A voltage transformer real-time diagnosis method based on a honkong system

Abstract

The application provides a voltage transformer real-time diagnosis method based on a Hongmeng system, and relates to the field of data processing. In the method, an original sampling sequence is acquired at a Hongmeng system edge computing node and an adaptive sliding window is constructed, a phase reference signal is extracted based on a standard voltage transformer secondary voltage sampling sequence and a power grid actual operating frequency is dynamically estimated, and a variable-length sampling window is generated; under the window, a windowed interpolation spectrum analysis process is performed to obtain a fundamental wave phasor estimation result, a ratio difference result and an angle difference result are calculated, test working condition information is combined to form an error characteristic data record, an error anomaly diagnosis result is obtained through error characteristic knowledge graph and graph neural network reasoning, and a differential processing strategy is executed to complete real-time diagnosis. By implementing the technical solution provided in the application, real-time diagnosis of error sources can be realized while ensuring the calculation accuracy of the ratio difference result and the angle difference result.

Description

Technical Field

[0001] This application relates to the technical field of data processing, specifically to a real-time diagnostic method for voltage transformers based on the HarmonyOS system. Background Technology

[0002] As an important device in the power system for voltage measurement, energy metering and relay protection, the voltage transformer's ratio difference and phase difference results are directly related to the metering accuracy and system operation reliability. Therefore, in the field inspection process, it is usually necessary to accurately analyze the voltage response relationship between the voltage transformer under test and the standard voltage transformer.

[0003] In existing technologies, a fixed sampling rate is often used to acquire the secondary voltage sampling sequences of the tested voltage transformer and the standard voltage transformer. Spectral analysis is then performed within a preset fixed window to extract the fundamental phasor and calculate the ratio difference and angle difference results. However, in actual operating environments, the actual operating frequency of the power grid is not constant but fluctuates continuously. When the fixed window does not match the actual power grid cycle, asynchronous sampling can easily be introduced, leading to spectral leakage and the picket fence effect. This causes deviations in the fundamental phasor estimation results, ultimately affecting the accuracy of the ratio difference and angle difference calculations.

[0004] Therefore, under the condition of power grid frequency fluctuations, how to achieve real-time diagnosis of error sources while ensuring the accuracy of the calculation of ratio difference and angle difference results has become a key problem that urgently needs to be solved. Summary of the Invention

[0005] This application provides a real-time diagnostic method for voltage transformers based on the HarmonyOS system, which facilitates real-time diagnosis of error sources while ensuring the accuracy of the ratio difference and angle difference calculation results.

[0006] The first aspect of this application provides a real-time diagnostic method for voltage transformers based on the HarmonyOS system. The method includes: at an edge computing node of the HarmonyOS system, acquiring an original sampling sequence sent by a voltage transformer calibrator, containing a secondary voltage sampling sequence of the voltage transformer under test and a secondary voltage sampling sequence of a standard voltage transformer, via a distributed soft bus, and constructing an adaptive sliding window; extracting a phase reference signal from the secondary voltage sampling sequence of the standard voltage transformer according to the adaptive sliding window, and dynamically estimating the actual operating frequency of the power grid to generate a variable-length sampling window synchronized with the actual operating frequency of the power grid; and performing operations on the secondary voltage sampling sequence of the voltage transformer under test and the secondary voltage sampling sequence of the standard voltage transformer respectively according to the variable-length sampling window. Windowed interpolation spectral analysis is performed to obtain the fundamental phasor estimation results of the tested voltage transformer and the standard voltage transformer. Based on the fundamental phasor estimation results of the tested and standard voltage transformers, the ratio difference and angle difference results are calculated, and error feature data records are formed by combining the test condition information. The error feature data records are mapped to an error feature knowledge graph, and a graph neural network is used to reason about error type nodes, influencing factor nodes, and equipment status nodes to obtain error anomaly diagnosis results including anomaly judgment results, error source location results, and confidence assessment results. A differentiated processing strategy is applied to the error anomaly diagnosis results and the original sampling sequence to complete the real-time diagnosis of the voltage transformer based on the HarmonyOS system.

[0007] A second aspect of this application provides a real-time diagnostic device for voltage transformers based on the HarmonyOS system. The device includes an acquisition module and a processing module. The acquisition module is configured to acquire, at the edge computing node of the HarmonyOS system, an original sampling sequence sent by a voltage transformer calibrator, containing a sampling sequence of the secondary voltage of the voltage transformer under test and a sampling sequence of the secondary voltage of a standard voltage transformer, and construct an adaptive sliding window. The processing module is configured to extract a phase reference signal from the sampling sequence of the secondary voltage of the standard voltage transformer according to the adaptive sliding window, and dynamically estimate the actual operating frequency of the power grid to generate a variable-length sampling window synchronized with the actual operating frequency of the power grid. The processing module is further configured to sample the secondary voltage of the voltage transformer under test and the secondary voltage of the standard voltage transformer according to the variable-length sampling window. The sequences are subjected to windowed interpolation spectral analysis to obtain the fundamental phasor estimation results of the tested voltage transformer and the standard voltage transformer. The processing module is further configured to calculate the ratio difference and angle difference based on the fundamental phasor estimation results of the tested and standard voltage transformers, and combine this with test condition information to form error feature data records. The processing module is also configured to map the error feature data records to an error feature knowledge graph, and use a graph neural network to reason about error type nodes, influencing factor nodes, and equipment status nodes to obtain error anomaly diagnosis results including anomaly determination results, error source location results, and confidence assessment results. The processing module is further configured to apply a differentiated processing strategy to the error anomaly diagnosis results and the original sampling sequence to complete real-time diagnosis of the voltage transformer based on the HarmonyOS system.

[0008] A third aspect of this application provides an electronic device including a processor, a memory, a user interface, and a network interface. The memory is used to store instructions, and both the user interface and the network interface are used to communicate with other devices. The processor is used to execute the instructions stored in the memory to cause the electronic device to perform the method described above.

[0009] A fourth aspect of this application provides a non-transitory computer-readable storage medium storing instructions that, when executed, perform the method described above.

[0010] In summary, one or more technical solutions provided in this application have at least the following technical effects or advantages: By directly acquiring the original sampling sequence through edge computing nodes of the HarmonyOS system combined with a distributed soft bus, the sampling sequences of the secondary voltage of the tested voltage transformer and the secondary voltage of the standard voltage transformer can achieve real-time access and synchronization alignment under the same time scale system. This avoids the problems of complex data interfacing between multiple devices and difficulty in unifying timing in traditional methods, ensuring data consistency and reliability for subsequent analysis from the source. By combining an adaptive sliding window with dynamic estimation of the actual operating frequency of the power grid, a variable-length sampling window synchronized with the actual power grid cycle is constructed, ensuring that the sampling window always maintains full-cycle coverage. This fundamentally suppresses spectral leakage and the picket fence effect caused by asynchronous sampling, significantly improving the accuracy of the fundamental phasor estimation results and providing a stable foundation for high-precision calculation of the ratio difference and angle difference results.

[0011] By employing windowed interpolation spectral analysis, the discrete spectrum is finely corrected while suppressing spectral sidelobe diffusion. This results in higher amplitude and phase resolution accuracy for both the fundamental phasor estimation results of the tested voltage transformer and the standard voltage transformer, ensuring that the ratio and phase difference results accurately reflect the metering performance status of the voltage transformer. By fusing the ratio and phase difference results with test condition information, error feature data records with contextual semantics are formed. Errors are no longer isolated values ​​but structured information associated with operating conditions, load status, and environmental factors, providing richer semantic support for subsequent error analysis. Through the combination of error feature knowledge graphs and graph neural networks, error type nodes, influencing factor nodes, and equipment status nodes are correlated and reasoned, achieving a direct mapping from error results to error causes. This not only outputs anomaly judgment results but also provides error source location results and their confidence assessment results, transforming the traditional offline analysis process relying on human experience into a real-time intelligent diagnostic process at the edge.

[0012] By applying differentiated processing strategies to the error anomaly diagnosis results and the original sampling sequences, and by classifying and scheduling data according to the degree of anomaly, key anomaly data is prioritized for retention and processing, while reducing invalid data transmission and computational overhead. This improves overall processing efficiency and system responsiveness while ensuring diagnostic accuracy. Therefore, it facilitates real-time diagnosis of error sources while maintaining the accuracy of comparison and angle difference calculations. Attached Figure Description

[0013] Figure 1 A flowchart illustrating a real-time diagnostic method for voltage transformers based on the HarmonyOS system, provided for an embodiment of this application; Figure 2 A schematic diagram of a real-time diagnostic device for voltage transformers based on the HarmonyOS system provided in this application embodiment; Figure 3This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application.

[0014] Explanation of reference numerals in the attached figures: 21. Acquisition module; 22. Processing module; 31. Processor; 32. Communication bus; 33. User interface; 34. Network interface; 35. Memory. Detailed Implementation

[0015] To enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments.

[0016] In the description of the embodiments of this application, the words "for example" or "for instance" are used to indicate examples, illustrations, or explanations. Any embodiment or design that is described as "for example" or "for instance" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design options. Rather, the use of the words "for example" or "for instance" is intended to present the relevant concepts in a specific manner.

[0017] In the description of the embodiments of this application, the term "multiple" means two or more. For example, multiple systems means two or more systems, and multiple screen terminals means two or more screen terminals. In addition, the terms "comprising," "including," "having," and variations thereof all mean "including but not limited to," unless otherwise specifically emphasized.

[0018] To address the aforementioned technical problems, this application provides a real-time diagnostic method for voltage transformers based on the HarmonyOS system, referring to... Figure 1 , Figure 1 This is a flowchart illustrating a real-time diagnostic method for voltage transformers based on the HarmonyOS system, provided in an embodiment of this application. The method is applied to a server and includes steps S110 to S160, as follows:

[0019] S110. At the edge computing node of the HarmonyOS system, the original sampling sequence sent by the voltage transformer calibrator, which includes the secondary voltage sampling sequence of the voltage transformer under test and the secondary voltage sampling sequence of the standard voltage transformer, is obtained through the distributed soft bus, and an adaptive sliding window is constructed.

[0020] Specifically, the server refers to the computing device in this embodiment that undertakes the responsibilities of data access, data processing, task scheduling, and result output, and is used to execute software processes related to real-time voltage transformer diagnosis at the edge computing node of the HarmonyOS system. After the server is deployed on the edge computing node of the HarmonyOS system, it first enters the cooperative domain listening state and calls the device discovery capability provided by the distributed soft bus to continuously scan the surrounding communicable devices. The edge computing node of the HarmonyOS system refers to the computing device deployed at the inspection site that can directly connect to the voltage transformer calibrator and complete data processing and diagnostic operations locally; the distributed soft bus refers to the unified communication base in the HarmonyOS system used to realize cross-device automatic discovery, connection establishment, data transmission, and cooperative scheduling; the cooperative domain refers to a set of devices that can complete mutual recognition and cooperation through the distributed soft bus in the same communication environment; the voltage transformer calibrator refers to the field inspection device used to output the secondary voltage sampling sequence of the tested voltage transformer and the secondary voltage sampling sequence of the standard voltage transformer. After discovering the voltage transformer calibrator, the server first reads the device identifier, service capabilities, sampling channel configuration, and data output description, and then establishes a data connection with the voltage transformer calibrator. After establishing the connection, channel registration and field binding are performed on the data stream. Channel registration involves assigning fixed logical receiving channels to the secondary voltage sampling sequences of the tested voltage transformer and the standard voltage transformer, respectively, to avoid channel confusion during subsequent processing. Field binding involves associating fields such as sample value, sampling time, device identifier, channel identifier, and sampling batch with the corresponding channel to form a unified data organization format. To facilitate subsequent dual-channel synchronous processing, the data record received at any given time can be expressed as follows:

[0021]

[0022] in, Indicates the first One sampling record; The device identifier for the voltage transformer calibrator is used to distinguish the data source; This indicates the sampling channel identifier, used to distinguish between the secondary voltage sampling sequence of the voltage transformer under test and the secondary voltage sampling sequence of the standard voltage transformer. This indicates the sampling time corresponding to this sampling record; This indicates the voltage sample value corresponding to this sampling record; The status flag for this sampling record indicates whether it is normal, missing, complete, or abnormal. Through the above processing, the server can stably acquire the original sampling sequence at the edge computing node of the HarmonyOS system and provide a unified data entry point for the subsequent formation of a dual-channel synchronous data structure.

[0023] After device discovery and connection establishment are completed, the server performs time synchronization alignment, integrity checks, and order reordering on the received dual-channel synchronous data structure to generate a time-continuous and channel-corresponding original sampling sequence. The dual-channel synchronous data structure refers to a data organization form that simultaneously accommodates the secondary voltage sampling sequences of the tested voltage transformer and the standard voltage transformer, constrained by a unified time axis. Time synchronization alignment refers to pairing the two sampling records according to their sampling time, ensuring a logical correspondence between two data streams at the same time or within the same sampling period. Integrity checks determine whether both the tested channel data and the standard channel data exist simultaneously at any given time, and whether any fields are missing. Order reordering involves reordering the data according to sampling time when network jitter, buffer out-of-order processing, or batch transmission causes a discrepancy between the data arrival order and the sampling order. The alignment result of the tested channel and the standard channel on a unified time axis can be represented as:

[0024]

[0025] in, Indicates the aligned first One dual-channel sampling unit; Represents the first on the unified timeline Each sampling time; Indicates at time The corresponding sampled value of the secondary voltage of the tested voltage transformer; Indicates at time The corresponding standard voltage transformer secondary voltage sampling value; This represents the matching status flag of the sampling unit, used to identify whether the dual-channel data at that moment is complete, whether it has been padded, and whether there are any anomalies. To quantify the completeness assessment, the dual-channel completeness coefficient can be further defined as:

[0026]

[0027] in, The integrity coefficient is used to measure the effectiveness of the dual-channel synchronous data structure within the current processing interval. Its value ranges from greater than or equal to zero to less than or equal to one. The closer the value is to one, the higher the integrity. This indicates the number of sampling units that simultaneously have valid secondary voltage sample values ​​of the tested voltage transformer and valid secondary voltage sample values ​​of the standard voltage transformer within the current processing interval. This represents the total number of sampling units that should theoretically exist within the current processing interval. The principle behind the above formula is that the proportion of effective dual-channel sampling units to theoretical sampling units reflects whether the current original sampling sequence meets the basic conditions for continuing into the window construction stage. When the integrity coefficient is lower than a preset threshold, the server can anomaly mark the missing positions or wait for subsequent retransmissions; when the integrity coefficient meets the requirements, it outputs the original sampling sequence that is temporally continuous and corresponds to the channels.

[0028] After obtaining the time-continuous and channel-corresponding original sampling sequences, the server performs joint cache organization based on the original sampling sequences. This ensures that the secondary voltage sampling sequences of the tested voltage transformer and the standard voltage transformer form corresponding storage segments along a unified time axis, and adds window boundary markers and data validity markers to these corresponding storage segments. Joint cache organization means that instead of caching the two sampling sequences independently, they are written in pairs to corresponding positions in the same logical cache area along a unified time axis, thus ensuring natural alignment of the two data streams during subsequent window truncation. A corresponding storage segment refers to a continuous data interval divided according to time continuity within the cache space. Each interval simultaneously stores the secondary voltage sampling sequences of the tested voltage transformer and the standard voltage transformer. Window boundary markers are control markers indicating whether a data interval may serve as the start or end point of a window. Data validity markers are status markers indicating whether the data within the interval is complete and suitable for frequency tracking and spectrum analysis. Any corresponding storage segment in the joint cache area can be represented as:

[0029]

[0030] in, Indicates the first Each corresponding storage segment; This indicates the set of time periods covered by the corresponding storage segment; This represents the set of secondary voltage sampling sequences of the tested voltage transformer within the corresponding storage segment; This represents the set of standard voltage transformer secondary voltage sampling sequences within the corresponding storage segment; This indicates the window boundary identifier corresponding to the storage segment; This represents the data validity flag corresponding to the specified storage segment. To quantify the validity of this storage segment, the segment validity can be defined as:

[0031]

[0032] in, Indicates the first The validity of a corresponding storage segment is used to comprehensively reflect whether the segment is suitable as the basis for subsequent window construction. The value range is usually set between zero and one, and the larger the value, the higher the quality of the segment. This indicates the integrity evaluation value of the corresponding storage segment; This represents the sequential consistency evaluation value of the corresponding storage segment, used to reflect whether the data is out of order; This represents the dual-channel synchronization evaluation value of the corresponding storage segment, used to reflect the degree of correspondence between the secondary voltage sampling sequence of the tested voltage transformer and the secondary voltage sampling sequence of the standard voltage transformer on the time axis. , , This represents the weighting coefficients used to adjust the influence of integrity, sequence consistency, and synchronicity in the overall evaluation. Each coefficient is a value greater than or equal to zero, and the sum is usually set to one. The principle of this formula is to weight and fuse multiple factors that determine the quality of a segment to form a single evaluation value, thereby providing an objective basis for the subsequent generation of window boundaries. When the segment's validity meets the threshold, the server retains it as a valid candidate segment; when the segment's validity is low, it is supplemented, reorganized, or discarded.

[0033] After the corresponding storage segment is formed and quality screening is completed, the server generates an initial window boundary based on the corresponding storage segment, according to the preset window coverage and window update rhythm. An adaptive sliding window is constructed by retaining overlapping data segments and incorporating newly added original sampling sequences. The preset window coverage refers to the target time length or number of target sampling points to be covered by a single window, typically requiring coverage of multiple power frequency cycles to improve the stability of subsequent frequency tracking and fundamental phasor estimation. The window update rhythm refers to the advance interval between two adjacent windows, which can be understood as the window sliding step size. The initial window boundary refers to the window's start and end positions determined based on the existing data in the current buffer. Overlapping data segments refer to the common data portion retained after the previous window ends and still required to participate in the calculation of the next window. The adaptive sliding window is a window structure that can be continuously updated as new original sampling sequences are added, and whose effective length can be further adjusted based on subsequent frequency estimation results. The relationship between the initial window length and the update step size can be described using the number of sampling points:

[0034]

[0035]

[0036] in, Indicates the number of sampling points covered by the initial window; This indicates the target time length corresponding to the preset window coverage area, and its value is set according to the on-site signal stability and target frequency resolution requirements; This represents the sampling rate, which indicates the number of sampling points acquired per unit of time. This indicates rounding down, used to convert the time length into an integer number of sampling points; This indicates the number of sampling points corresponding to the window update step size; This represents the target advancement time length corresponding to the window update rhythm. The principle behind the above formula is to convert the window coverage requirements and window advancement requirements in the time dimension into the number of directly operable sampling points in the buffer, thereby facilitating the server to accurately capture the window boundaries in the joint buffer.

[0037] After determining the initial window boundaries, the first... An adaptive sliding window is represented as:

[0038] in, Indicates the first An adaptive sliding window; This indicates the starting dual-channel sampling unit of the window; This indicates the termination of the dual-channel sampling unit in this window; Indicates the first The starting index position of each adaptive sliding window on a unified time axis; This represents the total number of sampling points covered by the window. To ensure both continuity and smooth transition between windows, adjacent windows typically retain overlapping portions, and the overlap rate can be expressed as:

[0039]

[0040] in, This represents the overlap rate between two adjacent adaptive sliding windows. Its value ranges from greater than or equal to zero to less than one. The larger the value, the more data the adjacent windows share. Indicates the total number of sampling points in the window; This represents the number of sampling points corresponding to the window update step size. The principle behind this formula is to measure the proportion of historical information retained during window advancement by using the ratio of the number of shared sampling points between adjacent windows to the total number of sampling points in the window. By reasonably setting the overlap rate, it is possible to reduce abrupt changes in results between windows while ensuring continuous window advancement, thereby improving the smoothness of subsequent frequency tracking. Each time the server receives a new original sampling sequence, it first retains the overlapping data segments from the previous window, then appends the new original sampling sequence to the end of the buffer, and regenerates the next window according to the new starting index, thus achieving the continuous construction of an adaptive sliding window.

[0041] Furthermore, to ensure that the constructed adaptive sliding window not only meets the length requirement but also the conditions for subsequent frequency tracking, the server can perform a window availability assessment for each adaptive sliding window. Window availability can be represented as:

[0042]

[0043] in, Indicates the first The availability evaluation value of an adaptive sliding window is used to determine whether the window is suitable for entering the subsequent phase reference signal extraction and dynamic estimation process of the actual operating frequency of the power grid. This indicates the overall validity of the data segment covered by the adaptive sliding window, which can be obtained by summarizing the validity of the aforementioned segments; This represents the waveform continuity evaluation value within the adaptive sliding window, used to reflect whether there are obvious breaks or jumps in the secondary voltage sampling sequence of the standard voltage transformer within the window. This represents the periodic stability evaluation value within the adaptive sliding window, used to reflect whether the periodic structure of the voltage waveform within the window is clear; , , This represents the weighting coefficients, used to adjust the proportion of influence of data validity, waveform continuity, and periodic stability on window availability. Each coefficient is a value greater than or equal to zero, and is usually set to sum to one. The principle of this formula is to comprehensively evaluate window quality, avoiding sending windows that meet the length condition but not the signal quality condition directly to subsequent processing. Only when the window availability reaches a preset threshold will the server determine the window as a valid adaptive sliding window, and use it in the next step to extract the phase reference signal from the secondary voltage sampling sequence of the standard voltage transformer.

[0044] S120. Based on the adaptive sliding window, extract the phase reference signal from the secondary voltage sampling sequence of the standard voltage transformer, dynamically estimate the actual operating frequency of the power grid, and generate a variable-length sampling window that is synchronized with the actual operating frequency of the power grid.

[0045] Specifically, after constructing the adaptive sliding window, the standard voltage transformer secondary voltage sampling sequence is first retrieved from the current adaptive sliding window and designated as the sole reference data source. The adaptive sliding window refers to a data window that continuously advances along a unified time axis on the original sampling sequence, providing a stable analysis interval for subsequent frequency tracking. The standard voltage transformer secondary voltage sampling sequence refers to the continuous voltage sampling value sequence output from the corresponding channel of the standard voltage transformer, serving to provide a phase reference for frequency tracking. The sole reference data source means that during the current frequency dynamic estimation process, only the standard voltage transformer secondary voltage sampling sequence is allowed to participate in phase reference signal extraction, and the tested voltage transformer secondary voltage sampling sequence is no longer mixed in, to avoid the tested side's own errors, local waveform distortions, or transient disturbances affecting frequency tracking stability. At the edge computing node of the HarmonyOS system, the server records the current adaptive sliding window as:

[0046]

[0047] in, Indicates the first An adaptive sliding window; This indicates the sampling sequence of the secondary voltage of the standard voltage transformer at the 1st... The sampled values ​​corresponding to each sampling position; Indicates the first The starting sampling index of an adaptive sliding window; Indicates the first The number of sampling points contained in an adaptive sliding window. The principle behind the above expression is that a continuous set of sampling points clarifies the current data range for frequency tracking, ensuring that subsequent waveform purification, phase reference signal extraction, and dynamic frequency estimation are all based on the same analysis window. By determining the secondary voltage sampling sequence of the standard voltage transformer as the sole reference data source, it can be guaranteed that subsequent phase evolution relationships are always established around a unified benchmark, thereby making the dynamic frequency estimation results more consistent.

[0048] After the standard voltage transformer secondary voltage sampling sequence is determined as the sole reference data source, waveform cleansing is performed on the standard voltage transformer secondary voltage sampling sequence to generate a reference sampling sequence that satisfies phase continuity and period stability. Waveform cleansing refers to suppressing glitches, isolated anomalies, local high-frequency disturbances, and small-scale waveform jitter in the original standard voltage transformer secondary voltage sampling sequence, while preserving the fundamental main period structure and phase evolution trend as much as possible. Phase continuity means that the waveform phases corresponding to adjacent sampling times should maintain a smooth transition without non-physical jumps. Period stability means that within an adaptive sliding window, the periodic structure of the waveform should remain basically consistent, facilitating the subsequent extraction of a reliable phase reference signal. The reference sampling sequence refers to the standard voltage transformer secondary voltage sampling sequence specifically used for phase reference signal extraction after waveform cleansing. The server can construct the reference sampling sequence using a combination of sliding smoothing and anomaly suppression. The cleaned reference sampling sequence can be represented as:

[0049]

[0050] in, Indicates the first Reference sampling sequence values ​​obtained at each sampling location; This represents the smoothing weight coefficient used in waveform cleanup, and its value is determined by the selected smoothing strategy. Indicates the first The sampling location is the center, and the sampling location within the neighborhood is the first... The sampled values ​​of the secondary voltage of the standard voltage transformer corresponding to each sampling position; This represents the half-width of the cleaned neighborhood, used to control the range of neighboring sampling points participating in the smoothing calculation. The principle behind this formula is that by weighting and fusing several sampling values ​​near the current sampling point, the impact of a single outlier on the overall waveform is weakened, while preserving the continuous contour of the fundamental waveform. To further determine whether the reference sampling sequence after waveform cleansing meets the requirements for subsequent phase extraction, a periodic stability evaluation metric can be defined as:

[0051]

[0052] in, Indicates the first The periodic stability evaluation quantity of the reference sampling sequence within an adaptive sliding window, whose value is usually between zero and one, and the larger the value, the more stable the periodic structure. Indicates the first The amount of change in the reference sampling sequence at each sampling position relative to the previous sampling position; This represents the average value of the change in the reference sampling sequence within the current adaptive sliding window; This represents the upper bound or empirical scale value of the change used for normalization. The principle behind this formula is to use the dispersion of the change at adjacent sampling points relative to the average change to measure whether the current reference sampling sequence maintains a stable periodic evolution. When the periodic stability evaluation value reaches the threshold, the generated reference sampling sequence can be considered to have met the conditions for subsequent phase reference signal extraction.

[0053] After generating the reference sampling sequence, a phase reference signal that is continuously updated over time is extracted based on the reference sampling sequence. Continuous tracking and error iteration processing are then performed on the phase reference signal to obtain a dynamic estimate of the actual operating frequency of the power grid. The phase reference signal is a continuous signal extracted from the reference sampling sequence that characterizes the current voltage waveform phase position and its time evolution. Continuous tracking refers to continuously updating the estimate of the current phase state as the reference sampling sequence progresses over time. Error iteration processing involves continuously feeding back the deviation between the current phase reference signal and the internally estimated phase to correct the frequency estimate. The dynamic estimate of the actual operating frequency of the power grid refers to the actual power frequency estimate that is continuously updated over time. The server can first construct an analytical phase representation and then form the phase reference signal. The phase reference signal can be represented as:

[0054]

[0055] in, Indicates the first Phase reference signal corresponding to each sampling position; This represents the arctangent function used to calculate the instantaneous phase based on the quadrature and in-phase components; This represents the orthogonal components obtained after orthogonal construction of the reference sampling sequence; Indicates the first The formula defines the reference sampling sequence value corresponding to each sampling position. The principle behind this formula is to combine the reference sampling sequence and its corresponding orthogonal components into a two-dimensional phase expression, thereby obtaining a continuous phase state at each sampling time. Based on the phase reference signal, the phase tracking error can be defined as:

[0056]

[0057] in, Indicates the first Phase tracking error corresponding to each sampling position; Indicates the first Phase reference signal corresponding to each sampling position; Indicates the first The formula defines the internal estimated phase corresponding to each sampling position. The principle behind this formula is to quantify the current tracking error by using the difference between the reference phase and the estimated phase. Subsequently, the server continuously corrects the frequency estimate using the phase tracking error; the frequency update relationship can be expressed as:

[0058]

[0059] in, Indicates the first Frequency estimates corresponding to each sampling location; This represents the frequency estimate corresponding to the previous sampling position; This represents the proportional adjustment coefficient, used to reflect the immediate strength of the effect of the current phase tracking error on frequency correction; This represents the integral adjustment coefficient, used to reflect the sustained effect of accumulated historical phase tracking errors on frequency correction; This indicates the distance from the starting sampling position of the current adaptive sliding window to the nth The formula is the cumulative amount of phase tracking error between sampling positions. The principle behind this formula is to introduce the current deviation and the historical cumulative deviation into the frequency correction process, so that the frequency estimation can both respond quickly to fluctuations and gradually approach the stable value, thereby forming a dynamic estimation result of the actual operating frequency of the power grid.

[0060] After obtaining the dynamic estimation result of the actual operating frequency of the power grid, a convergence determination process is performed on the dynamic estimation result. If the preset stability conditions are met, the dynamic estimation result is determined as a valid frequency estimation result. The convergence determination process refers to performing a stability test on the frequency estimation trajectory over a continuous period to determine whether the current dynamic frequency estimation result has overcome initial disturbances and local jitter, and can serve as a reliable basis for window synchronization reconstruction. The preset stability conditions refer to several pre-set convergence judgment thresholds, such as frequency change amplitude thresholds, short-time fluctuation thresholds, and estimation residual thresholds. A valid frequency estimation result refers to a frequency estimation result that meets the above stability conditions and can be directly used for generating variable-length sampling windows. The server can perform the most recent... Statistical analysis was performed on the frequency estimates, and the short-time frequency fluctuation was defined as:

[0061]

[0062] in, Indicates as of the date The short-term frequency fluctuation over a window period is used to measure the dispersion of recent frequency dynamic estimation results; Indicates the first Frequency estimation results corresponding to each window; Indicates as of the date The window, recent The average of the frequency estimates; This represents the number of windows involved in the convergence determination. The principle behind this formula is to determine whether the current frequency dynamic estimation result has reached a stable state by examining the fluctuation of the frequency estimation results relative to the average value in the most recent windows. To further incorporate phase error convergence, the convergence determination function can be defined as:

[0063]

[0064] in, Indicates the first The comprehensive convergence judgment quantity corresponding to each window; and This represents the weighting coefficient, used to adjust the influence of the frequency fluctuation term and the phase error term in the judgment; Indicates the first The formula represents the average phase tracking error magnitude corresponding to each window. The principle behind this formula is that it simultaneously incorporates the stationarity of the frequency estimation trajectory and the phase tracking accuracy into the judgment, making the convergence determination more reliable. When When the result is less than the preset threshold, it means that the current dynamic estimation result has met the preset stability condition. At this time, the server determines it as a valid frequency estimation result and writes it into the shared processing area for subsequent window length synchronization reconstruction calls.

[0065] After the effective frequency estimation results are determined, a synchronous reconstruction process is performed on the window length based on the effective frequency estimation results to generate a variable-length sampling window that forms an integer-period matching relationship with the actual operating frequency of the power grid. Synchronous reconstruction means that instead of using the fixed length of the original adaptive sliding window, the number of sampling points to be covered by the window is recalculated based on the current effective frequency estimation results, so that the window length can correspond to several complete actual power frequency cycles. Integer-period matching means that the new window length exactly covers an integer number of actual cycles, thus avoiding additional spectral leakage introduced by cycle boundary truncation. A variable-length sampling window refers to a new analysis window whose length can be dynamically adjusted according to changes in the effective frequency estimation results. The server first calculates the number of sampling points corresponding to a single actual cycle based on the effective frequency estimation results, and then obtains the total number of sampling points for the variable-length sampling window based on the preset cycle coverage number, specifically expressed as follows:

[0066]

[0067] in, Indicates the number of sampling points in a variable-length sampling window; This represents the rounding function, used to convert the theoretical number of sampling points into an integer number of sampling points; This indicates the preset number of periods to cover, used to represent how many complete actual periods a variable-length sampling window is expected to cover. Its value is usually an integer greater than or equal to one. This represents the sampling rate, used to indicate the number of sampling points acquired per unit of time. This represents the effective frequency estimation result, used to indicate the actual operating frequency of the power grid currently considered reliable. The principle behind this formula is to first determine the theoretical number of sampling points corresponding to one actual cycle using the sampling rate and the effective frequency estimation result, then multiply this by a preset cycle coverage number to ensure the generated window covers an integer number of actual cycles. Finally, the integer window length, which can be directly used for sample extraction, is obtained by rounding. Subsequently, the server recalibrates the start and end boundaries on the original sampling sequence covered by the current adaptive sliding window, constructing a variable-length sampling window, which can be expressed as:

[0068]

[0069] in, Indicates the first A variable-length sampling window; Indicates the first The sampled values ​​of the secondary voltage of the standard voltage transformer corresponding to each sampling position; Indicates the first The sampled values ​​of the secondary voltage of the voltage transformer under test corresponding to each sampling location; Indicates the starting sampling index of the current window; This indicates the number of sampling points in the variable-length sampling window. The principle behind this formula is that, under the constraints of a unified time axis and a unified starting position, corresponding sample values ​​from both the standard channel and the tested channel are simultaneously extracted, thus forming a dual-channel analysis window synchronized with the actual operating frequency of the power grid. This variable-length sampling window inherits the continuous sampling foundation provided by the preceding adaptive sliding window and introduces an integer-cycle constraint synchronized with the actual power frequency cycle, thereby providing a high-quality input window for subsequent windowed interpolation spectrum analysis.

[0070] S130. Based on the variable-length sampling window, perform windowed interpolation spectrum analysis on the secondary voltage sampling sequence of the tested voltage transformer and the secondary voltage sampling sequence of the standard voltage transformer to obtain the fundamental phasor estimation results of the tested voltage transformer and the fundamental phasor estimation results of the standard voltage transformer.

[0071] Specifically, after generating the variable-length sampling window, the secondary voltage sampling sequences of the tested voltage transformer and the standard voltage transformer are retrieved separately based on the variable-length sampling window, and a unified window loading process is performed to ensure that the two sampling sequences enter the spectrum analysis process under the same window boundaries. The variable-length sampling window refers to a data window reconstructed based on the effective frequency estimation results and matched with the actual operating frequency of the power grid to form an integer period relationship; the unified window loading process refers to simultaneously extracting the sample intervals corresponding to the secondary voltage sampling sequences of the tested voltage transformer and the standard voltage transformer under the same starting sampling index and the same ending sampling index constraint, so that the two sampling sequences have completely consistent time coverage and sample boundaries in subsequent spectrum analysis. Let the first... The starting sampling index of each variable-length sampling window is Window length is Then, the subject channel and standard channel samples after the unified window loading can be represented as follows:

[0072]

[0073]

[0074] in, Indicates the first The secondary voltage sampling sequence of the tested voltage transformer loaded within a variable-length sampling window; Indicates the first The standard voltage transformer secondary voltage sampling sequence is loaded within a variable-length sampling window; Indicates the first The sampled values ​​of the secondary voltage of the voltage transformer under test corresponding to each sampling location; Indicates the first The sampled values ​​of the secondary voltage of the standard voltage transformer corresponding to each sampling position; Indicates the first The starting sampling index of a variable-length sampling window; This indicates the number of sampling points in the variable-length sampling window. The principle behind the above expression is that by applying the same window boundary constraints to both sampling sequences, the subsequent spectral analysis results can be established on the same time base and the same whole cycle, thereby ensuring that the final extracted fundamental phasor estimation results of the tested voltage transformer and the fundamental phasor estimation results of the standard voltage transformer are directly comparable.

[0075] After completing the unified window loading process, a window validity review is performed on the secondary voltage sampling sequences of the tested voltage transformer and the standard voltage transformer to generate valid sampling sequences that meet the requirements of waveform continuity and channel synchronization. The window validity review involves quality control of the two sampling sequences within the current variable-length sampling window, focusing on checking for missing samples, out-of-order samples, local spikes, dual-channel misalignment, and waveform jumps. Waveform continuity means that within the same channel, waveform changes between adjacent sampling points should maintain a smooth evolution, without non-physical breaks or abrupt jumps. Channel synchronization means that the time points corresponding to the secondary voltage sampling sequences of the tested and standard voltage transformers at the same sampling position should be consistent. A valid sampling sequence is a sample sequence that, after passing the window validity review, can proceed to windowing and spectral transformation processing. The window continuity evaluation metric can be defined as:

[0076]

[0077] in, Indicates the first The waveform continuity evaluation value corresponding to a variable length sampling window is usually between zero and one. The larger the value, the better the waveform continuity. Indicates the first The waveform change at each sampling position relative to the previous sampling position can be calculated for the secondary voltage sampling sequence of the voltage transformer under test and the secondary voltage sampling sequence of the standard voltage transformer, respectively. This represents the average value of the waveform changes within the current window; This represents the magnitude scale used for normalization. The principle behind this formula is to evaluate the smoothness and continuity of the waveform within a window by examining the deviation of the changes at adjacent sampling points from the average change. The channel synchronicity evaluation metric can be further defined as:

[0078]

[0079] in, Indicates the first The channel synchronization evaluation value corresponding to a variable-length sampling window is usually between zero and one. The larger the value, the better the time correspondence between the two channels. This indicates that the secondary voltage sampling sequence of the tested voltage transformer is in the [number]th [year]. The time identifier corresponding to each sampling location; This indicates the sampling sequence of the secondary voltage of the standard voltage transformer at the 1st... The time identifier corresponding to each sampling location; This represents the time deviation scale used for normalization. The principle behind this formula is to quantify whether the two sampling sequences within the current window remain synchronized by normalizing the sampling time difference between the two channels. The server integrates the waveform continuity evaluation and the channel synchronization evaluation to perform processes such as removal, interpolation, marking, or rearrangement on abnormal samples, thereby outputting a valid sampling sequence that meets the requirements of spectrum analysis.

[0080] After generating the valid sampling sequences, windowing is applied to each valid sampling sequence to suppress spectral sidelobe spread caused by window boundaries. Windowing refers to applying weights to each sampling point within the window before the time-domain samples enter the spectral transform, so that the contribution of sampling points near the window boundary gradually decreases, thereby weakening the edge discontinuities caused by window truncation. Spectral sidelobe spread refers to the phenomenon that frequency domain energy spreads from the dominant frequency to surrounding frequencies due to the truncation of the time-domain signal by a finite-length window. The server can use a Hanning window to process the secondary voltage sampling sequences of the tested voltage transformer and the secondary voltage sampling sequences of the standard voltage transformer separately. The window function can be expressed as:

[0081]

[0082] in, Indicates the first The window function weights corresponding to each window position; This represents the position index within the window, with values ​​ranging from zero to... ; Indicates the number of sampling points in a variable-length sampling window; This represents the constant pi. The principle behind this formula is to gradually decrease the weight of the window function at the ends of the window while maintaining a larger weight in the middle, thereby reducing abrupt changes at the boundary. The windowed subject channel and standard channel sequences are represented as follows:

[0083]

[0084]

[0085] in, This indicates that the secondary voltage sampling sequence of the tested voltage transformer, after windowing processing, is at the [number]th [time / period]. Sample values ​​corresponding to each window position; This indicates that the secondary voltage sampling sequence of the standard voltage transformer, after windowing processing, is at the [number]th [number]th [number]. Sample values ​​corresponding to each window position; This represents the sampled value of the secondary voltage of the voltage transformer under test before the window is applied; This represents the sampled value of the secondary voltage of the standard voltage transformer before the window is applied; This represents the window function weight at the corresponding position. The principle behind the above formula is to reduce the abrupt changes in samples at the window edges through time-domain weighting, thereby obtaining a more concentrated main lobe response and weaker side lobe spread in subsequent spectral transformation processing, creating better frequency domain conditions for peak spectral line identification and interpolation correction processing.

[0086] After windowing, a spectrum transform is performed on the windowed sampled sequence, and the main response region is located within the fundamental frequency neighborhood to obtain the main response region. Spectrum transform converts the time-domain sampled sequence into a discrete frequency-domain spectral representation, allowing the frequency components, amplitude distribution, and phase distribution of the signal to be explicitly expressed in the frequency domain. The fundamental frequency neighborhood is a finite search interval defined around the actual operating frequency of the power grid, used to avoid interference from higher harmonics and noise peaks in the fundamental frequency location. The main response region is the frequency domain region within the fundamental frequency neighborhood where the energy is most concentrated and most likely corresponds to the true fundamental frequency components. The server can perform discrete Fourier transforms on the test channel and the standard channel respectively, with the spectral expression as follows:

[0087]

[0088]

[0089] in, This indicates that the secondary voltage sampling sequence of the tested voltage transformer is in the [number]th [year]. The spectral values ​​corresponding to each frequency point; This indicates the sampling sequence of the secondary voltage of the standard voltage transformer at the 1st... The spectral values ​​corresponding to each frequency point; This represents the frequency index, with values ​​ranging from zero to... ; This represents the windowing result of the secondary voltage sampling sequence of the tested voltage transformer; This represents the windowed result of the secondary voltage sampling sequence of a standard voltage transformer; Represents the imaginary unit; This represents an exponential function. The principle behind this formula is that each time-domain sampling point is projected onto different discrete frequency points, thus forming a comprehensive representation of the frequency domain amplitude and phase. To locate the main response region, the theoretical center frequency of the fundamental wave is first determined based on the effective frequency estimation results:

[0090]

[0091] in, Indicates the center frequency of the fundamental frequency theory; This represents the rounding function; This represents the effective frequency estimation result; Indicates the number of sampling points in a variable-length sampling window; This represents the sampling rate. The principle behind this formula is to map the true fundamental frequency to a discrete spectral index position based on the current effective frequency estimation result. Subsequently, in Set the surrounding neighborhood range and in The main response region is searched within the interval to ensure that subsequent peak spectral line identification and processing are focused on the most likely fundamental wave region.

[0092] After obtaining the main response region, peak spectral line identification processing is performed on the main response region, and interpolation correction processing is performed based on the peak spectral line and its adjacent spectral lines to obtain the interpolation correction result. Peak spectral line identification processing refers to finding the discrete frequency point with the largest amplitude within the main response region as a coarse localization result closest to the true fundamental frequency. The peak spectral line refers to the discrete frequency point with the largest amplitude within the main response region. The adjacent spectral lines refer to the frequency points located to the left and right of the peak spectral line, used to describe the degree of offset of the true fundamental frequency relative to the discrete frequency grid. Interpolation correction processing refers to using the amplitude relationship between the peak spectral line and its adjacent spectral lines to refine the estimation of the true fundamental frequency, amplitude, and phase, in order to reduce the picket fence effect caused by the discrete frequency grid. The peak spectral line can be represented as:

[0093]

[0094] in, This indicates the frequency index corresponding to the peak spectral line; This represents the variable that maximizes the objective function; Indicates the first The spectral amplitude corresponding to each frequency point; Indicates the center frequency of the fundamental frequency theory; This represents the half-width of the main response region. The principle behind this formula is to find the discrete frequency point with the largest amplitude within a predetermined fundamental frequency neighborhood, thus coarsely locating the fundamental frequency. Subsequently, interpolation correction coefficients can be constructed using the amplitude relationship between the peak spectral line and its adjacent spectral lines. The three-spectral-line interpolation correction coefficients can be expressed as:

[0095]

[0096] in, This represents the interpolation correction factor, used to characterize the offset of the true peak frequency relative to the center position of the peak spectral line. Its value is usually between -0.5 and +0.5. This indicates the spectral amplitude corresponding to the peak spectral line; This indicates the spectral amplitude corresponding to the spectral line adjacent to the left of the peak spectral line; This represents the spectral amplitude corresponding to the spectral line adjacent to the right of the peak spectral line. The principle behind this formula is to determine the direction and degree of the true fundamental frequency's offset in the discrete frequency grid by comparing the response differences between adjacent spectral lines to the left and right of the peak spectral line. Based on the interpolation correction coefficients, the corrected frequency position can be further obtained.

[0097]

[0098] in, Indicates the position of continuous frequencies after interpolation correction; This indicates the index of the discrete frequency point corresponding to the peak spectral line; This represents the interpolation correction coefficient. The principle behind this formula is to superimpose the discrete frequency point position with the sub-frequency point offset, thereby obtaining a continuous position expression that is closer to the true fundamental frequency. Based on this, more accurate frequency support can be provided for subsequent fundamental phasor extraction.

[0099] After obtaining the interpolation correction results, the fundamental phasor estimation results of the tested voltage transformer and the standard voltage transformer are extracted based on the interpolation correction results. The fundamental phasor estimation result of the tested voltage transformer refers to a complex quantity expression composed of the fundamental frequency, fundamental amplitude, and fundamental phase of the tested channel within the current variable-length sampling window; the fundamental phasor estimation result of the standard voltage transformer refers to a complex quantity expression composed of the fundamental frequency, fundamental amplitude, and fundamental phase of the standard channel within the current variable-length sampling window. The fundamental phasor estimation results are essentially a unified vector expression of the fundamental components, which can be directly used for subsequent calculations of the ratio difference and angle difference results. The corrected fundamental frequency can be expressed as:

[0100]

[0101] in, This indicates the fundamental frequency after interpolation correction; Indicates the position of continuous frequencies after interpolation correction; Indicates the sampling rate; This represents the number of sampling points in a variable-length sampling window. The principle behind this formula is to remap continuous positions in the frequency domain back to actual frequency values. The corrected fundamental phasor can be expressed as: ,in, This indicates the fundamental phasor estimation result; This represents the fundamental amplitude after interpolation correction; This indicates the fundamental phase after interpolation correction; This represents the unit phase factor determined by the fundamental wave phase. The principle behind this formula is that it combines the amplitude and phase of the fundamental wave into a complex quantity, thus uniformly characterizing the magnitude and direction of the fundamental wave component. For the subject channel and the standard channel, the server obtained:

[0102]

[0103]

[0104] in, This represents the estimation result of the fundamental phasor of the tested voltage transformer; This represents the estimation results of the fundamental phasor of a standard voltage transformer; and These represent the fundamental amplitude estimation results for the subject channel and the standard channel, respectively; and These represent the fundamental phase estimation results for the tested channel and the standard channel, respectively. Through the above processing, the server can generate the fundamental phasor estimation results for the tested voltage transformer and the standard voltage transformer under the same window boundaries, the same integer cycle constraints, and the same interpolation correction logic, providing a unified and high-precision fundamental phasor input for the next step of calculating the ratio difference and angle difference results.

[0105] S140. Based on the fundamental phasor estimation results of the tested voltage transformer and the fundamental phasor estimation results of the standard voltage transformer, calculate the ratio difference and angle difference results, and combine them with the test condition information to form error characteristic data records.

[0106] Specifically, after obtaining the fundamental phasor estimation results of the tested voltage transformer and the standard voltage transformer, phasor decomposition processing is first performed on both to obtain amplitude expression sequences and phase expression sequences, respectively. The fundamental phasor estimation result of the tested voltage transformer refers to the fundamental complex quantity expression obtained after windowed interpolation spectral analysis processing of the secondary voltage sampling sequence of the tested voltage transformer; the fundamental phasor estimation result of the standard voltage transformer refers to the fundamental complex quantity expression obtained after the same processing of the secondary voltage sampling sequence of the standard voltage transformer. Phasor decomposition processing refers to decomposing the complex quantity form of the fundamental phasor into amplitude expression sequences representing the magnitude of the amplitude and phase expression sequences representing the phase position. Let the first... The fundamental phasor estimation results of the tested voltage transformer corresponding to each variable-length sampling window are as follows: The standard voltage transformer fundamental phasor estimation result is as follows: Then its phasor decomposition result can be expressed as:

[0107]

[0108]

[0109]

[0110]

[0111] in, Indicates the first The amplitude expression results corresponding to the fundamental phasor estimation results of the tested voltage transformer under a variable length sampling window; Indicates the first Phase expression results corresponding to the fundamental phasor estimation results of the tested voltage transformer under a variable length sampling window; Indicates the first The amplitude expression results corresponding to the fundamental phasor estimation results of a standard voltage transformer under a variable length sampling window; Indicates the first Phase expression results corresponding to the fundamental phasor estimation results of a standard voltage transformer under a variable length sampling window; Modular operations representing complex quantities are used to extract the amplitude information of the fundamental phasor; The argument operation, representing a complex number, is used to extract the phase information of the fundamental phasor. The principle behind the above formula is that the fundamental phasor in complex form is decomposed into mutually independent but consistent amplitude and phase information, so that subsequent amplitude comparison processing and phase comparison processing can be carried out separately around the corresponding fundamental components under the same window.

[0112] After phasor decomposition, amplitude comparison processing is performed based on the amplitude expression sequence to generate a ratio difference result characterizing the degree of amplitude response deviation, and phase comparison processing is performed based on the phase expression sequence to generate an angle difference result characterizing the degree of phase response shift. The amplitude expression sequence refers to the set of amplitude expression results under multiple variable-length sampling windows arranged along the window advancement direction; the phase expression sequence refers to the set of phase expression results under multiple variable-length sampling windows arranged along the window advancement direction; amplitude comparison processing involves performing a ratio deviation analysis between the amplitude expression results corresponding to the fundamental phasor estimation results of the tested voltage transformer and those corresponding to the fundamental phasor estimation results of the standard voltage transformer within the same window; the ratio difference result is the error quantity characterizing the degree of deviation of the tested channel's response relative to the standard channel in the fundamental amplitude; phase comparison processing involves performing a phase difference analysis between the phase expression results corresponding to the fundamental phasor estimation results of the tested voltage transformer and those corresponding to the fundamental phasor estimation results of the standard voltage transformer; the angle difference result is the error quantity characterizing the degree of shift in the tested channel's response relative to the standard channel in the fundamental phase. The ratio difference result under a variable length sampling window can be expressed as:

[0113]

[0114] in, Indicates the first The ratio difference results under a variable length sampling window are used to reflect the degree of deviation of the fundamental amplitude of the tested voltage transformer from that of the standard voltage transformer. Indicates the first The amplitude expression results corresponding to the fundamental phasor estimation results of the tested voltage transformer under a variable length sampling window; Indicates the first The amplitude expression result corresponding to the fundamental phasor estimation result of the standard voltage transformer under a variable-length sampling window. The principle of this formula is that the difference between the tested amplitude and the standard amplitude is normalized relative to the standard amplitude to form a result that can directly characterize the strength of the amplitude deviation. The angle difference result under a variable length sampling window can be expressed as:

[0115]

[0116] in, Indicates the first The angle difference results under a variable length sampling window are used to reflect the degree of deviation of the fundamental phase of the tested voltage transformer relative to the fundamental phase of the standard voltage transformer. Indicates the first Phase expression results corresponding to the fundamental phasor estimation results of the tested voltage transformer under a variable length sampling window; Indicates the first The phase expression result corresponding to the fundamental phasor estimation result of a standard voltage transformer under a variable length sampling window. The principle of this formula is to directly describe the phase offset relationship using the difference between the two fundamental phases under the same window. Considering the periodic wrap-around characteristic of the phase, in order to avoid the distortion of the angle difference result caused by phase jump, the angle difference result can be further processed by phase normalization, and the expression is:

[0117]

[0118] in, Indicates the first Angular difference results after phase normalization under a variable-length sampling window; This represents the phase normalization function, used to map the original angular difference results to a preset principal value interval, ensuring that the angular difference results remain continuous and comparable. The principle behind this formula is to bring the difference values ​​that cross the phase period boundary back to the unified analysis interval, thereby ensuring that the angular difference results truly reflect the physical phase shift rather than mathematical looping.

[0119] After generating the ratio difference and angle difference results, error validity verification is performed on them to determine the valid error results for the current window. Error validity verification involves checking the rationality, continuity, and consistency of the ratio difference and angle difference results generated within the current window to determine whether these two error quantities can truly represent the voltage transformer metering error state under the current window. Valid error results refer to those error results that, after passing the error validity verification, are deemed suitable for subsequent error feature data recording and construction processes. The server can construct error change smoothness evaluation quantities for the ratio difference and angle difference results separately. For example, for the ratio difference results, the ratio difference continuity evaluation quantity can be defined as:

[0120]

[0121] in, Indicates the first The larger the value of the ratio difference continuity evaluation quantity corresponding to the variable length sampling window, the smoother the change between the current ratio difference result and the ratio difference result of the previous window. Indicates the first Comparison results under a variable-length sampling window; This indicates the difference result under the previous variable-length sampling window; This represents the scale of change in the ratio difference used for normalization. The principle behind this formula is to determine whether there are any abrupt changes in the current ratio difference result by examining the magnitude of change in the ratio difference results between adjacent windows. For angle difference results, the angle difference continuity evaluation quantity can be defined as:

[0122]

[0123] in, Indicates the first The angle difference continuity evaluation quantity corresponding to a variable length sampling window, the larger the value, the smoother the change between the current angle difference result and the angle difference result of the previous window; Indicates the first Angular difference results after phase normalization under a variable-length sampling window; This represents the angular difference result after phase normalization under the previous variable length sampling window; This represents the scale of angle difference variation used for normalization. The principle behind this formula is to determine whether the current angle difference result is affected by sudden noise or abnormal waveforms by examining the magnitude of change in angle difference results between adjacent windows. The server can fuse the ratio difference continuity evaluation quantity, the angle difference continuity evaluation quantity, and the current window phasor consistency state to construct an error validity evaluation quantity:

[0124]

[0125] in, Indicates the first Error validity evaluation quantity corresponding to a variable length sampling window; , , This represents the weighting coefficient, used to adjust the degree of influence of ratio difference continuity, angle difference continuity, and phasor consistency on the overall evaluation; Indicates the first The phasor consistency evaluation value corresponding to each variable-length sampling window is used to reflect whether the fundamental phasor estimation results of the current tested voltage transformer and the fundamental phasor estimation results of the standard voltage transformer meet the basic consistency constraints under the same analysis state. The principle of this formula is to combine the smoothness of the error results with the upstream phasor analysis state to determine whether the current ratio difference and angle difference results are reliable. When the error validity evaluation value reaches a preset threshold, the server confirms the current ratio difference and angle difference results as valid error results under the current window.

[0126] After determining the valid error results in the current window, the corresponding test condition information is simultaneously acquired, and a condition-based binding process is performed between the test condition information and the ratio and angle difference results. Test condition information refers to a set of contextual information characterizing the current inspection environment and operating status, such as primary voltage percentage, secondary load status, ambient temperature, ambient humidity, inspection batch, test mode, and on-site equipment operation identifiers. Condition-based binding refers to establishing a one-to-one correspondence between the valid error results in the current window and the test condition information collected within the same time interval, on the same equipment, and in the same inspection round, giving the error results a clear semantic context. The server can then... The test condition information corresponding to each variable-length sampling window is represented as follows:

[0127]

[0128] in, Indicates the first A set of test condition information corresponding to a variable-length sampling window; Indicates the first The primary voltage percentage information corresponding to each variable-length sampling window is used to characterize the ratio of the current applied primary voltage to the rated value. Indicates the first The secondary load status information corresponding to each variable-length sampling window is used to characterize the current load configuration of the test loop. Indicates the first Ambient temperature information corresponding to a variable-length sampling window; Indicates the first Environmental humidity information corresponding to a variable-length sampling window; Indicates the first The test mode information corresponding to each variable-length sampling window is used to characterize the current test execution method. To ensure the consistency between test condition information and effective error results on the time axis, a condition alignment consistency evaluation metric can be defined as:

[0129]

[0130] in, Indicates the first The larger the value of the working condition alignment consistency evaluation quantity under a variable length sampling window, the more consistent the test working condition information and the effective error result are in time. Indicates the first Time markers corresponding to the effective error results under a variable-length sampling window; Indicates the first Time identifiers corresponding to test condition information under a variable-length sampling window; This represents the time deviation scale used for normalization. The principle behind this formula is to determine whether the effective error result time marker and the test condition information time marker are in the same diagnostic context by using the difference between them. Only when the condition alignment consistency evaluation quantity meets the requirements will the server formally bind the test condition information with the ratio difference result and angle difference result to avoid the problem of mismatch between the error result and the condition background.

[0131] After completing the condition-based binding process, the difference results, angle difference results, and test condition information are uniformly encapsulated to generate error feature data records containing window identifiers, time identifiers, and device identifiers. Uniform encapsulation refers to organizing the error quantity, condition background, and identifier information under the same window into standard record units that can be stored, retrieved, and mapped according to a unified field structure. Error feature data records are structured input data units that will be mapped to the error feature knowledge graph; they contain not only the error value itself but also the contextual semantics and source link when the error occurred. The server can then... The error feature data record corresponding to each variable-length sampling window is represented as follows:

[0132]

[0133] in, Indicates the first Error characteristic data records corresponding to each variable-length sampling window; Indicates the first Each variable-length sampling window has a corresponding window identifier, which is used to uniquely identify the current analysis window; Indicates the first The time identifier corresponding to each variable-length sampling window is used to characterize the time position of the current error feature data record. Indicates the first The device identifier corresponding to each variable-length sampling window is used to characterize which voltage transformer verification object the current error feature data record originates from. Indicates the first The ratio difference results corresponding to each variable-length sampling window; Indicates the first The angle difference results corresponding to each variable-length sampling window; Indicates the first A set of test condition information corresponding to a variable-length sampling window; Indicates the first The set of additional state fields corresponding to each variable-length sampling window is used to record the error validity status, operating condition binding status, and data source status. The principle behind the above expression is that it unifies the result value, context, and identification information into a complete error description unit, allowing subsequent error feature knowledge graph mapping processing to directly obtain complete input without having to backtrack the original analysis chain. To further evaluate whether the encapsulated error feature data record meets the requirements of subsequent graph inference, a record integrity evaluation metric can be defined as:

[0134]

[0135] in, Indicates the first The record integrity evaluation quantity corresponding to a variable-length sampling window; , , This represents the weighting coefficient, used to adjust the degree of influence of field completeness, working condition alignment consistency, and error validity on the overall evaluation; Indicates the first The field integrity evaluation metric corresponding to each variable-length sampling window is used to reflect whether any necessary fields are missing in the error characteristic data record. Indicates the first The working condition alignment consistency evaluation quantity corresponding to a variable length sampling window; Indicates the first The formula is based on an error validity evaluation metric corresponding to a variable-length sampling window. It takes into account structural integrity, semantic consistency, and result validity simultaneously, thereby ensuring that the error feature data records entering the subsequent error feature knowledge graph mapping process are both complete and reliable.

[0136] S150. Map the error feature data records to the error feature knowledge graph, and use a graph neural network to reason about the error type node, influencing factor node and equipment status node to obtain the error anomaly diagnosis result, which includes the anomaly judgment result, the error source location result and the confidence assessment result.

[0137] Specifically, after obtaining the error feature data records, the error feature knowledge graph is first loaded and graph activation processing is performed to ensure that error type nodes, influencing factor nodes, and equipment status nodes are in a mappable state. The error feature knowledge graph is a pre-constructed graph structure knowledge set around the error formation mechanism of voltage transformers. Internally, it organizes the relationships between error manifestations, operating conditions, and equipment status in the form of nodes and associated edges. Error type nodes are graph nodes used to characterize ratio anomalies, angle anomalies, composite anomalies, and other error manifestation categories. Influencing factor nodes are graph nodes used to characterize primary voltage percentage, secondary load status, ambient temperature, ambient humidity, external interference status, and other test operating condition information. Equipment status nodes are graph nodes used to characterize core status, winding status, wiring status, insulation status, and other internal equipment statuses. Graph activation processing involves switching the set of nodes, the set of associated edges, and the initial attributes of the nodes involved in the current diagnostic task from a static storage state to an online state that can participate in mapping, propagation, and inference. At the edge computing node of the HarmonyOS system, the server can represent the error feature knowledge graph as follows:

[0138]

[0139] in, Represents a knowledge graph of error characteristics; This represents the set of nodes in the graph, which includes the set of error type nodes, the set of influencing factor nodes, and the set of equipment status nodes. This represents the set of associated edges in the graph, used to describe whether a diagnostic association exists between any two nodes; This represents the set of associated edge weights, used to characterize the influence strength, support strength, or coupling strength corresponding to the associated edges. The principle behind the above expression is that it abstracts the error feature knowledge graph into a structured knowledge space composed of nodes, connections, and connection strengths, enabling subsequent error feature data records to be mapped based on a defined topology. To determine whether the graph activation process is complete, the mappable node coverage rate for the current diagnostic task can be defined as:

[0140]

[0141] in, This represents the mappable node coverage, which reflects whether the nodes required for the current diagnostic task are in a mappable state. Its value ranges from greater than or equal to zero to less than or equal to one. The closer the value is to one, the more fully the graph is activated. This indicates the number of nodes that are currently active and can participate in the mapping process; This indicates the number of nodes theoretically required for the diagnostic task corresponding to the current error feature data record. The principle behind this formula is to measure whether the graph is ready for mapping by using the ratio of the number of activated nodes to the number of required nodes. When the mappable node coverage meets the threshold, it indicates that error type nodes, influencing factor nodes, and device status nodes are all in a mappable state, and subsequent field decomposition and semantic processing can proceed.

[0142] After the graph activation process is complete, field decomposition and semantic organization are performed on the error feature data records to generate error performance features, operating condition influence features, and context constraint features. Error feature data records refer to structured data units composed of comparison results, angle difference results, test operating condition information, window identifiers, time identifiers, and device identifiers from the previous processing stage. Field decomposition and semantic organization refers to re-splitting and merging the multi-field data, originally encapsulated in a record format, according to diagnostic semantics into a feature set that directly corresponds to the graph node type. Error performance features refer to a feature set composed of comparison results, angle difference results, and their derived states that directly describe the current error morphology. Operating condition influence features refer to a feature set composed of primary voltage percentage, secondary load status, ambient temperature, ambient humidity, and other test operating condition information that characterizes the external influence conditions of the error. Context constraint features refer to a feature set composed of window identifiers, time identifiers, device identifiers, test rounds, and analysis status identifiers used to limit the boundaries and scope of this diagnosis. The server can then... The error characteristic data records corresponding to each variable-length sampling window are denoted as follows: The field decomposition results are expressed as follows:

[0143]

[0144]

[0145]

[0146] in, Indicates the first Error characteristics corresponding to a variable-length sampling window; Indicates the first The ratio difference results corresponding to each variable-length sampling window; Indicates the first The angle difference results corresponding to each variable-length sampling window; Indicates the first Error validity evaluation quantity corresponding to a variable length sampling window; Indicates the first The error state description quantity corresponding to each variable-length sampling window is used to characterize whether the current error is in a stable, abrupt, or complex fluctuation state. Indicates the first The operating condition influence characteristics corresponding to a variable-length sampling window; Indicates the percentage of primary voltage; Indicates the secondary load status; Indicates ambient temperature; Indicates ambient humidity; Indicates the test mode; Indicates the first Context constraint features corresponding to a variable-length sampling window; Indicates the window identifier; Indicates time stamp; Indicates equipment identification; Indicates the test round identifier; This indicates the analysis status identifier. The principle behind the above expression is to reorganize the numerical fields, operating condition fields, and identifier fields in the error feature data record into feature groups with clear diagnostic semantics, thereby providing a unified, clear, and graph-oriented input expression for subsequent node mapping processing.

[0147] After obtaining the error performance features, operating condition impact features, and context constraint features, node mapping processing is performed based on these features to form corresponding local diagnostic subgraphs. Node mapping processing refers to finding the most relevant error type nodes, influencing factor nodes, and equipment status nodes in the error feature knowledge graph based on the semantic content of the current error feature, and establishing the connection between the current feature and candidate nodes. A local diagnostic subgraph is a small-scale diagnostic graph structure extracted from the entire error feature knowledge graph and focused solely on the current error feature data record. It retains nodes and edges highly relevant to the current error performance, current test conditions, and current equipment status, thereby reducing interference from irrelevant knowledge and improving reasoning focus. The server can first construct the similarity between the error performance features, operating condition impact features, and context constraint features and the graph nodes, respectively. Let any feature vector be denoted as... The prototype vector of any graph node is denoted as The mapping similarity between the two can be expressed as:

[0148]

[0149] in, Indicates the first The input feature and the first The mapping similarity between the prototypes of graph nodes is a value that ranges from negative one to positive one. The larger the value, the closer the two are semantically. This represents the current input feature vector, which can correspond to any element among error performance features, operating condition influence features, or context constraint features; Indicating the first in the atlas The prototype vector of each node is used to represent the semantic center of that node; This represents the dot product of two vectors; and Let represent the magnitudes of the two vectors respectively. The principle behind this formula is that it measures the semantic similarity between the feature vector and the node prototype vector by the degree of directional consistency, thus providing a quantitative basis for node mapping. Let the set of nodes that satisfy the similarity threshold be . The corresponding set of associated edges is Then the first The local diagnostic sub-map corresponding to a variable-length sampling window can be represented as:

[0150] ; in, Indicates the first Local diagnostic sub-images corresponding to a variable-length sampling window; Represents the set of nodes that have been mapped and retained; This represents the set of edges connecting these nodes; This represents the set of weights corresponding to these associated edges. The principle behind this expression is to extract the most relevant local structures for the current diagnosis from the global error feature knowledge graph, enabling subsequent graph neural network inference to unfold within a more compact and targeted knowledge scope.

[0151] After the local diagnostic subgraph is formed, feature injection processing is performed on it to write error performance features, operating condition influence features, and context constraint features into the error type node, influencing factor node, and equipment status node, respectively. Feature injection processing refers to writing the real-time feature values ​​of the current object to be diagnosed into the attributes of the selected local diagnostic subgraph nodes, enabling static nodes that originally only stored prior knowledge to acquire the dynamic diagnostic status under the current test round. Writing to the error type node means writing error performance features such as the ratio difference result, angle difference result, error effectiveness evaluation quantity, and error status description quantity into the error type node attributes. Writing to the influencing factor node means writing operating condition influence features such as primary voltage percentage, secondary load status, ambient temperature, ambient humidity, and test mode into the influencing factor node attributes. Writing to the equipment status node means writing context constraint features such as equipment identifier, time identifier, test round identifier, and analysis status identifier into the equipment status node attributes, ensuring that the equipment status node remains bound to the current diagnostic object during subsequent inference. Let the first... The original node attribute vector of each node is The injected node attribute vector is The injection relationship can then be represented as:

[0152]

[0153] in, Indicates the first The first variable-length sampling window The node attribute vector of each node after feature injection processing; Indicates the first The original node attribute vector of each node before feature injection processing; Indicates the first The feature injection mapping matrix corresponding to each node is used to transform the current input features into the node attribute space; Indicates the first Under the variable length sampling window and the first The formula is based on the input feature vector corresponding to each node. The principle of this formula is to superimpose and fuse static prior knowledge and current dynamic diagnostic information in the same node attribute space, so that the graph neural network can utilize the long-accumulated knowledge structure and perceive the real-time error state and working condition background of the current test object during propagation.

[0154] After feature injection, graph neural network inference is performed based on the local diagnostic subgraph. Anomaly detection is performed on error type nodes to generate anomaly detection results, correlation analysis is performed on influencing factor nodes and equipment status nodes to generate error source location results, and confidence assessment is performed on the anomaly detection results and error source location results to generate confidence assessment results. Graph neural network inference refers to performing multiple rounds of neighborhood information propagation, aggregation, and state updates on the local diagnostic subgraph, enabling each node to form a new node representation by integrating the attribute states of its neighboring nodes and the weights of the associated edges. Anomaly detection refers to determining whether the current test object has an anomaly and what type of error it belongs to based on the updated node representations of error type nodes. Correlation analysis refers to evaluating the support strength of various operating conditions and equipment status for the current anomaly based on the updated node representations of influencing factor nodes and equipment status nodes, thereby locating the error source. Confidence assessment refers to quantifying the credibility of the anomaly detection results and error source location results. Let the... The first layer after propagation Each node is characterized as Then, the update of one round of propagation in a graph neural network can be represented as:

[0155]

[0156] in, Indicates the first The node at the th Node representation after layer propagation; This represents a non-linear activation function used to enhance the model's ability to characterize complex relationships; Indicates the first The linear transformation matrix of the current node's own representation during layer propagation; Indicates the first The linear transformation matrix representing adjacent nodes in layer propagation; Indicates the first The set of adjacent nodes of a node; Indicates the first In the propagation of layers from the first layer The node passes the message to the first node. The information weight of each node is used to characterize the strength of the influence of neighboring nodes on the current node; Indicates the first The node at the th Node representation after layer propagation. The principle of this formula is that by integrating its own information and neighbor information, error type nodes gradually absorb the operating condition influence and equipment status information associated with them, influencing factor nodes gradually absorb the error performance and equipment status information associated with them, and equipment status nodes gradually absorb the error performance and operating condition influence information associated with them, thereby forming a comprehensive node representation oriented towards the current diagnostic task.

[0157] After multiple rounds of propagation, anomaly detection processing can be performed on the representation of the error type node set. Let the final propagated aggregated representation of the error type nodes be denoted as... The anomaly detection output can then be represented as:

[0158]

[0159] in, Indicates the first Anomaly category probability distribution corresponding to a variable-length sampling window; This represents a normalized classification function used to convert scores for different anomaly categories into a probability distribution. This represents the anomaly classification matrix; This represents the anomaly detection bias vector; Indicates the first The formula aggregates the error type nodes corresponding to a variable-length sampling window. The principle behind this formula is to map the comprehensive representation of error type nodes to a competitive score of multiple anomaly categories, then normalize it to a probability distribution, thereby outputting the anomaly determination result. For error source localization, correlation analysis can be performed on the final representations of the influencing factor node set and the equipment status node set, respectively. If the final representation of the influencing factor nodes is denoted as... The final representation of the device state node is denoted as Then its source support score can be expressed as:

[0160]

[0161]

[0162] in, Indicates the first The first variable-length sampling window Each influencing factor node provides a support score for the source of the current anomaly; Indicates the first The first variable-length sampling window Each device status node provides a support score for the source of the current anomaly; This represents a compression function used to map source support scores to a range between zero and one. and These represent the scoring weight vectors corresponding to the influencing factor node and the device status node, respectively. and These represent the scoring biases corresponding to the influencing factor node and the device status node, respectively. Indicates the first The final representation of each influencing factor node; Indicates the first The final representation of each device status node. The principle of this formula is to use a unified scoring model to measure the explanatory power of each influencing factor node and device status node for the current anomaly diagnosis, thereby generating a ranking of candidate sources in the error source localization results. For confidence assessment processing, the anomaly determination results and error source localization results can be jointly incorporated into the comprehensive evaluation, and the confidence assessment result is defined as:

[0163]

[0164] in, Indicates the first Confidence assessment results corresponding to each variable-length sampling window; , , This represents the weighting coefficient, used to adjust the proportion of the influence of anomaly detection strength, source support strength, and input record integrity in the total confidence score; This represents the maximum probability value in the probability distribution of the anomaly category, used to reflect the degree of certainty in the anomaly determination result; The comprehensive value representing the source support score can be obtained by aggregating the support scores of the top-ranked influencing factor nodes and device status nodes; Indicates the first The formula is a record integrity evaluation metric corresponding to a variable-length sampling window. The principle behind this formula is that it simultaneously incorporates the class certainty, source interpretability, and input data quality of the model output into the confidence assessment, ensuring that the generated confidence assessment results reflect not only the strength of the model's judgment but also the reliability of the input data.

[0165] After obtaining the anomaly detection results, error source location results, and confidence assessment results, a result fusion process is performed on these results to generate an error anomaly diagnosis result. Result fusion processing refers to integrating diagnostic outputs from different sources and semantic levels into a unified structured diagnostic unit, allowing subsequent differentiated processing strategies to be directly invoked. The error anomaly diagnosis result is the final output record that comprehensively reflects whether the current test object is abnormal, the possible influencing factors and equipment status from which the anomaly may originate, and the current diagnostic reliability. The server can then... The error anomaly diagnosis results corresponding to each variable-length sampling window are represented as follows:

[0166]

[0167] in, Indicates the first Error anomaly diagnosis results corresponding to each variable-length sampling window; Indicates the window identifier; Indicates time stamp; Indicates equipment identification; This indicates the anomaly determination result, used to describe whether the error corresponding to the current window is normal and what type of anomaly it belongs to; This indicates the error source location result and is used to describe the sorted list of the main influencing factor nodes and device status nodes corresponding to the current window. This represents the confidence assessment result. The principle behind this expression is to encapsulate the anomaly determination result, error source location result, and confidence assessment result, along with the identification field, into a complete diagnostic output record. This allows the current error feature data record, after being mapped through the error feature knowledge graph and inferred by the graph neural network, to ultimately be transformed into an error anomaly diagnostic result that can directly drive subsequent differentiated processing strategies.

[0168] S160. Implement a differentiated processing strategy for the error anomaly diagnosis results and the original sampling sequence to complete the real-time diagnosis of the voltage transformer based on the HarmonyOS system.

[0169] Specifically, after obtaining the error anomaly diagnosis result, the first step is to perform result association confirmation processing between the error anomaly diagnosis result and the original sampling sequence to establish a correspondence under the same window identifier, time identifier, and device identifier. The error anomaly diagnosis result refers to the structured diagnostic output obtained after mapping the preceding error feature knowledge graph and inference using a graph neural network; the original sampling sequence refers to the secondary voltage sampling sequence of the tested voltage transformer and the secondary voltage sampling sequence of the standard voltage transformer, which are accessed from the voltage transformer calibrator and retained after time synchronization alignment and joint buffering; the result association confirmation processing refers to matching the diagnostic conclusion with the data source one-to-one according to a unified identifier system, ensuring that any diagnostic conclusion can be traced back to a unique original sampling interval. To achieve this processing, the server first reads the window identifier, time identifier, and device identifier from the error anomaly diagnosis result, then retrieves the sampling segments corresponding to the same set of identifiers from the original sampling sequence buffer, and constructs an association record. The association relationship can be represented as:

[0170]

[0171] in, Indicates the first The associated results for each window; Indicates the first The window identifier corresponding to each window; Indicates the first The time identifier corresponding to each window; Indicates the first The device identifier corresponding to each window; Indicates the first Error anomaly diagnosis results corresponding to each window; Indicates the first Each window corresponds to a segment of the original sampled sequence. The principle behind this formula is that by binding diagnostic results with the same identifier to the sampled data within the same recording unit, a unified processing entry point for subsequent differentiated processing strategies is formed. To determine whether the association is successful, an association consistency evaluation metric can also be defined:

[0172]

[0173] in, Indicates the first The association consistency evaluation value corresponding to each window ranges from zero to one. The closer the value is to one, the more accurate the association is. This indicates the window identifier matching status; a value of 1 is used when the window identifiers match, and a value of 0 is used otherwise. This indicates the time identifier matching status; a value of 1 is used if the time identifiers match, and a value of 0 is used otherwise. This indicates the device identifier matching status; a value of 1 is used if the device identifiers match, and 0 otherwise. The principle behind this formula is to measure whether a reliable correspondence has been established between the current error anomaly diagnosis result and the original sampling sequence by using the item-by-item matching results of the identifier field.

[0174] After completing the result correlation confirmation process, a diagnostic level discrimination process is performed based on the error anomaly diagnosis results to generate diagnostic level labels that drive differentiated processing strategies. Diagnostic level discrimination process refers to the hierarchical determination of the risk level, processing priority, and data retention strength of the current diagnostic task based on the anomaly determination results, error source location results, and confidence assessment results. The diagnostic level label refers to the classification label used for subsequent selection of processing paths and allocation of system resources. The server can integrate the anomaly category strength, source complexity, and confidence level to construct a diagnostic level score. The diagnostic level score can be expressed as:

[0175]

[0176] in, Indicates the first Each window corresponds to a diagnostic grade score; , , This represents the weighting coefficient, used to adjust the proportion of influence of anomaly intensity, source complexity, and confidence level in the total score; Indicates the first The anomaly intensity score corresponding to each window is used to characterize the severity of the anomaly determination result; Indicates the first The source complexity score corresponding to each window is used to characterize the number of factors, the depth of association, or the scope of influence involved in the error source localization result; Indicates the first The confidence assessment results correspond to each window. The principle behind this formula is to integrate the degree of abnormality, source complexity, and model confidence of the diagnostic conclusion into a single-level score, allowing different diagnostic states to be ranked on a unified scale. Based on this score, the server can perform segmented mapping, dividing the current diagnostic task into a routine processing level, a key processing level, or an intensive processing level. The diagnostic level identifier can be represented as:

[0177]

[0178] in, Indicates the first Each window corresponds to a diagnostic level indicator; Indicates the standard processing level; Indicates the level of priority handling; Indicates the level of enhanced processing; and This represents the threshold for classifying diagnostic levels, and The principle behind this formula is that it maps continuous scores to discrete levels through threshold segmentation, so that subsequent differentiated processing strategies can be directly applied.

[0179] After obtaining the diagnostic level identifier, data value assessment processing is performed on the original sampled sequence based on the diagnostic level identifier to determine the retention range and processing granularity of the original sampled sequence, resulting in the data value assessment processing result. Data value assessment processing refers to the quantitative judgment made based on the diagnostic level, anomaly evolution location, and error source localization results to determine which segments in the original sampled sequence have retention value, which segments only require summary retention, and which segments need to be retained in their entirety. The retention range refers to the length of the sampling interval that should be saved or uploaded on the timeline. The processing granularity refers to which of the three different processing depths—full processing, key segment processing, or summary feature processing—is applied to the original sampled sequence. Servers typically include anomaly occurrence segments, anomaly preceding segments, and anomaly recovery segments in a unified assessment to avoid losing evolutionary context by only retaining the anomaly center segment. The data value score can be expressed as:

[0180]

[0181] in, Indicates the first The data value score corresponding to each window; , , This represents the weighting coefficient, used to adjust the proportion of influence of diagnostic level, source complexity, and original data integrity in the total score; This represents the level mapping value corresponding to the diagnostic level identifier; the higher the level, the larger the mapping value. Indicates the first The source complexity score corresponding to each window; Indicates the first The integrity evaluation metric for the original sampling sequence corresponding to each window is used to characterize whether the original sampling segment is continuous, without missing data, and synchronized across both channels. The principle behind this formula is that diagnostic importance, source interpretation complexity, and original data quality jointly determine the retention value of the current sampled data. Based on this score, the server can further determine the retention length:

[0182]

[0183] in, Indicates the first The retention length of the sampled data corresponding to each window; Indicates the basic retention length; This represents the scaling factor from the score to the retention length; Indicates the first Each window corresponds to a data value score. The principle behind this formula is to allow high-value data to have a longer context retention period, while low-value data is only retained in the smallest traceable fragment, thereby improving storage and transmission efficiency.

[0184] After obtaining the data value assessment results, hierarchical encapsulation processing is performed on the error anomaly diagnosis results and the original sampled sequences to generate diagnostic result encapsulation units and sampled data encapsulation units. Hierarchical encapsulation processing refers to encapsulating the diagnostic results and original sampled sequences using different field sizes and content depths based on the diagnostic level identifier and the data value assessment results. The diagnostic result encapsulation unit is a structured output unit that uniformly organizes the anomaly determination results, error source location results, confidence assessment results, and corresponding identifier fields. The sampled data encapsulation unit is a data output unit that organizes the original sampled sequences according to the retention range and processing granularity. For the diagnostic result encapsulation unit, the server can construct:

[0185]

[0186] in, Indicates the first Each window corresponds to a diagnostic result encapsulation unit; Indicates the first Each window corresponds to a diagnostic level indicator; Indicates the first The anomaly detection results corresponding to each window; Indicates the first The error source location results corresponding to each window; Indicates the first The confidence assessment results correspond to each window. For the sampled data encapsulation unit, the server can extract the corresponding segment according to the retention length and processing granularity, and construct:

[0187]

[0188] in, Indicates the first Each window corresponds to a sampling data encapsulation unit; Indicates the first The retention length of the sampled data corresponding to each window; Indicates the first Each window has a corresponding processing granularity identifier, used to distinguish between full retention, key segment retention, and summary feature retention; This represents the sampled data content after being extracted and organized according to the retention range. The principle behind the two formulas above is to convert diagnostic conclusions and sampled data into standard units that can be independently scheduled, independently stored, and jointly uploaded, providing a unified data interface for subsequent processing path selection and task allocation.

[0189] After generating the diagnostic result encapsulation unit and the sampled data encapsulation unit, resource awareness processing is performed on the collaborative domain where the HarmonyOS edge computing node is located to obtain network link status, local processing load status, and collaborative node idle status, thus obtaining the resource awareness processing results. Resource awareness processing refers to the dynamic detection and quantitative evaluation of available communication resources, computing resources, and node resources in the current collaborative domain; network link status refers to the bandwidth, latency, jitter, and packet loss levels between the current edge computing node and the cloud or collaborative nodes; local processing load status refers to the processor usage, cache usage, task queue length, and power consumption levels of the current edge computing node; and collaborative node idle status refers to whether other HarmonyOS edge computing nodes within the same collaborative domain have the idle capacity to handle tasks. The server can manage a single collaborative node... The comprehensive resource index is expressed as:

[0190]

[0191] in, Indicates the first The comprehensive resource index of each collaborative node; , , , This represents the weighting coefficient, used to adjust the proportion of influence of bandwidth status, latency status, load status, and idle status in resource evaluation; Indicates the first The normalized available bandwidth level of each cooperating node; the larger the value, the more sufficient the bandwidth. Indicates the first The normalized latency level of each coordinating node; a larger value indicates greater latency, therefore, the normalized latency level is adopted. This demonstrates the advantage of low latency; Indicates the first The normalized local load level of each cooperating node; a larger value indicates a heavier load, therefore... This indicates the advantage of low load. Indicates the first The idle status score of each cooperating node is used to determine its suitability for undertaking offloading tasks. The principle behind this formula is to integrate network quality and node processing capacity into a single resource metric, so as to quickly select the target node most suitable for participating in the differentiated processing strategy.

[0192] After obtaining the resource awareness processing results, processing path selection is performed based on the diagnostic level identifier, data value assessment processing results, and resource awareness processing results to determine the local storage processing path, summary upload processing path, or enhanced upload processing path, resulting in the processing path selection processing result. Processing path selection processing refers to determining the final processing flow for the current diagnostic task under the constraints of three factors: diagnostic importance, data value, and system resources. The local storage processing path means that archiving and caching are completed only within the current HarmonyOS edge computing node, without remote upload. The summary upload processing path means that only the diagnostic conclusions, key status fields, and representative sample segments are uploaded. The enhanced upload processing path means that the diagnostic conclusions and the complete sample segment covering the abnormal evolution process are uploaded, and cloud-based deep analysis is triggered first. The server can construct a processing path scoring function:

[0193]

[0194] in, Indicates the first The window in the first Scoring on each candidate processing path; , , This represents a weighting coefficient used to adjust the proportion of influence of diagnostic level, data value, and resource status in path selection; Indicates the diagnostic grade label for the first An adaptation function for each candidate processing path; Indicates the data value score for the first An adaptation function for each candidate processing path; This indicates that the resource-aware processing result is related to the first... An adaptation function for each candidate processing path; Indicates the first The formula calculates the overall resource awareness result for each window. The principle behind this formula is to calculate the overall adaptability for different processing paths, and then select the path with the highest score as the actual execution path. The final path selection result can be expressed as:

[0195]

[0196] in, Indicates the first Each window corresponds to a processing path for selecting the processing result; value one represents the local storage processing path, value two represents the summary upload processing path, and value three represents the enhanced upload processing path. This indicates the candidate path number that yields the highest score. The principle behind this formula is to automatically select the current optimal processing path based on the maximum fit score principle.

[0197] After obtaining the processing path selection result, task allocation processing is performed on the diagnostic result encapsulation unit and the sampled data encapsulation unit based on the processing path selection result. This allows data processing and task offloading to be performed between edge computing nodes or collaborative nodes of the HarmonyOS system, generating task execution results. Task allocation processing refers to the reasonable allocation of the encapsulated diagnostic results and sampled data processing tasks to local nodes or collaborative nodes for execution based on the selected processing path, the current local node load, and the resource status of collaborative nodes. Task offloading refers to migrating processing subtasks such as data compression, feature extraction, encapsulation verification, fragmented upload preparation, or intermediate statistical analysis from the current node to collaborative nodes. The server can first define the... The candidate node pair Estimated completion time for each task:

[0198]

[0199] in, Indicates the first The candidate node executes the first... The estimated completion time for each task; This indicates that the task data will be transferred to the first... The estimated time required for each candidate node; Indicates the first The estimated computation time required for each candidate node to process this task; Indicates the first The estimated time required for each candidate node to return intermediate results to its local node. The principle behind this formula is to uniformly calculate the transmission, computation, and return stages into a total completion time, used to select the optimal unloading node. The optimal task allocation node can be represented as:

[0200]

[0201] in, Indicates the first The optimal allocation node number corresponding to each task; This indicates the node number that minimizes the estimated completion time. The principle behind this formula is to prioritize the local or collaborating node that can complete the current task fastest, thereby reducing overall processing latency. After the task is completed, the server receives the task execution results, including the local processing results, the results returned by the collaborating nodes, and the corresponding status flags.

[0202] After obtaining the task execution result, the result is processed through result synthesis and consistency verification to generate a valid processing result. Result synthesis refers to reassembling the intermediate processing results, compressed results, fragmented results, and status results output by local and collaborating nodes according to a unified identifier to form a complete processing output. Consistency verification checks whether the synthesized result is consistent with the original diagnostic result encapsulation unit and sampled data encapsulation unit in terms of identifier, content, order, and statistical characteristics. A valid processing result is the final processing result that, after consistency verification, is deemed suitable for direct output or upload. The server can construct a consistency verification score:

[0203]

[0204] in, Indicates the first The output consistency score corresponds to each window; , , This represents the weighting coefficient, used to adjust the proportion of influence of identifier consistency, sequence consistency, and content consistency in the overall score; Indicates the first The consistency evaluation value for each window is used to check whether the window identifier, time identifier, and device identifier are consistent. Indicates the first The sequential consistency evaluation quantity corresponding to each window is used to check whether the data fragment splicing order is correct; Indicates the first The content consistency evaluation score corresponding to each window is used to check whether the statistical content is reasonably consistent before and after compression, fragmentation, and aggregation. The principle of this formula is to jointly judge whether the current processing output is reliable through multi-dimensional consistency indicators. When the output consistency score is not lower than the threshold, the server confirms the current result as a valid processing result; if the output consistency score is lower than the threshold, the relevant tasks are backtracked and re-executed.

[0205] After generating valid processing results, final output processing is performed based on these results to complete real-time diagnostics of voltage transformers based on the HarmonyOS system. Final output processing refers to selecting processing results according to the processing path, writing valid processing results to local storage, sending them to the cloud, or simultaneously completing local archiving and remote upload. Completing real-time diagnostics of voltage transformers based on the HarmonyOS system means that the current diagnostic task forms a closed loop from raw sampling access, spectrum analysis, error feature generation, spectrum inference to differentiated processing and result delivery. If the current processing path is a local storage processing path, the server writes the valid processing results to the edge diagnostic result library and creates an index; if the current processing path is a summary upload processing path, the server uploads key segments from the diagnostic result encapsulation unit and the sample data encapsulation unit to the cloud collaborative analysis area; if the current processing path is an enhanced upload processing path, the server uploads the complete valid processing results to the cloud deep analysis area and retains the local receipt. The final output result can be represented as:

[0206]

[0207] in, Indicates the first The final output results corresponding to each window; Indicates the first Each window corresponds to a processing path selection and processing result; Indicates the first Each window corresponds to a diagnostic result encapsulation unit; Indicates the first Each window corresponds to an output data unit that may correspond to local storage content, summary upload content, or enhanced upload content. Indicates the first The set of task execution statuses corresponding to each window is used to record the processing completion status, upload status, and collaborative node participation status. The principle of this formula is to encapsulate the output content, output path, and execution status of the entire differentiated processing strategy into a single result unit, thereby enabling the real-time diagnosis of the current window to be directly traced, verified, and invoked by subsequent systems.

[0208] This application also provides a real-time diagnostic device for voltage transformers based on the HarmonyOS system, referring to... Figure 2 , Figure 2This application provides a schematic diagram of a real-time voltage transformer diagnostic device based on the HarmonyOS system. The device is a server, comprising an acquisition module 21 and a processing module 22. The acquisition module 21 acquires, at the edge computing node of the HarmonyOS system, the original sampling sequence sent by the voltage transformer calibrator, containing the secondary voltage sampling sequence of the tested voltage transformer and the secondary voltage sampling sequence of the standard voltage transformer via a distributed soft bus, and constructs an adaptive sliding window. The processing module 22 extracts a phase reference signal from the secondary voltage sampling sequence of the standard voltage transformer according to the adaptive sliding window, dynamically estimates the actual operating frequency of the power grid, and generates a variable-length sampling window synchronized with the actual operating frequency of the power grid. The processing module 22 is also used to process the secondary voltage sampling sequence of the tested voltage transformer and the standard voltage transformer according to the variable-length sampling window. The secondary voltage sampling sequences are subjected to windowed interpolation spectrum analysis to obtain the fundamental phasor estimation results of the tested voltage transformer and the standard voltage transformer. The processing module 22 is also used to calculate the ratio difference and angle difference results based on the fundamental phasor estimation results of the tested voltage transformer and the standard voltage transformer, and to form error feature data records in combination with the test condition information. The processing module 22 is also used to map the error feature data records to the error feature knowledge graph, and to perform reasoning on the error type nodes, influencing factor nodes and equipment status nodes through graph neural networks to obtain error anomaly diagnosis results including anomaly judgment results, error source location results and confidence assessment results. The processing module 22 is also used to perform differentiated processing strategies on the error anomaly diagnosis results and the original sampling sequences to complete the real-time diagnosis of the voltage transformer based on the HarmonyOS system.

[0209] It should be noted that the above embodiments of the apparatus are only illustrated by the division of the above functional modules. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. In addition, the apparatus and method embodiments provided in the above embodiments belong to the same concept, and the specific implementation process can be found in the method embodiments, which will not be repeated here.

[0210] This application also provides an electronic device, with reference to... Figure 3 , Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. The electronic device may include: at least one processor 31, at least one network interface 34, a user interface 33, a memory 35, and at least one communication bus 32.

[0211] The communication bus 32 is used to enable communication between these components.

[0212] The user interface 33 may include a display screen and a camera. Optionally, the user interface 33 may also include a standard wired interface and a wireless interface.

[0213] The network interface 34 may optionally include a standard wired interface or a wireless interface (such as a Wi-Fi interface).

[0214] The processor 31 may include one or more processing cores. The processor 31 connects to various parts of the server via various interfaces and lines, executing instructions, programs, code sets, or instruction sets stored in the memory 35, and calling data stored in the memory 35 to perform various server functions and process data. Optionally, the processor 31 may be implemented using at least one hardware form of Digital Signal Processing (DSP), Field-Programmable Gate Array (FPGA), or Programmable Logic Array (PLA). The processor 31 may integrate one or a combination of several of the following: Central Processing Unit (CPU), Graphics Processing Unit (GPU), and modem. The CPU primarily handles the operating system, user interface, and applications; the GPU is responsible for rendering and drawing the content to be displayed on the screen; and the modem handles wireless communication. It is understood that the modem may also not be integrated into the processor 31 and may be implemented as a separate chip.

[0215] The memory 35 may include random access memory (RAM) or read-only memory. Optionally, the memory 35 may include a non-transitory computer-readable storage medium. The memory 35 can be used to store instructions, programs, code, code sets, or instruction sets. The memory 35 may include a program storage area and a data storage area, wherein the program storage area may store instructions for implementing an operating system, instructions for at least one function (such as touch function, sound playback function, image playback function, etc.), instructions for implementing the above-described method embodiments, etc.; the data storage area may store data involved in the above-described method embodiments, etc. Optionally, the memory 35 may also be at least one storage device located remotely from the aforementioned processor 31. Figure 3As shown, the memory 35, which serves as a computer storage medium, may include an operating system, a network communication module, a user interface module, and an application program for a real-time diagnostic method for voltage transformers based on the HarmonyOS system.

[0216] exist Figure 3 In the electronic device shown, the user interface 33 is mainly used to provide an input interface for the user and to obtain the user input data; while the processor 31 can be used to call the application program stored in the memory 35, which is a real-time diagnostic method for voltage transformers based on the HarmonyOS system. When executed by one or more processors, the electronic device executes one or more methods as described in the above embodiments.

[0217] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, as some steps may be performed in other orders or simultaneously according to this application. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to this application.

[0218] This application also provides a non-transitory computer-readable storage medium storing instructions. When executed by one or more processors, these instructions cause an electronic device to perform one or more of the methods described in the above embodiments.

[0219] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0220] In the several embodiments provided in this application, it should be understood that the disclosed apparatus can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the shown or discussed mutual couplings or direct couplings or communication connections may be through some service interfaces; indirect couplings or communication connections between apparatuses or units may be electrical or other forms.

[0221] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0222] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0223] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage device (CMD). Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a memory and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned memory includes various media capable of storing program code, such as USB flash drives, portable hard drives, magnetic disks, or optical disks.

[0224] The foregoing description is merely an exemplary embodiment of this disclosure and should not be construed as limiting the scope of this disclosure. Any equivalent changes and modifications made in accordance with the teachings of this disclosure shall still fall within the scope of this disclosure. Those skilled in the art will readily conceive of other embodiments of this disclosure upon considering the specification and the disclosure of practical truth. This application is intended to cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary techniques in the art not described in this disclosure. The specification and embodiments are considered exemplary only, and the scope and spirit of this disclosure are defined by the claims.

Claims

1. A real-time diagnostic method for voltage transformers based on the HarmonyOS system, characterized in that, The method includes: At the edge computing node of the HarmonyOS system, the original sampling sequence sent by the voltage transformer calibrator, which includes the secondary voltage sampling sequence of the voltage transformer under test and the secondary voltage sampling sequence of the standard voltage transformer, is obtained through the distributed soft bus, and an adaptive sliding window is constructed. Based on the adaptive sliding window, a phase reference signal is extracted from the secondary voltage sampling sequence of the standard voltage transformer, and the actual operating frequency of the power grid is dynamically estimated to generate a variable-length sampling window that is synchronized with the actual operating frequency of the power grid. Based on the variable-length sampling window, windowed interpolation spectrum analysis is performed on the secondary voltage sampling sequence of the tested voltage transformer and the secondary voltage sampling sequence of the standard voltage transformer to obtain the fundamental phasor estimation results of the tested voltage transformer and the fundamental phasor estimation results of the standard voltage transformer. Based on the fundamental phasor estimation results of the tested voltage transformer and the fundamental phasor estimation results of the standard voltage transformer, the ratio difference and angle difference results are calculated, and error characteristic data records are formed by combining the test condition information. The error feature data records are mapped to the error feature knowledge graph. The graph neural network is used to reason about the error type node, influencing factor node and equipment status node to obtain the error anomaly diagnosis result, which includes the anomaly judgment result, the error source location result and the confidence assessment result. A differentiated processing strategy is applied to the error anomaly diagnosis results and the original sampling sequence to complete the real-time diagnosis of the voltage transformer based on the HarmonyOS system.

2. The real-time diagnostic method for voltage transformers based on the HarmonyOS system according to claim 1, characterized in that, At the edge computing node of the HarmonyOS system, the original sampling sequence sent by the voltage transformer calibrator, which includes the secondary voltage sampling sequence of the tested voltage transformer and the secondary voltage sampling sequence of the standard voltage transformer, is obtained through a distributed soft bus, and an adaptive sliding window is constructed, specifically including: At the edge computing node of the HarmonyOS system, the distributed soft bus is used to perform device discovery and connection establishment for the voltage transformer calibrator in the collaborative domain, obtain the original sampling sequence containing the secondary voltage sampling sequence of the tested voltage transformer and the secondary voltage sampling sequence of the standard voltage transformer, and perform channel registration and field binding on the secondary voltage sampling sequence of the tested voltage transformer and the secondary voltage sampling sequence of the standard voltage transformer to form a dual-channel synchronous data structure. The dual-channel synchronous data structure is subjected to time synchronization alignment, integrity check and order rearrangement processing to generate a time-continuous original sampling sequence corresponding to each channel; Based on the original sampling sequence, a joint cache organization is performed so that the secondary voltage sampling sequence of the tested voltage transformer and the secondary voltage sampling sequence of the standard voltage transformer form corresponding storage segments according to a unified time axis, and window boundary markers and data validity markers are added to the corresponding storage segments. Based on the corresponding storage segment, an initial window boundary is generated according to the preset window coverage and window update rhythm, and the adaptive sliding window is constructed by retaining overlapping data segments and accessing newly added original sampling sequences.

3. The real-time diagnostic method for voltage transformers based on the HarmonyOS system according to claim 1, characterized in that, The step of extracting a phase reference signal from the secondary voltage sampling sequence of the standard voltage transformer according to the adaptive sliding window, and dynamically estimating the actual operating frequency of the power grid to generate a variable-length sampling window synchronized with the actual operating frequency of the power grid, specifically includes: The standard voltage transformer secondary voltage sampling sequence is retrieved from the adaptive sliding window, and the standard voltage transformer secondary voltage sampling sequence is determined as the sole reference data source. The waveform cleanup process is performed on the secondary voltage sampling sequence of the standard voltage transformer to generate a reference sampling sequence that satisfies phase continuity and period stability. Based on the reference sampling sequence, a phase reference signal that is continuously updated over time is extracted, and continuous tracking and error iteration processing are performed on the phase reference signal to obtain a dynamic estimation result of the actual operating frequency of the power grid. The dynamic estimation result is subjected to convergence determination processing. When the preset stability condition is met, the dynamic estimation result is determined as the effective frequency estimation result. Based on the effective frequency estimation results, a synchronous reconstruction process is performed on the window length to generate a variable-length sampling window that forms an integer period matching relationship with the actual operating frequency of the power grid.

4. The real-time diagnostic method for voltage transformers based on the HarmonyOS system according to claim 1, characterized in that, The step involves performing windowed interpolation spectral analysis on the secondary voltage sampling sequences of the tested voltage transformer and the standard voltage transformer, respectively, based on the variable-length sampling window, to obtain the fundamental phasor estimation results of the tested voltage transformer and the standard voltage transformer. Specifically, this includes: Based on the variable-length sampling window, the secondary voltage sampling sequence of the tested voltage transformer and the secondary voltage sampling sequence of the standard voltage transformer are retrieved respectively, and a unified window loading process is performed to ensure that the two sampling sequences enter the spectrum analysis process under the same window boundary. The window validity verification process is performed on the secondary voltage sampling sequence of the tested voltage transformer and the secondary voltage sampling sequence of the standard voltage transformer to generate a valid sampling sequence that satisfies waveform continuity and channel synchronization. Windowing processing is performed on each of the effective sampled sequences to suppress spectral sidelobe spread caused by window boundaries; Spectral transformation is performed on the windowed sampling sequence, and the main response region is located in the neighborhood of the fundamental wave to obtain the main response region. Peak spectral line identification processing is performed on the main response region, and interpolation correction processing is performed based on the peak spectral line and its adjacent spectral lines to obtain the interpolation correction processing result; Based on the interpolation correction results, the fundamental phasor estimation results of the tested voltage transformer and the fundamental phasor estimation results of the standard voltage transformer are extracted respectively.

5. The real-time diagnostic method for voltage transformers based on the HarmonyOS system according to claim 1, characterized in that, The method involves calculating the ratio difference and angle difference based on the fundamental phasor estimation results of the tested voltage transformer and the standard voltage transformer, and then combining this with test condition information to form error characteristic data records, specifically including: Phasor decomposition processing is performed on the fundamental phasor estimation results of the tested voltage transformer and the fundamental phasor estimation results of the standard voltage transformer to obtain the amplitude expression sequence and the phase expression sequence, respectively. Amplitude comparison processing is performed based on the amplitude expression sequence to generate a ratio difference result characterizing the degree of amplitude response deviation, and phase comparison processing is performed based on the phase expression sequence to generate an angle difference result characterizing the degree of phase response shift. Perform error validity verification on the comparison results and the angle difference results to determine the valid error results under the current window; Synchronously acquire the test condition information corresponding to the effective error result, and perform condition correspondence binding processing on the test condition information, the ratio difference result, and the angle difference result; The comparison result, the angle difference result, and the test condition information are subjected to unified encapsulation processing to generate error feature data records containing window identifier, time identifier, and device identifier.

6. The real-time diagnostic method for voltage transformers based on the HarmonyOS system according to claim 1, characterized in that, The step involves mapping the error feature data records to an error feature knowledge graph, and using a graph neural network to reason about error type nodes, influencing factor nodes, and equipment status nodes to obtain error anomaly diagnosis results that include anomaly determination results, error source location results, and confidence assessment results. Specifically, this includes: Load the error feature knowledge graph and perform graph activation processing to make the error type node, influencing factor node and device status node in a mappable state; The error feature data records are subjected to field decomposition and semantic organization to generate error performance features, operating condition impact features, and context constraint features. Node mapping processing is performed based on the error performance characteristics, the operating condition impact characteristics, and the context constraint characteristics to form a corresponding local diagnostic subgraph; The local diagnostic subgraph is subjected to feature injection processing so that error performance features, operating condition influence features and context constraint features are written into the error type node, influencing factor node and equipment status node, respectively. Based on the local diagnostic subgraph, graph neural network inference is performed to perform anomaly judgment processing on error type nodes to generate anomaly judgment results, correlation analysis processing is performed on influencing factor nodes and equipment status nodes to generate error source location results, and confidence assessment processing is performed on the anomaly judgment results and the error source location results to generate confidence assessment results. The anomaly determination result, the error source location result, and the confidence assessment result are subjected to result fusion processing to generate the error anomaly diagnosis result.

7. The real-time diagnostic method for voltage transformers based on the HarmonyOS system according to claim 1, characterized in that, The differential processing strategy applied to the error anomaly diagnosis results and the original sampling sequence to complete the real-time diagnosis of voltage transformers based on the HarmonyOS system specifically includes: The error anomaly diagnosis result is correlated and confirmed with the original sampling sequence to establish a correspondence under the same window identifier, time identifier, and device identifier; Based on the error anomaly diagnosis results, perform diagnostic level discrimination processing to generate diagnostic level identifiers for driving differentiated processing strategies; Based on the diagnostic level identifier, data value assessment processing is performed on the original sampled sequence to determine the retention range and processing granularity of the original sampled sequence, and the data value assessment processing result is obtained. Hierarchical encapsulation processing is performed on the error anomaly diagnosis result and the original sampling sequence respectively to generate a diagnosis result encapsulation unit and a sampling data encapsulation unit; Resource awareness processing is performed on the collaborative domain where the edge computing nodes of the HarmonyOS system are located to obtain the network link status, local processing load status and collaborative node idle status, and obtain the resource awareness processing results. Based on the diagnostic level identifier, the data value assessment processing result, and the resource awareness processing result, a processing path selection process is performed to determine the local storage processing path, the summary upload processing path, or the enhanced upload processing path, and a processing path selection processing result is obtained. Based on the processing path selection processing result, the diagnostic result encapsulation unit and the sampling data encapsulation unit perform task allocation processing to perform data processing and task unloading between edge computing nodes or collaborative nodes of the HarmonyOS system, and generate task execution results. The task execution results are then processed by result synthesis and consistency verification to generate valid results. Based on the effective processing results, the final output processing is performed to complete the real-time diagnosis of the voltage transformer based on the HarmonyOS system.

8. A real-time diagnostic device for voltage transformers based on the HarmonyOS system, characterized in that, The device is used to execute the real-time diagnostic method for voltage transformers based on the HarmonyOS system as described in any one of claims 1 to 7. The device includes an acquisition module and a processing module, wherein... The acquisition module is used to acquire the original sampling sequence sent by the voltage transformer calibrator, which includes the secondary voltage sampling sequence of the voltage transformer under test and the secondary voltage sampling sequence of the standard voltage transformer, at the edge computing node of the HarmonyOS system via a distributed soft bus, and to construct an adaptive sliding window. The processing module is used to extract a phase reference signal from the secondary voltage sampling sequence of the standard voltage transformer according to the adaptive sliding window, and to dynamically estimate the actual operating frequency of the power grid to generate a variable-length sampling window that is synchronized with the actual operating frequency of the power grid. The processing module is further configured to perform windowed interpolation spectrum analysis processing on the secondary voltage sampling sequence of the tested voltage transformer and the secondary voltage sampling sequence of the standard voltage transformer according to the variable length sampling window, so as to obtain the fundamental phasor estimation results of the tested voltage transformer and the fundamental phasor estimation results of the standard voltage transformer. The processing module is also used to calculate the ratio difference and angle difference based on the fundamental phasor estimation results of the tested voltage transformer and the fundamental phasor estimation results of the standard voltage transformer, and to form error characteristic data records in combination with the test condition information. The processing module is also used to map the error feature data records to an error feature knowledge graph, and to perform reasoning on error type nodes, influencing factor nodes and equipment status nodes through a graph neural network to obtain an error anomaly diagnosis result that includes anomaly judgment result, error source location result and confidence assessment result; The processing module is also used to perform differentiated processing strategies on the error anomaly diagnosis results and the original sampling sequence to complete the real-time diagnosis of the voltage transformer based on the HarmonyOS system.

9. An electronic device, characterized in that, The electronic device includes a processor, a memory, a user interface, and a network interface. The memory is used to store instructions. The user interface and the network interface are both used to communicate with other devices. The processor is used to execute the instructions stored in the memory to cause the electronic device to perform the method as described in any one of claims 1 to 7.

10. A non-transitory computer-readable storage medium, characterized in that, The non-transitory computer-readable storage medium stores instructions that, when executed, perform the method as described in any one of claims 1 to 7.

Images