A generator power telemetry bias source end online calibration measurement system

Abstract

The application relates to the technical field of electric variable measurement, and particularly discloses a generator power remote measurement deviation source end online calibration measurement system. The system acquires a first reference voltage signal, a first measurement voltage signal and first acquisition node information, performs time mark and phase synchronization and time-frequency domain conversion, constructs a first amplitude-phase joint difference sequence of gain deviation and time delay deviation according to corresponding frequency points, updates a first dynamic amplitude threshold value according to normalized amplitude deviation probability density peak value migration in a same frequency band continuous operation window, shunts over-limit data according to continuity, blocks calibration quantity generation corresponding to transient disturbance, and performs hardware aging identification, remote measurement topology matching and historical state confirmation on persistent structural deviation, and locks the source end online calibration object. The application can reduce deviation misjudgment caused by time sequence mismatch, operation baseline migration and transient disturbance, and improve positioning accuracy of actual deviation sources in cascaded remote measurement equipment.

Description

Technical Field

[0001] This invention relates to the field of electrical variable measurement, and more specifically to an online calibration measurement system for generator power telemetry deviation source. Background Technology

[0002] During generator online operation, the power telemetry channel typically consists of a reference measurement chain, telemetry transmitters, and multiple measurement or signal conditioning nodes. It is used to acquire, transform, and remotely output electrical variables such as voltage related to generator operation. To determine whether the telemetry transmitters and their measurement chain maintain permissible measurement deviations, it is necessary to perform an online comparison between the reference measurement results and the telemetry output results.

[0003] Existing technologies typically acquire reference voltage and measured voltage signals within a preset sampling window, using fixed tolerances or preset thresholds to determine whether amplitude, phase, or time delay is abnormal, and generating calibration, alarm, or maintenance basis when deviations exceed the threshold. Some solutions also identify abnormal nodes in the telemetry channel through frequency domain analysis, anomaly classification, or equipment topology location.

[0004] However, the reference measurement chain and telemetry transmitters may have mismatched acquisition time scales and phases, making it easy for direct comparisons to include timing differences in equipment deviations. Changes in generator operating status can also cause the deviation statistical baseline to shift, making fixed thresholds difficult to apply continuously. At the same time, both single-window voltage transient disturbances and persistent electrical performance degradation can manifest as amplitude exceeding limits, and existing solutions struggle to distinguish between the two before entering the calibration link. For multi-node cascaded telemetry channels, end-point gain deviations and time delay deviations may also be formed by the superposition of different nodes, making it difficult to determine the actual deviation source.

[0005] The information disclosed in this background section is intended only to enhance the understanding of the overall background of the invention and should not be construed as an admission or in any way implying that the information constitutes prior art known to those skilled in the art. Summary of the Invention

[0006] This invention provides an online calibration measurement system for generator power telemetry deviation sources, which solves the problems in the prior art where the acquisition time scale and phase mismatch between the reference measurement chain and the telemetry transmitter easily leads to measurement pseudo-deviations, the fixed threshold is difficult to adapt to the migration of the deviation statistical baseline within different operating windows, the voltage transient disturbance and the continuous electrical performance degradation in a single window are easily confused, and the actual deviation source is difficult to determine in a multi-node cascaded power telemetry channel.

[0007] The technical solution of this invention is as follows: An online calibration and measurement system for generator power telemetry deviation source end includes: a synchronization spectrum construction module: used to acquire the first reference voltage signal output by the reference measurement chain in the generator power telemetry channel, the first measurement voltage signal output by the telemetry transmitter, and the corresponding first acquisition node information, perform time-scale and phase synchronization and time-frequency domain conversion to obtain a first reference spectrum sequence and a first measurement spectrum sequence; an amplitude-phase difference construction module: used to calculate the gain deviation and time delay deviation of the two spectrum sequences according to the corresponding frequency points, and form a first amplitude-phase joint difference sequence in frequency order; a threshold update and deviation splitting module: used to update the initial amplitude threshold according to the peak migration of the normalized amplitude deviation probability density within a continuous window of the same frequency band to obtain a first dynamic amplitude threshold; and to perform gain deviation... After the difference exceeds the first dynamic amplitude threshold, current is shunted: if the limit is not exceeded continuously, the second transient error feature vector is extracted, and the generation of calibration quantity is blocked after confirming the voltage transient disturbance; if the limit is exceeded continuously, the first structural error feature vector is extracted; Distortion identification and topology matching module: used to classify and map the first structural error feature vector to determine the first signal distortion type; when the first signal distortion type is hardware aging type, the first frequency fluctuation feature of the first measured voltage signal after synchronization is extracted, and combined with the first acquisition node information, the first candidate deviation device list is obtained by power telemetry topology matching; Source calibration object determination module: used to compare the candidate devices with the historical operating status to determine the first power measurement deviation source, and lock the source online calibration object according to the device identifier.

[0008] Optionally, the synchronization spectrum construction module is further configured to: perform frequency domain conversion on the first reference voltage signal and the first measured voltage signal, extract the corresponding first reference phase and the first measured phase, and obtain an initial phase difference; when the initial phase difference is greater than a phase threshold, perform time-scale resampling and reference phase shift on the two voltage signals to determine the aligned time axis where the phases coincide; extract voltage signals at the same operating moment according to the aligned time axis, and perform time-frequency domain conversion to obtain the first reference spectrum sequence and the first measured spectrum sequence.

[0009] Optionally, the amplitude-phase difference construction module is further configured to: extract the frequency point set of the first reference spectrum sequence and the first measurement spectrum sequence, remove abnormal frequency points to form a second reference spectrum sequence and a second measurement spectrum sequence; calculate the gain deviation and delay deviation of the second reference spectrum sequence and the second measurement spectrum sequence at the corresponding frequency points, perform smoothing filtering on the gain deviation and the delay deviation, and arrange them in frequency order to form the first amplitude-phase joint difference sequence; after forming the first amplitude-phase joint difference sequence, compare the gain deviation and delay deviation of the corresponding frequency points with the gain threshold and the delay threshold respectively to determine the over-limit state of the corresponding frequency points.

[0010] Optionally, the threshold update and deviation splitting module is further configured to: acquire the electrical variable sequence corresponding to the first reference voltage signal and the first measured voltage signal within a preset time period, and divide the continuous operation window of the same frequency band according to the frequency component; calculate the normalized amplitude deviation between the reference voltage and the measured voltage within each continuous operation window of the same frequency band, use the normalized amplitude deviation as a gain deviation sample, construct the probability density function of the normalized amplitude deviation and extract the peak migration between adjacent windows; determine the threshold update amount when the peak migration amount is greater than a preset evolution threshold, and use the threshold update amount to compensate the initial amplitude threshold to obtain the first dynamic amplitude threshold used for gain deviation determination.

[0011] Optionally, the threshold update and deviation splitting module is further configured to: determine that there is a persistent amplitude deviation when the gain deviation is greater than the first dynamic amplitude threshold and continues to exceed the limit within the continuous operation window; perform mean removal processing on the corresponding first amplitude-phase joint difference sequence; construct a joint deviation matrix based on the gain deviation and time delay deviation; perform principal component dimensionality reduction on the joint deviation matrix to obtain the first structural error feature vector characterizing the long-term electrical performance degradation of the telemetry transmitter and used for deviation source determination.

[0012] Optionally, the threshold update and deviation shunting module is further configured to: determine the existence of a single-window transient deviation when the gain deviation is greater than the first dynamic amplitude threshold and does not continuously exceed the limit within the continuous operation window; extract high-frequency residuals from the corresponding first amplitude-phase joint difference sequence, determine the transient error term based on the high-frequency residuals, and form the second transient error feature vector; and block the generation of calibration values ​​for the first amplitude-phase joint difference sequence when a voltage transient disturbance is confirmed based on the second transient error feature vector.

[0013] Optionally, the distortion identification and topology matching module is further configured to: combine the first structural error feature vector with the corresponding gain deviation and time delay deviation into a joint feature vector; classify and map the joint feature vector and compare it with a preset distortion type label to determine the first signal distortion type; when the first signal distortion type belongs to the hardware aging type, extract the main frequency amplitude and harmonic components of the signal obtained by synchronous processing of the first measured voltage signal to form the first frequency fluctuation feature.

[0014] Optionally, the distortion identification and topology matching module is further configured to: traverse the topology of the generator power telemetry device step by step according to the first signal distortion type, the first acquisition node information, and the first frequency fluctuation characteristics to obtain a first topology node sequence; obtain the amplitude and phase deviation of each node in the first topology node sequence, identify nodes whose amplitude and phase deviation exceeds the deviation threshold as abnormal nodes, construct an amplitude and phase deviation propagation path according to the connection relationship between abnormal nodes; perform association matching on the amplitude and phase deviation propagation path, and include abnormal nodes whose matching degree exceeds the matching threshold in the first candidate deviation device list.

[0015] Optionally, the source calibration object determination module is further configured to: extract the historical operating status of each candidate device in the first candidate deviation device list to form a first historical operating status set; combine the current gain deviation, current delay deviation and current frequency fluctuation corresponding to each candidate device to form a current deviation feature set; calculate the state matching degree between the current deviation feature set and the corresponding historical state in the first historical operating status set, and determine the candidate device with the state matching degree exceeding the state threshold as the first power measurement deviation source.

[0016] Optionally, the source-end calibration object determination module is further configured to: extract the device identifier corresponding to the first power measurement deviation source, lock the source-end online calibration object according to the device identifier; and output the source-end online calibration object only when the first signal distortion type corresponding to the first structural error feature vector belongs to the hardware aging type and the state matching degree exceeds the state threshold.

[0017] The beneficial effects of this invention are as follows: This invention synchronizes the first reference voltage signal and the first measured voltage signal in terms of time and phase, and constructs a first amplitude-phase joint difference sequence of gain deviation and time delay deviation according to the corresponding frequency points. This allows deviation judgment to be based on the same operating time and frequency reference, reducing the impact of pseudo-deviations caused by timing mismatch. The first dynamic amplitude threshold is updated by the migration of the peak value of the normalized amplitude deviation probability density within the continuous operating window of the same frequency band, so that the judgment benchmark changes with the deviation statistical baseline. By shunting only the data that has exceeded the limit according to whether it has exceeded the limit continuously, the generation of calibration quantity triggered by the voltage transient disturbance in the single window is blocked, and the continuous amplitude deviation is included in the structural error extraction and hardware aging identification. By combining the information of the first acquisition node, the first frequency fluctuation characteristics, the power telemetry topology, and the historical operating status, the first power measurement deviation source is determined and the online calibration object at the source end is locked, thereby improving the accuracy of the actual deviation source location and reducing the risk of erroneous calibration. Attached Figure Description

[0018] Figure 1This is a schematic diagram of the logical architecture of an online calibration and measurement system for generator power telemetry deviation source provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of the timing and phase synchronization and spectrum construction of dual-channel voltage signals provided in one embodiment of the present invention; Figure 3 This is a schematic diagram of the construction of the first phase joint differential sequence and the update of the first dynamic amplitude threshold provided in an embodiment of the present invention; Figure 4 This is a schematic diagram of the interruption of out-of-limit data continuity shunting and calibration quantity generation provided in one embodiment of the present invention; Figure 5 This is a schematic diagram of topology matching and amplitude-phase deviation propagation path of a power telemetry device provided in one embodiment of the present invention; Figure 6 This is a schematic diagram of historical operation status matching and source-end online calibration object locking provided in an embodiment of the present invention. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only for explaining the invention and are not intended to limit the invention; that is, the described embodiments are merely some embodiments of the invention, and not all embodiments. The components of the embodiments of the invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0020] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0021] It should be noted that relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0022] As mentioned earlier, when the generator power telemetry channel is running online, there may be acquisition time scale and phase mismatch between the reference measurement chain and the telemetry transmitter. The statistical baseline of amplitude deviation may shift within different operating windows. The transient voltage disturbance in a single window and the continuous electrical performance degradation of the telemetry transmitter may manifest as similar amplitude over-limits. For a power telemetry channel formed by cascading multiple measurement nodes and signal transmission nodes, the gain deviation and time delay deviation detected at the end may also be formed by the superposition of different nodes, making it difficult to directly determine the actual deviation source.

[0023] To address this, the present invention provides an online calibration measurement system for generator power telemetry deviation at the source end. The system uses a first reference voltage signal, a first measured voltage signal, and corresponding first acquisition node information from the same generator power telemetry channel as front-end inputs. It sequentially performs time-scale and phase synchronization, time-frequency domain conversion, frequency point amplitude-phase difference construction, threshold update driven by normalized amplitude deviation probability density peak migration, continuous shunting after exceeding limits, transient calibration quantity generation blocking, persistent structural error identification, power telemetry topology matching triggered by hardware aging, and historical operating status confirmation. Finally, it locks and outputs the online calibration object at the source end according to the device identifier.

[0024] The following is combined Figures 1 to 6 This invention is described in detail.

[0025] Example 1: like Figure 1 As shown, this embodiment of the invention provides an online calibration measurement system for generator power telemetry deviation at the source end. The system includes a synchronous spectrum construction module, an amplitude-phase difference construction module, a threshold update and deviation splitting module, a distortion identification and topology matching module, and a source-end calibration object determination module. These modules are connected according to the online measurement sequence of generator power telemetry deviation. The spectrum sequence, amplitude-phase joint difference sequence, branching results, or candidate deviation device list generated by the previous module serve as the processing basis for the next module.

[0026] The synchronization spectrum construction module is used to acquire the first reference voltage signal output by the reference measurement chain in the generator power telemetry channel, the first measurement voltage signal output by the telemetry transmitter, and the corresponding first acquisition node information, and to perform time-scale and phase synchronization and time-frequency domain conversion to obtain the first reference spectrum sequence and the first measurement spectrum sequence.

[0027] Specifically, both the first reference voltage signal and the first measured voltage signal are discrete voltage sequences arranged according to sampling time. Each sampling record includes a voltage value, a timestamp, and a validity flag. The first acquisition node information and the first measured voltage signal are bound using the same channel identifier and operating window number, and include a device identifier, a channel identifier, and an adjacency identifier for accessing the power telemetry device topology. The synchronization spectrum construction module verifies the timestamp monotonicity, common valid time period, and node binding status of the two voltage signals, obtains the initial phase difference between the two signals, and forms an aligned time axis with overlapping phases when the synchronization processing conditions are met. Valid sequences at the same operating time are extracted, and time-frequency domain conversion is performed using the same frequency resolution. The first reference spectrum sequence, the first measured spectrum sequence, and the first acquisition node information are jointly output to the amplitude-phase difference construction module.

[0028] The amplitude-phase difference construction module is used to calculate the gain deviation and time delay deviation of the two spectral sequences according to the corresponding frequency points, and form the first amplitude-phase joint difference sequence in frequency order.

[0029] Specifically, the amplitude-phase difference construction module extracts frequency points from the first reference spectrum sequence and the first measurement spectrum sequence that can correspond one-to-one according to frequency value or frequency point index. Abnormal frequency points with excessively low reference amplitude, missing any spectrum value, or isolated jumps between adjacent frequency points are removed. For the retained corresponding frequency points, gain deviation and time delay deviation are calculated respectively, and smoothing is performed on both deviation dimensions. The processed gain deviation and time delay deviation retain their respective dimensions and physical meanings, and form a first amplitude-phase joint difference sequence in ascending frequency order. Then, the first amplitude-phase joint difference sequence and the over-limit status of each frequency point are sent to the threshold update and deviation splitting module.

[0030] The first phase-coordinated difference sequence is indexed by the operating window number and frequency point identifier. Each gain deviation and delay deviation within the same operating window corresponds to a valid frequency point. During continuous limit exceedance determination, the system fixes the same frequency point within the same frequency band or a pre-selected main frequency point, and obtains the gain deviation and the first dynamic amplitude threshold of the frequency point in sequence according to the operating window number. The window gain deviation in formula (3) represents the gain deviation of the fixed frequency point within the corresponding operating window. The frequency point subscript is omitted to simplify the variable representation, and the gain deviations of different frequencies are not counted together. The limit exceedance state of the delay deviation is used as an auxiliary state field of the same frequency point to construct the first structural error feature vector and the subsequent amplitude-phase deviation propagation path, and does not trigger transient or persistent branches separately.

[0031] The threshold update and deviation shunting module is used to update the initial amplitude threshold based on the peak migration of the normalized amplitude deviation probability density within a continuous window in the same frequency band to obtain the first dynamic amplitude threshold; after the gain deviation exceeds the first dynamic amplitude threshold, shunting is performed: when there is no continuous exceedance, the second transient error feature vector is extracted, and the generation of calibration is blocked after confirming the voltage transient disturbance; when there is continuous exceedance, the first structural error feature vector is extracted.

[0032] The initial amplitude threshold is determined by the upper quantile of the initial valid samples in the same frequency band or the gain calibration tolerance of the telemetry transmitter (for example, the initial amplitude threshold is 0.020, dimensionless, derived from the upper limit of the initial valid samples and the 2% gain calibration tolerance); the first dynamic amplitude threshold is formed by updating the initial amplitude threshold according to the peak shift limit, and its allowable boundary is determined by the allowable deviation range of the device (for example, the lower limit is 0.010 and the upper limit is 0.050, both dimensionless, derived from the allowable deviation range of the device).

[0033] Specifically, the threshold update and deviation splitting module organizes sampling windows with the same frequency band identifier, sequential time, and valid verification into continuous operating windows within the same frequency band. It calculates the normalized amplitude deviation between the reference voltage and the measured voltage within each window and forms a probability density based on the normalized amplitude deviation samples. This module reads the peak position of the probability density of adjacent valid windows and performs limiting compensation on the initial amplitude threshold based on the peak shift, obtaining the first dynamic amplitude threshold. If the current gain deviation does not exceed the first dynamic amplitude threshold, the current window ends and returns to the next monitoring window; if the current gain deviation has exceeded the first dynamic amplitude threshold, mutual exclusion splitting is performed based on whether the limit is continuously exceeded within the continuous operating window. When there are no consecutive limit violations, the threshold update and deviation shunting module extracts high-frequency residuals from the first phase joint difference sequence to form a second transient error feature vector, and writes the calibration quantity to generate a blocking state when the voltage transient disturbance is confirmed; when there are consecutive limit violations, the joint deviation data composed of gain deviation and time delay deviation is subjected to mean removal and principal component dimensionality reduction to form a first structural error feature vector, and outputs it to the distortion identification and topology matching module.

[0034] The distortion identification and topology matching module is used to classify and map the first structural error feature vector to determine the first signal distortion type; when the first signal distortion type is hardware aging type, the first frequency fluctuation feature of the first measured voltage signal after synchronization is extracted, and the first acquisition node information is combined with power telemetry topology matching to obtain the first candidate deviation device list.

[0035] Specifically, the distortion identification and topology matching module combines the first structural error feature vector with the corresponding gain deviation and time delay deviation into a joint feature vector, and compares it with a preset distortion type label. Only when the first signal distortion type belongs to the hardware aging type, this module extracts the main frequency amplitude and harmonic components from the synchronized first measurement voltage signal to form the first frequency fluctuation feature; then, taking the node corresponding to the first acquisition node information as the topology starting point, it traverses step by step along the physical signal direction of the power telemetry device to determine the abnormal nodes and the amplitude and phase deviation propagation paths between abnormal nodes, and includes the abnormal nodes whose association matching results meet the conditions in the first candidate deviation device list.

[0036] The source-end calibration object determination module is used to compare candidate devices with historical operating states, determine the first power measurement deviation source, and lock the source-end online calibration object according to the device identifier.

[0037] Specifically, the source-end calibration object determination module extracts the historical gain deviation, historical delay deviation, historical frequency fluctuation, status label, recording time, and corresponding frequency band of each candidate device according to the device identifier in the first candidate deviation device list, forming a first historical operating status set. Simultaneously, it constructs a current deviation feature set by combining the current gain deviation, current delay deviation, and current frequency fluctuation of the candidate device. The source-end calibration object determination module calculates the state matching degree between the current deviation feature set and the corresponding historical operating status, identifies the only candidate device in the first candidate deviation device list that meets the state confirmation conditions as the first power measurement deviation source, and reads its device identifier to set the object lock status. The system outputs the source-end online calibration object only when the first structural error feature vector is valid, the first signal distortion type belongs to the hardware aging type, and the state matching degree meets the output conditions. The input validity, first dynamic amplitude threshold, branch result, candidate device list, state matching degree, and object identifier for this cycle are written into the historical operating record.

[0038] The calibration quantity generation blocking status indicates that subsequent independent calibration processes must not generate calibration quantities for the current first phase combined difference sequence; the source-end online calibration object identifies the target device that needs to be processed in subsequent independent calibration processes. This system does not generate compensation coefficients in the main processing chain, does not write calibration parameters to the telemetry transmitter register, and does not use shutdown or alarm as the output result of the source-end online calibration object.

[0039] Based on the above system, this invention synchronizes the time and phase of the first reference voltage signal and the first measured voltage signal before deviation judgment, so that the gain deviation and delay deviation of the corresponding frequency points are established on the same operating time and unified frequency reference; updates the first dynamic amplitude threshold by the migration of the normalized amplitude deviation probability density peak value, and distinguishes between single-window transient deviation and continuous amplitude deviation according to continuity after the gain deviation exceeds the limit, so that voltage transient disturbances are blocked before entering the deviation source location; and forms a first candidate deviation device list by combining the first frequency fluctuation characteristics, the first acquisition node information and the power telemetry topology only when the hardware aging type is established, and then determines and locks the first power measurement deviation source by combining the historical operating status, thereby forming a continuous processing closed loop from dual-channel signal synchronization, amplitude and phase difference construction, deviation splitting, topology matching to the output of the source-end online calibration object.

[0040] Example 2: To provide a clearer and more complete explanation of the above embodiments, the present invention also provides an embodiment two. For example... Figures 2 to 6 As shown, in this second embodiment, each module processes the data in the following order: dual-channel voltage signal synchronization, amplitude-phase difference construction, first dynamic amplitude threshold update, over-limit continuous current shunting, hardware aging type confirmation, power telemetry topology matching, and historical status confirmation.

[0041] like Figure 2 As shown, the synchronous spectrum construction module sequentially performs initial phase difference extraction, time scale resampling, reference phase shift, common effective time period interception, and time frequency domain conversion on the first reference voltage signal and the first measurement voltage signal to form the first reference spectrum sequence and the first measurement spectrum sequence.

[0042] Preferably, the synchronization spectrum construction module is further used for: The first reference voltage signal and the first measured voltage signal are frequency domain converted to extract the corresponding first reference phase and first measured phase, and the initial phase difference is obtained. When the initial phase difference is greater than the phase threshold, the two voltage signals are time-scaled resampled and the reference phase is shifted to determine the aligned time axis for phase overlap; the phase threshold is determined by the upper limit of the phase difference of multiple effective synchronization windows during initialization or the phase difference allowed by the telemetry transmitter (for example, the phase threshold is 2°, derived from the smaller value between the upper limit of the stable phase difference in the effective window of initialization and the phase difference allowed by the device calibration).

[0043] Voltage signals at the same operating moment are extracted according to the aligned time axis and converted in the time-frequency domain to obtain the first reference spectrum sequence and the first measurement spectrum sequence.

[0044] Specifically, when the synchronous spectrum construction module forms a new sampling window on the same generator power telemetry channel, it synchronously receives the sampled values, timestamps, and validity flags of two voltage signals, and binds the first acquisition node information to the first measured voltage signal according to the channel identifier and operating window number. The sampling window length is jointly determined by the target frequency resolution and the online processing response time limit (for example, the sampling window length is 10s, derived from the window configuration and online monitoring cycle configuration required to distinguish the 50Hz main frequency and its lower harmonics).

[0045] The phase threshold is determined by the upper bound of the phase difference of multiple effective synchronization windows during initialization or the phase difference allowed by the telemetry transmitter (for example, a phase threshold of 2°, derived from the smaller of the upper bound of the stable phase difference and the phase difference allowed by the device calibration within the effective initialization window). The synchronization spectrum construction module extracts the first reference phase and the first measurement phase at the same target frequency point within a common effective time period, and obtains the initial phase difference after periodic wrapping of the phase difference. When the initial phase difference is not greater than the phase threshold, the original timestamp is retained; when the initial phase difference is greater than the phase threshold, an equally spaced aligned time axis is established based on the timestamps. The two voltage signals are interpolated and resampled. The acquisition time base offset is determined based on the offset of the two timestamps relative to the aligned time axis, and the acquisition time base offset is converted into a reference phase shift. The reference phase of the first reference voltage signal is then shifted. The reference phase shift is only used to compensate for the time base offset of the two acquisition records and does not eliminate the actual transmission phase difference formed by the telemetry transmitter and its measurement chain. Phase coincidence refers to the sampling reference times of the two signals coinciding on the aligned time axis.

[0046] Specifically, after the reference phase is shifted, the remaining phase difference is recalculated. When the remaining phase difference enters the phase threshold, the same valid operating time interval of the two signals is extracted, and the same window function, window length, and frequency resolution are used for time-frequency domain conversion. The remaining phase difference retains the phase change formed by the telemetry transmitter and its measurement chain, and is used for subsequent time delay deviation calculation. If the common valid interval is less than one sampling window, the timestamp resampling crosses an invalid data interval, the acquisition time base offset cannot be determined based on the timestamp, or the first acquisition node information binding fails, the current window is marked as invalid, and the first reference spectrum sequence and the first measurement spectrum sequence for subsequent threshold updates are not formed, and the most recently valid first dynamic amplitude threshold is retained.

[0047] For example, at a target frequency of 50Hz, with the first reference phase at 10° and the first measurement phase at 13°, the initial phase difference is 3°, which is greater than the phase threshold of 2°. The synchronization spectrum construction module performs time-scale resampling and reference phase shifting. When the remaining phase difference after shifting is 1°, the common effective time period of the two signals is extracted to form two spectrum sequences. The 50Hz, 10°, 13°, and 1° are all example operating data used to illustrate the processing relationship.

[0048] like Figure 3 As shown, after removing abnormal frequency points, the amplitude-phase difference construction module calculates the gain deviation and time delay deviation according to the corresponding frequency points and forms the first amplitude-phase joint difference sequence; the threshold update and deviation splitting module further updates the first dynamic amplitude threshold according to the peak migration of the normalized amplitude deviation probability density within the continuous operation window of the same frequency band.

[0049] Preferably, the amplitude-phase difference construction module is further used for: Extract the frequency point set of the first reference spectrum sequence and the first measurement spectrum sequence, remove abnormal frequency points, and form the second reference spectrum sequence and the second measurement spectrum sequence; The gain deviation and time delay deviation between the second reference spectrum sequence and the second measured spectrum sequence are calculated at the corresponding frequency points. The gain deviation and the time delay deviation are smoothed and filtered, and arranged in frequency order to form the first amplitude-phase joint difference sequence. After forming the first phase joint difference sequence, the gain deviation and time delay deviation at the corresponding frequency point are compared with the gain threshold and time delay threshold, respectively, to determine the over-limit status of the corresponding frequency point. The gain threshold is determined by the telemetry transmitter gain calibration tolerance or the upper limit of the gain deviation of the initial valid sample (for example, the gain threshold is 0.020, dimensionless, i.e., 2%, derived from the device gain calibration tolerance and the upper limit of the initial stable window); the time delay threshold is determined by the telemetry transmitter's allowable time delay error or the upper limit of the time delay deviation of the initial valid sample (for example, the time delay threshold is 100μs, derived from the device's allowable time delay error and the upper limit of the initial stable window).

[0050] Abnormal frequency points include at least those whose first reference spectral amplitude is lower than the effective lower limit of the reference amplitude, those with any missing spectral value, and those forming isolated jumps relative to adjacent frequency points. The effective lower limit of the reference amplitude is determined by the voltage acquisition range, noise floor, and the minimum stable amplitude of the initialized valid sample (for example, the effective lower limit of the reference amplitude is 5V, derived from the safe amplitude boundary above the acquisition range and noise floor); frequency points below this lower limit are not included in the gain deviation calculation to avoid the denominator approaching zero.

[0051] Before calculating the amplitude and phase deviation, the amplitude and phase difference construction module, according to the channel ratio or engineering scale configuration of the reference measurement chain and the telemetry transmitter, converts the first reference voltage signal and the first measured voltage signal to the same voltage characterization scale, and uses the same window function, window length, frequency resolution, and spectral amplitude normalization method to form the second reference spectrum sequence and the second measured spectrum sequence. The time delay deviation is calculated only for frequency points greater than zero that have passed validity verification; DC frequency points are not included in the time delay deviation calculation.

[0052] Specifically, the gain deviation and delay deviation at the corresponding frequency point are calculated using the following formula: (1) in, and The second reference spectrum sequence and the second measurement spectrum sequence are respectively in the second reference spectrum sequence and the second measurement spectrum sequence in the second reference spectrum sequence. The complex spectral values ​​at each frequency point have an amplitude unit of V; and These represent the corresponding phases, in rad. The corresponding frequency point is in Hz; This is the effective lower limit of the reference amplitude, in V. The absolute value of the gain deviation is dimensionless. The positive or negative relationship between the measured spectrum amplitude and the reference spectrum amplitude is reserved separately as the gain deviation direction field. This refers to the time delay deviation, measured in seconds (s). This indicates that the phase difference has been converted to the principal value interval. The calculated... and As two independent fields with the same frequency in the first combined differential sequence.

[0053] The window length of the smoothing filter is determined by the frequency resolution and the number of effective frequency points (for example, the smoothing window length is 3 adjacent frequency points, derived from the number of effective frequency points in the current frequency band and the requirement for isolated fluctuation suppression). The amplitude-phase difference construction module smooths the gain deviation sequence and the time delay deviation sequence respectively, without directly adding two deviations of different dimensions; the smoothed deviation pairs are arranged in ascending order of frequency to form the first amplitude-phase joint difference sequence.

[0054] The gain threshold is determined by the telemetry transmitter gain calibration tolerance or the upper bound of the gain deviation of the initial valid sample (e.g., a gain threshold of 0.020, dimensionless, i.e., 2%, derived from the device calibration tolerance and the upper bound of the initial stability window); the delay threshold is determined by the telemetry transmitter's allowable delay error or the upper bound of the delay deviation of the initial valid sample (e.g., a delay threshold of 100 μs, derived from the device's allowable delay error and the upper bound of the initial stability window). The gain threshold and delay threshold are only used to mark the out-of-limit state of the corresponding frequency point after the formation of the first phase joint difference sequence, and are not prerequisites for forming the first phase joint difference sequence.

[0055] For example, at a frequency of 50Hz, when the amplitude of the second reference spectrum is 100V and the amplitude of the second measurement spectrum is 102.5V, and the phase difference between the two is 0.0314rad, according to equation (1), the gain deviation is 0.025 and the time delay deviation is about 100μs. When the gain deviation exceeds the gain threshold of 0.020 and the time delay deviation reaches the example time delay threshold of 100μs, the system records the corresponding over-limit state and retains the two deviation fields of this frequency point in the first phase joint difference sequence.

[0056] Preferably, the threshold update and deviation splitting module is further used for: Obtain the electrical variable sequence corresponding to the first reference voltage signal and the first measured voltage signal within a preset time period, and divide the continuous operation window of the same frequency band according to the frequency component; Calculate the normalized amplitude deviation between the reference voltage and the measured voltage within each continuous operating window of the same frequency band, use the normalized amplitude deviation as a gain deviation sample, construct the probability density function of the normalized amplitude deviation, and extract the peak shift between adjacent windows. When the peak migration amount exceeds a preset evolution threshold, a threshold update amount is determined. This threshold update amount is then used to compensate for the initial amplitude threshold, resulting in the first dynamic amplitude threshold used for gain deviation determination. The preset evolution threshold is determined by the upper bound of the probability density peak fluctuation of adjacent stable windows during the initialization phase (e.g., the preset evolution threshold is 0.002, dimensionless). The threshold update amount is determined using a restricted proportional update (e.g., the threshold update coefficient is 0.50, dimensionless, derived from offline initialization verification).

[0057] The sampling window is the basic sampling window for forming a first reference spectrum sequence and a first measurement spectrum sequence; the threshold statistics window consists of multiple basic sampling windows with the same frequency band identifier, continuous time sequence, and valid verification within a preset time period; the continuous over-limit window is an adjacent basic sampling window used for continuous counting in formula (3). The preset time period is determined by the number of effective samples required to form a stable probability density and the system online update cycle (for example, the preset time period is 30 minutes, derived from the number of continuous effective samples under a 10s sampling window and the threshold update cycle configuration). Within the same threshold statistics window, the normalized amplitude deviation at the same frequency point in each basic sampling window forms a gain deviation sample, and normalized amplitude deviation samples from different frequency bands or different frequency points are not mixed.

[0058] The probability density function is formed using normalized frequencies with fixed bins. The number of bins is determined by the initial sample size and the effective range of normalized amplitude deviation (for example, 50 bins, derived from the effective sample size and peak resolution requirements within a preset 30-minute time period). Adjacent threshold statistical windows use the same frequency band identifier and the same bin boundaries to extract the probability density peak positions, so that the peak migration can be directly compared. The basic sampling window only provides probability density samples and consecutive limit-crossing counts, and does not separately form the probability density peak for formula (2).

[0059] The initial amplitude threshold is determined by the upper quantile of the valid samples initialized in the same frequency band or the gain calibration tolerance of the telemetry transmitter (e.g., the initial amplitude threshold is 0.020, dimensionless, derived from the upper bound of the valid samples initialized and the 2% gain calibration tolerance); the preset evolution threshold is determined by the upper bound of the probability density peak fluctuation of adjacent stable windows during the initialization phase (e.g., the preset evolution threshold is 0.002, dimensionless, derived from the statistical upper bound of the peak migration of the stable window); the threshold update coefficient is determined by the proportion of peak migration transferred to the judgment threshold during initialization verification (e.g., the threshold update coefficient is 0.50, dimensionless, derived from offline initialization verification); the lower and upper limits of the first dynamic amplitude threshold are determined by the allowable deviation range of the device and the effective historical boundary (e.g., the lower limit is 0.010 and the upper limit is 0.050, both dimensionless, derived from the allowable deviation range of the device).

[0060] Specifically, the peak migration and the first dynamic amplitude threshold are determined by the following formula: (2) in, and These are the peak positions of the probability density of the current threshold statistical window and the previous effective threshold statistical window of the same frequency band, respectively, both of which are dimensionless; The peak migration is dimensionless. The preset evolution threshold is dimensionless. The threshold update coefficient is dimensionless. The first effective dynamic amplitude threshold is the previous effective threshold, and the first effective threshold statistics window is substituted with the initial amplitude threshold. and These are the lower and upper limits of the first dynamic amplitude threshold, respectively, both of which are dimensionless; The current first dynamic amplitude threshold is dimensionless; This indicates that the update result will be limited to the upper and lower limits. The calculated... Starting from the next basic sampling window after the current threshold statistics window ends, it is used to determine the gain deviation of the corresponding frequency band; the basic sampling window within the current threshold statistics window still uses the previous valid first dynamic amplitude threshold before the update.

[0061] When multiple peak values ​​with the same maximum value exist in the probability density, the peak value closest to the peak value of the previous effective window in the same frequency band is selected as the current peak value; if it still cannot be uniquely determined, the first dynamic amplitude threshold is not updated for the current window. When there are insufficient effective samples, the peak migration amount does not exceed the preset evolution threshold, or the update result reaches the upper or lower limit, the system either uses the previous effective threshold or restricts the update result to the corresponding boundary.

[0062] For example, the peak position of the probability density of the previous effective window of the same frequency band is 0.012, the peak position of the probability density of the current window is 0.015, the peak shift is 0.003, which is greater than the preset evolution threshold of 0.002; when the previous first dynamic amplitude threshold is 0.020 and the threshold update coefficient is 0.50, the current first dynamic amplitude threshold is 0.0215 according to equation (2), which is between the upper and lower limits of the example of 0.010 to 0.050.

[0063] like Figure 4 As shown, the threshold update and deviation shunting module performs shunting based on whether the gain deviation exceeds the first dynamic amplitude threshold within the continuous operation window; when the limit is continuously exceeded, a first structural error feature vector is formed; when the limit is not continuously exceeded, a second transient error feature vector is formed, and the generation of calibration quantity is blocked after confirming the voltage transient disturbance.

[0064] More preferably, the threshold update and deviation splitting module is further used for: When the gain deviation is greater than the first dynamic amplitude threshold and continues to exceed the limit within the continuous operation window, it is determined that there is a persistent amplitude deviation. The first phase joint difference sequence is subjected to mean removal processing. A joint deviation matrix is ​​constructed based on gain deviation and time delay deviation. Principal component dimensionality reduction is performed on the joint deviation matrix to obtain the first structural error feature vector that characterizes the long-term electrical performance degradation of the telemetry transmitter and is used for deviation source determination.

[0065] The continuous window threshold is determined by the length of a single sampling window and the shortest verifiable performance degradation duration of the telemetry transmitter (for example, the continuous window threshold is 3 valid windows, corresponding to 30 seconds, derived from the configuration of a 10-second sampling window and a 30-second shortest acknowledgment duration). The threshold update and deviation shunting module only checks continuity when the current gain deviation is greater than the current first dynamic amplitude threshold; windows that have not exceeded the limit do not enter transient or structural branches.

[0066] Specifically, the continuous over-limit state is determined by the following formula: (3) in, This is the threshold for consecutive windows, in units of [number]. This represents the gain deviation for the corresponding window, and is dimensionless. This is the first dynamic amplitude threshold for the corresponding window, and is dimensionless. For indicator functions; For the most recent The number of out-of-limit windows in a window, expressed in units of [number]. When The system determines if the current window has exceeded the limit. The system determines whether there are consecutive limit violations. If any intermediate window is marked as invalid or a valid limit-free state occurs, the consecutive count restarts.

[0067] In the continuous out-of-limit branch, the threshold update and deviation splitting module assembles a joint deviation matrix from the gain deviation and time delay deviation in the most recent continuous out-of-limit window according to a fixed field order. It then performs mean removal and scaling on the gain deviation column and time delay deviation column respectively, and calculates the covariance matrix and eigenvector. Principal components are selected according to the eigenvalues ​​from largest to smallest. The cumulative variance contribution threshold is determined by the requirement of retaining the main continuous changes and suppressing minor random fluctuations (e.g., the cumulative variance contribution threshold is 85%, derived from the principal component contribution distribution of the initial valid samples). The principal component projection results required to reach this threshold form the first structural error feature vector. This first structural error feature vector is used to characterize the main continuous change direction of the gain deviation and time delay deviation within the continuous out-of-limit window, and serves as input for subsequent first signal distortion type identification and deviation source determination. It is not used alone as a confirmation result that the telemetry transmitter has experienced long-term electrical performance degradation.

[0068] When the joint bias matrix has insufficient rank, any field scale is zero, or the cumulative variance contribution cannot reach the 85% threshold, the first structural error feature vector is marked as low confidence and subsequent distortion identification is stopped to prevent insufficient samples from forming incorrect hardware aging types.

[0069] More preferably, the threshold update and deviation splitting module is further used for: When the gain deviation is greater than the first dynamic amplitude threshold and does not continuously exceed the limit within the continuous operation window, it is determined that there is a single-window transient deviation. High-frequency residuals are extracted from the corresponding first phase joint difference sequence, and transient error terms are determined based on the high-frequency residuals to form the second transient error feature vector; When confirming voltage transient disturbances based on the second transient error eigenvector, the generation of calibration values ​​for the first phase-to-phase joint difference sequence is blocked.

[0070] The high-frequency residual is formed by subtracting the low-frequency trend obtained by short-window smoothing from the first amplitude-phase joint difference sequence. The short-window smoothing length follows the example configuration of three adjacent frequency points in the amplitude-phase difference construction stage. The transient error term includes the high-frequency residual peak value, the high-frequency residual root mean square, the high-frequency residual energy ratio, and the residual duration. The high-frequency residual energy ratio threshold is determined by the residual energy distribution of the confirmed voltage transient disturbance window (e.g., the high-frequency residual energy ratio threshold is 0.60, dimensionless, derived from the lower bound of the energy ratio of the confirmed transient window in the initialization stage). The short duration condition is determined by the duration range of the single-window transient disturbance (e.g., the duration does not exceed one 10s sampling window, derived from the single-window transient branch definition).

[0071] When the current basic sampling window has exceeded the limit and the number of consecutive limit exceedances in formula (3) is less than the continuous window threshold, the system records the current state as the single-window transient deviation pending confirmation state, retains the consecutive limit exceedance count, and extracts the second transient error feature vector. Here, the single-window transient deviation pending confirmation state is a status identifier for the processing branch where the current basic sampling window is located. It indicates that the current basic sampling window has not yet met the continuous limit exceedance condition, does not indicate that only one basic sampling window in the most recent continuous window has exceeded the limit, and does not directly indicate that the voltage transient disturbance has been confirmed.

[0072] When the proportion of high-frequency residual energy in the second transient error feature vector reaches the example threshold of 0.60 and the residual duration does not exceed one sampling window, the voltage transient disturbance is confirmed and a calibration quantity generation blocking state is set. This blocking state prevents the current first phase joint difference sequence from entering hardware aging type identification, power telemetry topology matching, and source-end online calibration object determination; subsequent new effective windows are processed again from the synchronization spectrum construction module. If the subsequent effective basic sampling window continues to exceed the limit, the single-window transient deviation pending confirmation state of the previous basic sampling window is canceled, and continuous limit exceeding counting continues; after reaching the continuous window threshold, it enters the continuous limit exceeding branch. If the second transient error feature vector does not meet the voltage transient disturbance confirmation condition, a calibration quantity generation blocking state is not set.

[0073] For example, if the current gain deviation is 0.030 and the current first dynamic amplitude threshold is 0.0215, but only the current window exceeds the limit in the last 3 valid windows, the result of equation (3) is less than 3, and the system enters the single-window transient deviation confirmation branch; if the high-frequency residual energy ratio is 0.68 and the duration is 1 sampling window, then the calibration quantity generation blocking state is written, and the first candidate deviation device list is not formed.

[0074] Preferably, the distortion identification and topology matching module is further used for: The first structural error feature vector is combined with the corresponding gain deviation and time delay deviation to form a joint feature vector. The joint feature vector is then classified and mapped and compared with a preset distortion type label to determine the first signal distortion type. When the first signal distortion type belongs to the hardware aging type, the main frequency amplitude and harmonic components of the signal obtained by synchronous processing of the first measured voltage signal are extracted to form the first frequency fluctuation feature.

[0075] The preset distortion type label includes a distortion type identifier, the order of joint feature vector fields, the reference center or allowed interval of each field, and the label matching condition. The preset distortion type label is formed from historical operation records whose status type has been validated and confirmed (for example, the label matching threshold for hardware aging type is 0.80, dimensionless, derived from the matching degree distribution boundary between confirmed hardware aging records and non-aging records).

[0076] Specifically, the distortion identification and topology matching module compares the first structural error feature vector, gain bias, and time delay bias using the same field order and scale as the preset distortion type labels. For numerical fields with a reference center and an allowable range, the field matching value is determined by the degree of deviation of the field from the reference center. If the field value is within the allowable range, the corresponding matching value is retained; if it exceeds the allowable range, the field is recorded as a mismatch. For state fields, a match is recorded when the current state is consistent with the label state, and a mismatch is recorded when they are inconsistent. The matching values ​​of each valid field are combined to form the label matching degree according to the non-negative weights of the corresponding fields in the preset distortion type labels. Missing fields are not included in the synthesis, and the weights of the remaining valid fields are renormalized. If the number of valid fields is insufficient to meet the label matching conditions, the first signal distortion type is recorded as undetermined. The classification mapping is a rule comparison based on the preset distortion type label fields and matching conditions.

[0077] If the tag matching degree does not reach the example threshold of 0.80, the first signal distortion type is recorded as undetermined and the current processing cycle ends. If the tag matching degree reaches the threshold, the first signal distortion type is determined based on the tag with the highest matching degree and a unique correspondence. If two or more distortion type tags simultaneously reach the tag matching threshold but cannot form a unique highest matching result, the first signal distortion type is recorded as undetermined and does not enter the power telemetry topology matching process.

[0078] When the first signal distortion type belongs to the hardware aging type, the distortion identification and topology matching module reads the dominant frequency amplitude and selected harmonic components from the spectrum of the synchronized first measured voltage signal. The selected harmonic range is determined based on the effective frequency response range of the power telemetry channel and the validity of the initialized spectrum (for example, selecting the 2nd to 5th harmonics, which are derived from the effective frequency response range of the current telemetry channel); the first frequency fluctuation characteristics include the inter-window variation of the dominant frequency amplitude, the amplitude ratio of each selected harmonic relative to the dominant frequency, and the inter-window variation of their respective harmonics.

[0079] like Figure 5 As shown, the distortion identification and topology matching module takes the node corresponding to the first acquisition node information as the topology starting point, traverses the power telemetry device topology level by level along the physical signal direction, determines abnormal nodes based on the amplitude and phase deviation of each node and constructs the amplitude and phase deviation propagation path, and forms the first candidate deviation device list through association matching.

[0080] More preferably, the distortion identification and topology matching module is further used for: Based on the first signal distortion type, the first acquisition node information, and the first frequency fluctuation characteristics, the first topology node sequence is obtained by traversing the topology of the generator power telemetry equipment level by level. The amplitude and phase deviation of each node in the first topological node sequence is obtained. Nodes whose amplitude and phase deviation exceeds the deviation threshold are identified as abnormal nodes. The amplitude and phase deviation propagation path is constructed based on the connection relationship between abnormal nodes. The deviation threshold is determined by the normalized result of the node gain deviation and node delay deviation relative to their respective allowable boundaries (for example, the normalized deviation threshold is 1.0, dimensionless, derived from the normalized boundary of the gain threshold of 0.020 and the delay threshold of 100μs).

[0081] The amplitude-phase deviation propagation paths are correlated and matched, and abnormal nodes with a matching degree exceeding a matching threshold are included in the first candidate deviation device list. The matching threshold is determined by the distribution boundary of the correlation matching degree between confirmed propagation paths and non-propagation paths (for example, the matching threshold is 0.75, dimensionless, derived from the lower bound of the matching degree of historical confirmed paths).

[0082] The generator power telemetry equipment topology records the device identifier, channel identifier, adjacent node identifier, and physical signal direction of each node. The distortion identification and topology matching module takes the node corresponding to the first acquisition node information as the starting point, visits adjacent nodes step by step along the physical signal direction, and uses the set of visited nodes to truncate the topology loop, forming the first topology node sequence.

[0083] The node gain deviation and node delay deviation of each node in the first topology node sequence are obtained from the node current deviation record bound to the corresponding device identifier, channel identifier, operating window number, and frequency identifier. The node current deviation record is formed by the input-side voltage record and output-side voltage record of the corresponding node in the same operating window according to the same frequency matching and amplitude-phase deviation calculation method as formula (1), and the record validity status is retained. When the node current deviation record is inconsistent with the current operating window, the record field is missing, or the record validity status does not meet the requirements, the corresponding node is marked as node deviation record unavailable, and the amplitude-phase deviation propagation path is not constructed based on the node.

[0084] The amplitude and phase deviation of each node is jointly determined by the ratio of the node gain deviation to the gain threshold and the ratio of the node delay deviation to the delay threshold. The normalized deviation threshold is determined when the two ratios reach their respective allowable boundaries (e.g., the normalized deviation threshold is 1.0, dimensionless, derived from the normalization boundary of the gain threshold of 0.020 and the delay threshold of 100 μs). The influence weights of path continuity, deviation direction consistency, similarity of first frequency fluctuation characteristics, and node adjacency are determined by the reliability of each factor in the confirmed deviation propagation record (e.g., the four weights are 0.30, 0.25, 0.30, and 0.15 respectively, all dimensionless, derived from the field reliability configuration of historical confirmed paths); the matching threshold is determined by the distribution boundary of the association matching degree between confirmed propagation paths and non-propagation paths (e.g., the matching threshold is 0.75, dimensionless, derived from the lower bound of the matching degree of historical confirmed paths).

[0085] While calculating the absolute value of the gain deviation, the amplitude-phase difference construction module retains the positive and negative relationship between the measured spectrum amplitude and the reference spectrum amplitude, forming a gain deviation direction field. This gain deviation direction field does not participate in the gain deviation exceeding limit judgment; it is only used for node deviation direction consistency scoring. The positive or negative sign of the time delay deviation serves as the time delay deviation direction field.

[0086] Specifically, the correlation degree between the node amplitude and phase deviation and the amplitude and phase deviation propagation path is determined by the following formula: (4) in, and The first The gain bias and delay bias of each topology node are dimensionless and s, respectively; and These are the gain threshold and the delay threshold, respectively, and are dimensionless and s, respectively; This is the nodal amplitude and phase deviation, dimensionless. The corresponding node will be identified as an abnormal node. , , and The scores are normalized scores for propagation path continuity, node deviation direction consistency, first frequency fluctuation feature similarity, and node adjacency relationship, all of which are dimensionless. to The corresponding influence weights are all dimensionless; The correlation matching degree is dimensionless. The calculated... Compared with the matching threshold, abnormal nodes that exceed the matching threshold are included in the first candidate deviation device list.

[0087] The propagation path continuity score is based on the path from the topological starting point to the... In a path with a number of nodes, the ratio of the number of adjacent relationships between actually continuously connected abnormal nodes to the total number of adjacent relationships that the path should have is determined; the node deviation direction consistency score is determined based on the ratio of the number of comparable fields with consistent directions in the gain deviation direction field and delay deviation direction field of adjacent abnormal nodes within the path to the total number of valid comparable fields; the first frequency fluctuation feature similarity score is based on the... The normalized difference between the main frequency amplitude change and the selected harmonic amplitude ratio of each node and the corresponding field of the first frequency fluctuation characteristic is determined. The smaller the field difference, the higher the similarity score; the node adjacency relationship score is determined in the first... A node is assigned a value of 1 if it has a direct connection to the previous anomalous node as shown in the topology record and conforms to the physical signal direction; otherwise, it is assigned a value of 0. If any score lacks necessary input, it is not replaced by a preset constant; instead, the corresponding path is marked as having insufficient score fields and is not included in the first candidate deviation device list.

[0088] The amplitude and phase deviation propagation path is constructed according to the actual connection relationship between abnormal nodes. Continuously connected abnormal nodes with the same direction of amplitude and phase deviation change form the same path; isolated abnormal nodes can be retained, but their path continuity score is reduced. When there is a topology break, missing node deviation records, or the correlation matching degree of multiple paths does not exceed the example matching threshold of 0.75, a new source-end online calibration object is not formed, but a low-confidence or pending verification record is retained.

[0089] For example, when the gain deviation of a certain topology node is 0.030 and the delay deviation is 80μs, substituting the gain threshold of 0.020 and the delay threshold of 100μs into equation (4), the node amplitude and phase deviation is 1.5, which is greater than the normalized deviation threshold of 1.0, and the node is identified as an abnormal node; when the four normalized scores of the corresponding propagation path are 0.90, 0.80, 0.85 and 1.00 respectively, the association matching degree is 0.875 with example weights of 0.30, 0.25, 0.30 and 0.15, which exceeds the matching threshold of 0.75, and the abnormal node enters the first candidate deviation device list.

[0090] like Figure 6 As shown, the source calibration object determination module matches the current deviation characteristics of each candidate device in the first candidate deviation device list with the corresponding historical operating status. When the state matching degree of the unique candidate device exceeds the state threshold, the first power measurement deviation source is determined, and the source online calibration object is locked according to its device identifier.

[0091] Preferably, the source-end calibration object determination module is further configured to: Extract the historical operating status of each candidate device from the first candidate deviation device list to form a first historical operating status set; Combine the current gain deviation, current delay deviation, and current frequency fluctuation of each candidate device to form the current deviation feature set; The state matching degree between the current deviation feature set and the corresponding historical state in the first historical operating state set is calculated respectively. Candidate devices whose state matching degree exceeds the state threshold are identified as the first power measurement deviation source. The state threshold is determined by the boundary of the matching degree distribution between the confirmed degraded state and the non-degraded state (for example, the state threshold is 0.80, dimensionless, derived from the lower bound of the matching degree of the confirmed historical state).

[0092] The first historical operating status set contains only historical status records that correspond one-to-one with the candidate device identifier, have the same frequency band, or are comparable and have passed validity verification. Each historical status record includes historical gain deviation, historical delay deviation, historical frequency fluctuation, status label, recording time, and corresponding frequency band; candidate devices lacking confirmed degradation status records are marked as having insufficient historical samples and are not replaced by the normal historical average value as the degradation status template.

[0093] For the same candidate device and the same frequency band, the effective lower and upper boundaries of each field—gain deviation, delay deviation, and frequency fluctuation—are determined based on the historical state records that have been validated. The same set of effective boundaries is then used to perform dimensionless normalization on both the current and historical fields. If the effective upper and lower boundaries of a field are the same, that field is marked as non-normalizable and does not participate in state matching. The historical degradation state template is composed of the median values ​​of each normalized field in the confirmed degradation state records of the candidate device, to reduce the impact of a single abnormal historical record on the template. If the number of confirmed degradation state records is insufficient to form a valid template, the corresponding candidate device is marked as having insufficient historical samples.

[0094] The weights of the gain deviation, delay deviation, and frequency fluctuation in the state matching degree are determined by the stability and validity of the corresponding fields in confirmed historical states (e.g., the weights of the three fields are 0.40, 0.30, and 0.30 respectively, all dimensionless, derived from the reliability configuration of historical state fields); the state threshold is determined by the boundary of the matching degree distribution between confirmed degraded and non-degraded states (e.g., the state threshold is 0.80, dimensionless, derived from the lower bound of the matching degree of confirmed historical states). When some fields are marked as non-normalizable or missing, only the remaining valid fields are retained, and their field weights are renormalized; if the number of valid fields is insufficient, the state matching degree is not calculated.

[0095] Specifically, the state matching degree of the candidate device is calculated using the following formula: (5) in, For the first The first candidate device in the current deviation feature set A normalized field value, dimensionless; The first of the historical degradation state templates corresponding to this candidate device A normalized field value, dimensionless; For the first The non-negative weights of each field are dimensionless. The number of fields involved in the matching, expressed in units of individual fields. The state matching degree is dimensionless. The calculated value is... Compared with the state threshold, if and only if a candidate device When the state threshold is exceeded, the candidate device is identified as the first source of power measurement deviation; if there are two or more candidate devices... At the same time, when the state threshold is exceeded, the candidate device and its corresponding state matching degree are retained and the state to be reviewed is output, without locking the source online calibration object.

[0096] For example, the current gain deviation, current delay deviation, and current frequency fluctuation of a candidate device are 0.80, 0.70, and 0.75 after being normalized by historical scale, respectively. The corresponding historical degradation state templates are 0.75, 0.65, and 0.70, respectively. When using example field weights of 0.40, 0.30, and 0.30, the state matching degree obtained according to Equation (5) is approximately 0.951, which exceeds the state threshold of 0.80. Therefore, the candidate device meets the state confirmation condition of the first power measurement deviation source.

[0097] More preferably, the source-end calibration object determination module is further used for: Extract the device identifier corresponding to the first power measurement deviation source, and locate the online calibration object at the source end based on the device identifier; The source-end online calibration object is output only when the first signal distortion type corresponding to the first structural error feature vector belongs to the hardware aging type and the state matching degree exceeds the state threshold.

[0098] Specifically, when only one candidate device in the first candidate deviation device list has a state matching degree exceeding the state threshold, the source calibration object determination module identifies that candidate device as the first power measurement deviation source, reads its unique device identifier, and sets the object lock state to prevent the output object within the same processing cycle from being overwritten by other nodes. When multiple candidate devices exceed the state threshold simultaneously, the system retains the candidate ranking and corresponding state matching degree and outputs a status pending review, without forcibly locking the source online calibration object.

[0099] Before outputting the source-end online calibration object, the source-end calibration object determination module re-verifies the effective state of the first structural error feature vector, the first signal distortion type, and the state matching degree; if any condition is not met, no new source-end online calibration object is output. When the output interface is temporarily unavailable, the system records the device identifier, object lock status, and current cycle results as data to be output, and continues outputting after the interface is restored, without generating or issuing actual calibration quantities.

[0100] When the first reference voltage signal and the first measured voltage signal do not share a common valid time period, the probability density samples are insufficient, the first structural error feature vector has low confidence, the first signal distortion type does not belong to the hardware aging type, the power telemetry equipment topology is broken, or the historical operating status of the candidate equipment is insufficient, the system stops the corresponding subsequent processing and retains the invalid, insufficient sample, low confidence, undetermined, topology abnormal, or insufficient historical sample states, respectively. The next valid sampling window starts again from the synchronization spectrum construction module, without directly inheriting the transient or structural classification of the previous window, and only calls the most recently valid first dynamic amplitude threshold and the historical operating status that has been validated.

[0101] Accordingly, this invention establishes a correspondence between the first reference voltage signal, the first measured voltage signal, and the first acquisition node information at the front end through a synchronous spectrum construction module, and performs time-scale resampling and reference phase shift when the phase difference exceeds the limit, so that subsequent gain deviation and delay deviation are formed based on the corresponding frequency points of the same operating time; after removing abnormal frequency points, the amplitude-phase difference construction module forms a first amplitude-phase joint difference sequence, and the threshold update and deviation splitting module updates the first dynamic amplitude threshold using the probability density peak migration within the continuous operating window of the same frequency band, and only splits the data that has exceeded the limit according to continuity, so that... A single-window voltage transient disturbance triggers the generation and blocking of calibration quantities, causing continuously exceeding-limit data to form the first structural error feature vector. Based on this, the distortion identification and topology matching module extracts the first frequency fluctuation feature, constructs the amplitude and phase deviation propagation path, and forms the first candidate deviation device list only when the hardware aging type is established. Then, the source-end calibration object determination module confirms the first power measurement deviation source by combining historical operating status and locks the source-end online calibration object according to the device identifier. This makes the dual-channel signal synchronization, dynamic judgment, anti-misdirection, topology positioning, and historical confirmation form a source-end online calibration measurement chain that is connected from front to back.

[0102] The above description is merely a preferred embodiment of the technical solution of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A generator power telemetry deviation source-end online calibration and measurement system, characterized in that, include: Synchronous spectrum construction module: used to acquire the first reference voltage signal output by the reference measurement chain in the generator power telemetry channel, the first measurement voltage signal output by the telemetry transmitter and the corresponding first acquisition node information, perform time-scale and phase synchronization and time-frequency domain conversion to obtain the first reference spectrum sequence and the first measurement spectrum sequence; Amplitude-phase difference construction module: used to calculate the gain deviation and time delay deviation of two spectral sequences according to the corresponding frequency points, and form the first amplitude-phase joint difference sequence in frequency order; Threshold update and deviation splitting module: used to update the initial amplitude threshold based on the peak migration of the normalized amplitude deviation probability density within a continuous window in the same frequency band, to obtain the first dynamic amplitude threshold; Shunting is performed after the gain deviation exceeds the first dynamic amplitude threshold: when there is no continuous over-limit, the second transient error feature vector is extracted, and the generation of calibration quantity is blocked after confirming the voltage transient disturbance; when there is a continuous over-limit, the first structural error feature vector is extracted. Distortion identification and topology matching module: used to classify and map the first structural error feature vector to determine the first signal distortion type; when the first signal distortion type is hardware aging type, extract the first frequency fluctuation feature of the first measured voltage signal after synchronization, and combine it with the first acquisition node information along the power telemetry topology matching to obtain the first candidate deviation device list; Source calibration target determination module: used to compare candidate devices with historical operating status, determine the first power measurement deviation source, and lock the source online calibration target according to the device identification.

2. The system according to claim 1, characterized in that, The synchronization spectrum construction module is further used for: The first reference voltage signal and the first measured voltage signal are frequency domain converted to extract the corresponding first reference phase and first measured phase, and the initial phase difference is obtained. When the initial phase difference is greater than the phase threshold, the two voltage signals are time-scaled resampled and reference phase shifted to determine the aligned time axis where the phases coincide. Voltage signals at the same operating moment are extracted according to the aligned time axis and converted in the time-frequency domain to obtain the first reference spectrum sequence and the first measurement spectrum sequence.

3. The system according to claim 1, characterized in that, The amplitude-phase difference construction module is further used for: Extract the frequency point set of the first reference spectrum sequence and the first measurement spectrum sequence, remove abnormal frequency points, and form the second reference spectrum sequence and the second measurement spectrum sequence; The gain deviation and time delay deviation between the second reference spectrum sequence and the second measured spectrum sequence are calculated at the corresponding frequency points. The gain deviation and the time delay deviation are smoothed and filtered, and arranged in frequency order to form the first amplitude-phase joint difference sequence. After forming the first phase-joint difference sequence, the gain deviation and delay deviation of the corresponding frequency point are compared with the gain threshold and delay threshold, respectively, to determine the over-limit status of the corresponding frequency point.

4. The system according to claim 1, characterized in that, The threshold update and deviation splitting module is further used for: Obtain the electrical variable sequence corresponding to the first reference voltage signal and the first measured voltage signal within a preset time period, and divide the continuous operation window of the same frequency band according to the frequency component; Calculate the normalized amplitude deviation between the reference voltage and the measured voltage within each continuous operating window of the same frequency band, use the normalized amplitude deviation as a gain deviation sample, construct the probability density function of the normalized amplitude deviation, and extract the peak shift between adjacent windows. When the peak migration amount is greater than a preset evolution threshold, a threshold update amount is determined, and the threshold update amount is used to compensate the initial amplitude threshold to obtain the first dynamic amplitude threshold for gain deviation determination.

5. The system according to claim 4, characterized in that, The threshold update and deviation splitting module is further used for: When the gain deviation is greater than the first dynamic amplitude threshold and continues to exceed the limit within the continuous operation window, it is determined that there is a persistent amplitude deviation. The first phase joint difference sequence is subjected to mean removal processing. A joint deviation matrix is ​​constructed based on gain deviation and time delay deviation. Principal component dimensionality reduction is performed on the joint deviation matrix to obtain the first structural error feature vector that characterizes the long-term electrical performance degradation of the telemetry transmitter and is used for deviation source determination.

6. The system according to claim 4, characterized in that, The threshold update and deviation splitting module is further used for: When the gain deviation is greater than the first dynamic amplitude threshold and does not continuously exceed the limit within the continuous operation window, it is determined that there is a single-window transient deviation. High-frequency residuals are extracted from the corresponding first phase joint difference sequence, and transient error terms are determined based on the high-frequency residuals to form the second transient error feature vector; When confirming voltage transient disturbances based on the second transient error eigenvector, the generation of calibration values ​​for the first phase-to-phase joint difference sequence is blocked.

7. The system according to claim 1, characterized in that, The distortion identification and topology matching module is further used for: The first structural error feature vector is combined with the corresponding gain deviation and time delay deviation to form a joint feature vector. The joint feature vector is then classified and mapped and compared with a preset distortion type label to determine the first signal distortion type. When the first signal distortion type belongs to the hardware aging type, the main frequency amplitude and harmonic components of the signal obtained by synchronous processing of the first measured voltage signal are extracted to form the first frequency fluctuation feature.

8. The system according to claim 7, characterized in that, The distortion identification and topology matching module is further used for: Based on the first signal distortion type, the first acquisition node information, and the first frequency fluctuation characteristics, the first topology node sequence is obtained by traversing the topology of the generator power telemetry equipment level by level. Obtain the amplitude and phase deviation of each node in the first topological node sequence, identify nodes whose amplitude and phase deviation exceeds the deviation threshold as abnormal nodes, and construct the amplitude and phase deviation propagation path based on the connection relationship between abnormal nodes. The amplitude and phase deviation propagation paths are correlated and matched, and abnormal nodes with a matching degree exceeding the matching threshold are included in the first candidate deviation device list.

9. The system according to claim 8, characterized in that, The source-end calibration object determination module is further used for: Extract the historical operating status of each candidate device from the first candidate deviation device list to form a first historical operating status set; Combine the current gain deviation, current delay deviation, and current frequency fluctuation of each candidate device to form the current deviation feature set; Calculate the state matching degree between the current deviation feature set and the corresponding historical state in the first historical operating state set, and determine the candidate device whose state matching degree exceeds the state threshold as the first power measurement deviation source.

10. The system according to claim 9, characterized in that, The source-end calibration object determination module is further used for: Extract the device identifier corresponding to the first power measurement deviation source, and locate the online calibration object at the source end based on the device identifier; The source-end online calibration object is output only when the first signal distortion type corresponding to the first structural error feature vector belongs to the hardware aging type and the state matching degree exceeds the state threshold.

Images