action according to the event level, such as output interruption, signal bypass or data freeze.

[0121] Through this judgment logic, the system realizes dynamic identification of "recoverable state" and "response-required state", thereby postponing the response strategy until sufficient judgment basis is established, reducing the probability of false response and improving the accuracy of system judgment.

[0122] For example, during a UPS power switching operation, the system's integrated signal detected a moderate current jump within two cycles, which the fluctuation identification unit identified as a critical fluctuation. The error fitting unit output an adjustment factor of 0.10, corresponding to a 15-millisecond delay. The system entered a delay state, disabling the power-off command and marking the status output as medium risk. During the delay period, the fitted trend remained unchanged, so the system automatically extended the delay by 5 milliseconds. After reassessment, the residual increased, indicating an increased risk. The system then exited the delay state and entered the fault response logic, executing signal bypass.

[0123] After entering the delayed state, the protection delay unit generates a delay behavior state label. This label contains the delay time interval, the current fusion signal fitting residual value, the amplitude change trend parameter, and the fitting regression level information. The label is embedded in the output signal stream in real time and transmitted to the fault response unit.

[0124] The delayed behavior status tag is automatically updated every two sampling cycles to assist the fault response unit in executing the response level determination process. When the fitted regression level in the status tag shows a deteriorating trend or the residual value continues to expand, the fault response unit loads the response determination conditions in advance and enters the candidate preparation state.

[0125] The delay behavior status tag also serves as one of the input fields for generating the fault event identification tag. Its content is written into the event identification structure after the fault response behavior is triggered, supporting subsequent external fault trend analysis and multi-tag clustering processes. Through the generation and dynamic update of the delay behavior status tag, a data sharing mechanism is established between the protection delay unit and the fault response unit, realizing feedforward judgment of the response path and cross-unit linkage.

[0126] In a preferred embodiment of the present invention, after entering the delay state, the protection delay unit generates a structured data description, called a delay behavior state tag, to support system tracking of current delay behavior and decision optimization. This tag, a dynamically updated state identifier, provides a quantitative description of the current state to other functional units in the system during the protection delay phase, serving as an auxiliary basis for subsequent fault diagnosis and response triggering.

[0127] The "delay behavior status label" refers to a data set containing key judgment parameters generated during the delay response phase. This label mainly includes the following four fields:

[0128] Delay time interval: refers to the cumulative time that the current delay state has been running since it was started, as well as the maximum allowable delay time range set by the system, in milliseconds;

[0129] Current fusion signal fitting residual value: the current error value between the fusion signal and the regression prediction trend, which is used to reflect whether the current signal state is stable;

[0130] Amplitude change trend parameter: reflects the change direction and rate of the current fusion signal in continuous cycles, such as whether it is in an increasing trend or whether there is an accelerated rise phenomenon;

[0131] Fitting regression level information: The regression trend level given by the error fitting unit after evaluation can usually be divided into three levels: "stable", "weak change", and "deterioration", corresponding to different degrees of response recommendations.

[0132] These parameter fields together constitute a complete tag information. Each tag is time-sensitive and only represents the system status within the current two sampling periods.

[0133] Real-time tag embedding and cross-unit transmission: The tag automatically refreshes every two sampling cycles, re-evaluating and updating the field value every two minimum detection cycles to ensure the system has the latest status snapshot. Tag data is embedded in the output signal stream via an internal communication link, appending the tag information to the main system's data output channel for transmission to the fault response unit.

[0134] Through this mechanism, the system realizes data linkage between the protection delay unit and the fault response unit, so that the response behavior not only depends on the final judgment conclusion, but also has the ability to perceive the "current delay state evolution trend" and has the feedforward control capability of the response path.

[0135] Tags assist the fault response unit: After receiving a tag, the fault response unit extracts its field content to assist in determining the response level. The response level refers to the intensity of the response action selected by the system based on the severity of the fault. Common options include signal bypass, data freeze, and forced disconnect.

[0136] The following fields in the tag have direct triggering significance: when the fitted regression level shows a deteriorating trend, meaning the system has identified an increasing trend slope for consecutive periods, indicating that the fault is worsening; or when the residual value continues to expand, meaning the fused signal deviates further and further from the reference model, indicating that the error evolution is risky; if any of these conditions are met, the fault response unit no longer waits for the delay state to end, but instead preemptively loads the response judgment conditions, elevating the current state to a "candidate response state" and preparing to enter the formal response process. This mechanism improves the system's agility in responding to critical conditions, enabling the system to move beyond passive waiting and instead enable proactive reasoning and judgment.

[0137] The role of tags in the fault event identification structure: Delay behavior state tags not only serve as a reference for judgment but also as a component field in the fault event identification tag generation process. When the system finally enters the fault response process, the generated event identification structure will contain complete response context information, including: the current trigger action type; system runtime reference model information; fitted trend evolution path; and historical content records of delay behavior state tags. This information is output as a data encapsulation of a single event, supporting subsequent system-level fault trend analysis and tag clustering comparison for purposes such as identifying high-frequency fault sources and optimizing prediction models.

[0138] For example, during a backup power supply switchover, the system detected a critical medium-amplitude fluctuation in the current signal, prompting the protection delay unit to enter a delay state with an initial delay time of 15 milliseconds. After the label is generated, the following information appears: Delay time interval: 0–15ms; Current residual: 0.045; Amplitude trend: Increasing; Fit level: Weak change. Two cycles later, the following information is updated: Delay time interval: 0–17ms; Current residual: 0.061; Fit level changes to: Deteriorating. At this point, the fault response unit loads the response judgment conditions, pre-evaluates the risk as high, and marks it as a candidate response state. Subsequently, at 18 milliseconds, the system jumps to the response phase, executes signal bypass, and generates an event label containing the aforementioned delay label field for background system cluster analysis.

[0139] Before executing the emergency response process, the fault response unit introduces a delayed behavior status label as a preprocessing information source for response judgment. The fitting regression level and residual trend in the status label are used as one of the judgment factors. The offset between the current fusion signal and the reference model, the duration, and the fitting trend change are combined to determine the response behavior level.

[0140] The response behaviors are divided into three categories: interrupt output, signal bypass, and data freeze. If the residual value in the delay behavior state label increases continuously for five cycles, and the fusion signal deviation degree is greater than 15% for three consecutive cycles, an interrupt output is triggered.

[0141] If the tag trend does not deteriorate significantly but the fluctuation lasts less than ten milliseconds, the data freeze strategy is executed;

[0142] If the fitting trend shows nonlinear growth and the status label level is medium risk, the signal bypass logic is executed to redirect the current data flow to the redundant path output channel; this response mechanism introduces the delayed behavior status label as a linkage condition to improve the stability and reliability of the response trigger judgment.

[0143] In this embodiment of the present invention, the fault response unit is the core structure responsible for determining the final protection strategy execution method when an abnormal event occurs during current transformer monitoring. To improve judgment accuracy and avoid erroneous protection responses caused by single data errors, the fault response unit first introduces delayed behavior status tags as a preprocessing information source for response determination before executing the emergency response process.

[0144] The delay behavior status tag is a structured status data packet generated by the protection delay unit during the response delay phase. It contains multiple fields that reflect the current signal status and trend characteristics. Key fields specifically used for fault response prejudgment include: Fitting regression level: This indicates the classification of the current error fitting trend, generally set to "stable," "medium risk," and "deteriorating," reflecting whether the system trend is accelerating; Residual trend: This indicates whether the error between the fused signal and the reference model continues to increase over the current few cycles. This information provides the fault response unit with external input to determine whether the system is "in the process of deterioration." Combined with the signal data collected by the unit, it forms a multi-source joint judgment basis.

[0145] The "fusion signal" is a high-reliability output data stream constructed by combining the sampling values of the main path and the high-speed sampling values of the backup path, after time synchronization, phase calibration, and distortion elimination in the error fitting unit. It is used to reflect the current status in real time.

[0146] When determining whether to execute a response action, the system not only considers the delay behavior status label but also comprehensively analyzes the following: the degree of offset between the fused signal and the reference model: the percentage change in the current signal relative to the normal value over multiple cycles; the duration of the offset: whether the offset state persists for a certain period to avoid transient anomalies triggering protection; and the fitted trend change: the rate of change of the current error trend, indicating whether it is linear growth, nonlinear fluctuation, or stabilization. These factors, combined with the label field, form a multi-factor decision system, improving the stability and confidence of the response action determination.

[0147] Response behavior classification and execution conditions: In the present invention, response behaviors are divided into the following three categories:

[0148] Interrupt output definition: Immediately interrupt the current signal output, usually corresponding to the strongest protection level; Trigger condition: The residual value in the delayed behavior state label increases continuously for five cycles; at the same time, the fusion signal offset is greater than 15% for three consecutive cycles; Significance: It is judged that the fault trend is clear and worsening, and the system safety must be protected immediately.

[0149] Data freeze definition: Pauses data refresh, maintains the current output value, and prevents further intervention; trigger condition: No significant deterioration in the tag trend (fitting level is stable or medium risk); at the same time, the fluctuation duration is less than ten milliseconds, which is judged as a short-term disturbance; significance: If it is judged as short-term jitter or pseudo-fluctuation, there is no need to perform disconnection or bypass, only freeze and observe.

[0150] Signal bypass definition: Switching the output path, shutting off the primary path's output signal and using the backup path's data stream as the primary output. Triggering conditions: The fitted trend exhibits nonlinear growth, i.e., an exponential, curvilinear acceleration of the error slope; and the status label is rated medium risk. Significance: The primary path is considered abnormal, but the backup path data remains reliable, allowing path switching to avoid impacting the overall control strategy. Through this classification, the system implements a tiered protection system, from low to high intensity, ensuring appropriate response strategies in a variety of complex situations.

[0151] Suppose the system detects a fusion signal fluctuation during a power switching process, causing the protection delay unit to enter a delayed state and generate a status label. The current cycle label shows: Fit Level = Medium Risk, Residual Trend = Increasing; after the fifth cycle, the residual continues to increase; and the fusion signal offset is 18%. The system compares these conditions and meets the output interrupt triggering conditions. It immediately enters the emergency response process, executes an output interrupt, and shuts off the main output. In another scenario, the residual value in the label is small, the fit level is stable, and the duration is only 6ms. The system then determines that the disturbance is minor and only freezes the data, without taking any strong protection action.

[0152] When triggering protection behavior, the fault response unit generates a fault event identification tag. The tag structure includes fault type code, fluctuation behavior tag, protection action record, fusion signal fitting trajectory information and delay behavior status tag fields. It is encapsulated and output synchronously in the standard communication data frame format. Each event identification tag has a unique number for fault tracing and historical archiving. When similar fitting trends, similar delay behavior status fields and different fluctuation starting points exist in multiple event identification tags, the fault trend clustering logic is triggered and an early warning report is sent to the superior platform.

[0153] In this embodiment of the present invention, to achieve structured recording of fault behaviors, response traceability, and historical archiving, the fault response unit generates a complete, standardized data structure called a fault event identification tag when triggering a protection action. This tag is constructed simultaneously with the execution of the protection action, encapsulated in a standard communication data frame format, and synchronously output for inter-system sharing, platform recording, and monitoring by higher-level platforms.

[0154] In a preferred embodiment of the present invention, the offset between the fused signal and the reference data model is calculated using a percentage-based relative offset. The system sets a 10% warning zone boundary and a 15% fault response threshold. This means that if the fused signal offset exceeds 15% of its reference model value for three consecutive cycles, the system deems the offset to have exceeded the preset limit and triggers the fault response unit to generate a high-level protection instruction.

[0155] The fault event identification tag contains the following key fields, each of which has a clear technical function:

[0156] Fault type coding: The system assigns a classification code to each event based on the fault behavior category; Function: Quickly identifies types such as current overload, phase misalignment, sudden disturbance, and logic jump; Usage: Used for classification queries, statistical reports, and label clustering basic indexes.

[0157] Fluctuation behavior label: reflects the behavioral characteristics of the fused signal during the identification phase before the current event; includes: fluctuation amplitude range, duration classification, and slope level; function: supports determining whether the fault is triggered by the accumulation of continuous minor fluctuations or caused by a one-time drastic jump.

[0158] Protection action record: record of the response type and action level ultimately executed by the system; scope includes: interrupt output, signal bypass, data freeze; record content: response level, execution time, action duration.

[0159] Fusion signal fitting trajectory information: residual trend curve and regression model data summary output by the error fitting unit; usually adopts a simplified structure, such as slope value sequence, regression level transfer sequence, residual extreme value, etc.; function: provides quantitative indicators for fault trend identification, supports behavior analysis and model training.

[0160] Delay behavior status label field: protects the complete content of the label generated by the delay unit in the delayed response phase; includes: delay time, status level, trend evaluation value; function: reveals whether the system experienced a waiting period before the event and whether there is latent evolution of the trend.

[0161] At the same moment a protection action occurs, the fault response unit encapsulates all of the aforementioned fields into a standard communication data frame format, typically consisting of a header, tag body fields, a check digit, and a trailer. This ensures consistent identification and compatibility across the system bus or network. The tags are synchronously output to: a higher-level platform (such as a data center centralized control system); a local event logging module; and a fault history archive database. Each event identification tag is uniquely numbered, generated using a combination of a timestamp, a fault type bitcode, and a system node number, enabling full lifecycle fault tracing.

[0162] Fault Trend Clustering Logic: As the system continues to operate, it accumulates numerous fault event identification tags. To improve fault early warning capabilities, this paper designs a fault trend clustering logic to identify potential "repetitive but hidden" fault types in the system.

[0163] Clustering is triggered when the system detects the following three similarity features in multiple event identification tags: Similar fitting trend: There is trend slope and residual path consistency between the fusion signal and the fitting trajectory information in the tag; Similar delay behavior state field: The delay tag content in the tag (such as delay interval, level mark, trend direction) is statistically clustered; Different fluctuation starting points: Multiple tags are generated at different time points, different load conditions or power supply states, indicating non-accidental consistency.

[0164] Once the cluster trigger conditions are met, the system will deem the existence of a potential systemic risk evolution pattern and then send an early warning report to the superior platform (such as the data center monitoring and management system); the report includes: cluster label number, behavioral feature summary, historical distribution curve, recommended response strategy, etc.; it supports operation and maintenance personnel to conduct preventive inspections, model revisions or system structure optimization.

[0165] Example: Over three days, the system recorded seven event identification tags, five of which met the following characteristics: the fitted trend slopes were all positively increasing; the residual values continued to rise during the delay phase; the status level in the delay tags was all "medium risk"; and these events occurred at different time periods and during peak server loads. The system determined that this cluster had consistent characteristics, triggered the clustering logic, and generated an early warning report to the higher-level platform, indicating possible uneven current distribution between systems or abnormal power switching strategies, and recommended investigation.

[0166] It should be noted that: in order to facilitate understanding of the structural composition and response process of the current transformer intelligent control system proposed in the present invention, some control terms and system behavior instructions involved in the specification are defined and technically explained to ensure the consistency of the system structure and the collaborative logic between the functional units are open and clear.

[0167] "High-speed sampling instruction" refers to the trigger instruction issued by the fluctuation identification unit to the redundant acquisition unit after the system identifies that the current state is in a critical fluctuation type. It is used to start the backup sampling path and increase the sampling frequency to a higher than the set sampling rate of the main path, thereby achieving detailed capture of the current state during the fluctuation period and redundant verification. The triggering basis of this instruction is that the fluctuation identification unit's judgment result on the fusion signal characteristics meets the following conditions: the fluctuation intensity is within the system tolerance threshold, the frequency has not undergone serious deviation, and the trend has not shown continuous acceleration deviation. Although this type of fluctuation is not an immediate fault, it has the possibility of disturbance, so it is necessary to start the backup sampling to obtain higher-precision data. The high-speed sampling instruction has a one-time, single-cycle triggering feature. The sampling behavior is released after being completed on demand by the sampling drive logic in the redundant acquisition unit, and does not constitute continuous resource occupation.

[0168] A "high-priority protection response instruction" refers to a type of potential control instruction that can be invoked but not executed during the protection delay period, pre-set based on the initial assessment of the error fitting unit and the delay judgment mechanism when the system identifies a potential fault risk. This instruction is essentially a protective action response capability, such as current interruption, output disconnection, or trip signal transmission. However, until the fused signal state has completely deteriorated and the fitting trend has not formed a clear abnormality, it is locked and delayed by the behavioral priority conversion structure in the protection delay unit. The behavioral characteristics of this instruction are: "It has the conditions for execution preparation but is suppressed by the system strategy logic." Until the delay state is released or the response conditions are upgraded, the fault response unit will take over and determine whether to convert it into a formal execution instruction (i.e., a high-level protection instruction). By pre-setting and suppressing high-priority instructions, the system establishes a response "waiting zone" for protection actions, effectively preventing false operations when data judgment is insufficient, and improving the system's response accuracy.

[0169] "High-level protection instructions" are execution-type control commands that are ultimately generated and sent by the fault response unit when the system determines that the current current state has reached an intolerable level during the fault response phase. Its corresponding response behaviors include but are not limited to: signal interruption output (termination of the current control output signal flow); signal bypass (switching to the backup path output); data freeze (maintaining the current output and pausing updates), etc. The triggering conditions for this instruction are usually that multiple independent dimensions reach the execution threshold, including but not limited to: the residual value in the delay behavior state label continues to grow and exceeds the set cycle threshold; the fitting trend shows nonlinear accelerated changes; the offset amplitude between the fusion signal and the reference model is greater than the set tolerance. Once the high-level protection instruction is executed, it usually has the characteristics of irreversibility, system enforcement and cross-module synchronous response. Its response result is embedded in the event identification tag as part of the fault event record content and output, and is transmitted to the upper-level platform or storage system to form a closed loop of system protection behavior.

[0170] To ensure that these control signals are not isolated, the system's structural design incorporates the following sequential logic: The fluctuation identification unit identifies a critical fluctuation, triggers a high-speed sampling instruction, and the redundant acquisition unit samples the fused signal. The fused signal is analyzed by the error fitting unit. If the trend does not deteriorate, a high-priority protection response instruction is generated, locked by a delay mechanism. If the trend escalates or the delay fails, the fault response unit generates a high-level protection instruction, forcing the corresponding response action. This structure ensures a complete closed-loop response path from identification to suppression to confirmation to execution. Each of the three control signals has a trigger source, execution target, and system interface location, forming a control model with distinct functions, mutually exclusive policies, and non-conflicting behaviors.

[0171] The above formulas are all dimensionless and numerical calculations. The formulas are obtained by collecting a large amount of data and performing software simulation to obtain the most recent real situation. The preset parameters in the formulas are set by technicians in this field according to actual conditions.

[0172] It should be understood that in the various embodiments of the present application, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0173] Those skilled in the art will appreciate that the modules and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0174] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and modules described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0175] The above are only specific embodiments of the present application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

Claims

1. A current transformer intelligent control system, characterized in that: include: The fluctuation identification unit is used to obtain the current signal during the switching period through real-time sampling. Based on predefined instantaneous change identification criteria, it extracts the continuous change amplitude, rising edge slope, phase difference change, and frequency mutation value from the current signal, and completes the feature identification and classification of abnormal fluctuation behavior according to data statistical rules; The redundant acquisition unit, upon identifying the tolerable switching fluctuation type, immediately triggers the backup sampling path and enables high-speed sampling instructions to obtain redundant current information flow. It then performs a time axis comparison with the main path data and, while ensuring timing continuity, constructs a fused signal result to replace the distorted or drifted segments in the main signal. The error fitting unit continuously compares the fused signal with the preset reference data model to perform fitting regression on the error offset trend of the current fluctuation state. Based on the fitting results, the protection trigger threshold and response time are set to make real-time corrections to the main control strategy. The protection delay unit automatically delays the execution of protection when it determines that the fluctuation state is within an acceptable range, and outputs a fusion signal as an effective monitoring benchmark. At the same time, it suspends high-priority protection response instructions to prevent false operation due to transient errors and improve power continuity during switching. The fault response unit immediately executes the emergency response process when the offset between the fused signal and the reference data model exceeds the preset limit, or when the redundant sampling path cannot complete effective time alignment, generates a high-level protection instruction, interrupts the signal output process, and sends a fault event identifier to the external system for protection processing and logging of power switching control; The error fitting unit includes a dynamic regression correction structure, in which the regression behavior is analyzed based on the mean deviation trend of the fusion signal from the reference model within the current five sampling periods, while considering the consistency of the deviation direction and the stability of the duration; The regression strategy adopts one of two methods: interval minimum variance fitting method and exponential weighted moving average method. The former is used when the current fluctuation intensity is greater than 10%, and the latter is used when the fluctuation duration is less than 10 milliseconds. The offset is converted into a trigger adjustment amplitude factor, whose value range is 0.05 to 0.15, and is used to correct the protection trigger threshold; The error fitting unit also includes a trend segmentation tracking structure to isolate the impact of high-frequency disturbances. It analyzes behavioral trends through fitting residuals within each detection cycle. When the residual is less than the baseline value of 0.03 for three consecutive cycles, it enters the conservative correction phase of the fitting trend. In this phase, no rapid threshold regression adjustment is performed, and the protection strategy is updated based solely on the historical cumulative average error. When the residual reverses direction twice within five cycles, the automatic output signal does not participate in the threshold correction and only maintains the current delay behavior.

2. The intelligent control system for current transformer according to claim 1, characterized in that: The fluctuation identification unit includes an abnormal current fluctuation screening structure. It is used to set a multi-threshold condition judgment mechanism based on the extraction of continuous change amplitude, rising edge slope, phase difference change and frequency mutation value. Different characteristic parameters are combined and calculated in a time-aligned manner to construct a high-dimensional characteristic behavior discrimination space. Through the set three-value hierarchical decision logic, the detection results are marked as stable fluctuation, critical fluctuation or severe abnormality. In the time window slip judgment process, the first judgment factor is whether the fitting coefficient between the average amplitude change and the rising rate of three adjacent sampling periods is less than 0.1, and the second judgment factor is whether the instantaneous frequency change is within 5 Hz. When both conditions are met, the current state is determined to be a critical fluctuation. The critical fluctuation type is identified as the tolerable switching fluctuation type, which is used as the basis for starting the subsequent backup sampling path, so that the working behavior of the redundant acquisition unit has a traceable judgment basis based on the feature recognition results, thereby ensuring the continuity of the technical chain between fluctuation judgment and data compensation.

3. The intelligent control system for current transformer according to claim 2, characterized in that: The redundant acquisition unit establishes a data mirror relationship by running dual-path current signal channels in parallel. The error between the sampling period of the primary path and the sampling period of the backup path is set to no more than ten microseconds to ensure that the time alignment accuracy meets the data fusion requirements. The backup path sampling values need to undergo a preliminary noise removal process, where the removal method includes one of the three methods: maximum difference value method, root mean square limit method, and adaptive judgment based on sliding window filtering; Under the condition that the sampling values meet the redundant timing structure, an interleaved data stream is established by delaying one cycle to ensure the continuity of the redundant data and that it does not cause logical interference with the main path. At the same time, all redundant data will be brought into the phase reconstruction operation process to participate in the construction of the error fitting unit behavior.

4. The intelligent control system for current transformer according to claim 3, characterized in that: The protection delay unit includes a behavior priority conversion structure. When the fused signal is classified as a critical fluctuation type, it actively delays the protection execution time interval. The delay value is not less than 5 milliseconds and not more than 20 milliseconds. The specific delay time is obtained by linear mapping based on the adjustment amplitude factor output by the error fitting unit. During the delay period, all signal outputs are marked as medium risk and high-level interrupt control commands are prohibited from being sent; after the delay, if the regression value of the error fitting unit remains stable, the delay will continue; otherwise, the protection action call behavior in the fault response unit will be executed.

5. The intelligent control system for current transformer according to claim 4, characterized in that: After entering the delayed state, the protection delay unit generates a delay behavior state label. This label contains the delay time interval, the current fusion signal fitting residual value, the amplitude change trend parameter, and the fitting regression level information. The label is embedded in the output signal stream in real time and transmitted to the fault response unit. The delayed behavior status tag is automatically updated every two sampling cycles to assist the fault response unit in executing the response level determination process. When the fitted regression level in the status tag shows a deteriorating trend or the residual value continues to expand, the fault response unit loads the response determination conditions in advance and enters the candidate preparation state. The delay behavior status tag is also used as one of the input fields for generating the fault event identification tag, and its content is written into the event identification structure after the fault response behavior is triggered.

6. The intelligent control system for current transformer according to claim 5, characterized in that: Before executing the emergency response process, the fault response unit introduces a delayed behavior status label as a preprocessing information source for response judgment. The fitting regression level and residual trend in the status label are used as one of the judgment factors. The offset between the current fusion signal and the reference model, the duration, and the fitting trend change are combined to determine the response behavior level. The response behaviors are divided into three categories: interrupt output, signal bypass, and data freeze. If the residual value in the delay behavior state label increases continuously for five cycles, and the fusion signal deviation degree is greater than 15% for three consecutive cycles, an interrupt output is triggered. If the tag trend does not deteriorate significantly but the fluctuation lasts less than ten milliseconds, the data freeze strategy is executed; If the fitted trend shows nonlinear growth and the status label level is medium risk, the signal bypass logic is executed to redirect the current data flow to the redundant path output channel.

7. The intelligent control system for current transformer according to claim 6, characterized in that: When a protection action is triggered, the fault response unit generates a fault event identification tag. This tag structure includes the fault type code, fluctuation behavior tag, protection action record, fusion signal fitting trajectory information, and delay behavior status tag fields. It is encapsulated and output synchronously using a standard communication data frame format. Each event identification tag has a unique number for fault tracing and historical archiving. When multiple event identification tags contain similar fitting trends, similar delay behavior status fields, and different fluctuation starting points, the fault trend clustering logic is triggered and an early warning report is sent to the superior platform.

Images