A decentralized DTU station terminal system based on linear interpolation method

Abstract

The application discloses a kind of distributed DTU station based on linear interpolation method terminal system, including multiple interval units and a public unit;Interval unit gathers and sends to public unit with original sampling time label carrying bus voltage sampling sequence and feeder current sampling sequence;Public unit includes event determination module, time difference calculation module and data synchronization reconstruction module;Event determination module identifies the local switching event time of capacitor switching in each interval unit;Time difference calculation module calculates the time alignment offset of each interval unit according to local switching event time;Data synchronization reconstruction module uses time alignment offset to shift the original sampling time label of each interval unit, and executes linear interpolation when any interval unit is missing data at a certain time position of sampling point total set.

Description

Technical Field

[0001] The present invention relates to the field of power systems and power distribution, and more specifically, to a decentralized DTU (Distribution Terminal Unit) station terminal system based on the linear interpolation method. Background Art

[0002] In a distribution automation system, the DTU station terminal is an important node connecting primary distribution equipment and the dispatching / master station. It is commonly used to sample, upload, and record events of electrical quantities such as the bus voltage and the current of each feedback line in the station, and cooperate to complete functions such as alarm, waveform recording, fault location, and linkage control. With the increase in the scale of the station and the number of access circuits, a decentralized architecture gradually appears in engineering, that is, each interval configures an independent sampling and processing unit, and at the same time, a common unit aggregates the data of multiple intervals and communicates externally. In this type of architecture, whether the cross-interval data can be aligned on the same time axis directly affects the reliability of event review, fault waveform splicing, and the joint judgment of multi-source data by protection and automation algorithms.

[0003] In the prior art, regarding the decentralized station terminal and the distribution automation system, many works have proposed improvement schemes from the perspectives of system reliable operation and communication coordination. For example, the patent document "A Parameter Push Method for a Distributed Intelligent Distribution Terminal System" (Publication No. CN110336380A) proposes a distributed terminal parameter distribution, initialization synchronization, and rollback mechanism composed of a management unit and a front-end unit to solve the problem of difficult parameter synchronization when the online status of the front-end unit is unknown. Another patent document "A Self-Healing Method for Adaptive Differential Protection of a Distribution Network Based on Wireless Communication" (Publication No. CN113452000A) proposes to synchronize multi-terminal terminal data by combining global time synchronization and interpolation synchronization in a wireless communication scenario for ensuring data consistency in differential protection and self-healing control.

[0004] However, there is still a more subtle and difficult pain point in the engineering implementation of the station terminal: the decentralized interval units often sample independently and carry local time tags. Even if the sampling periods are the same, time axis misalignment will still occur due to local clock drift, startup phase difference, communication buffering, and message queuing, resulting in a millisecond-level or even larger misalignment in the records of the same physical event in different intervals. This misalignment is not noticeable within a single interval, but once it is necessary to perform correlation analysis on the bus voltage and the currents of multiple feedback lines under the same time reference, problems such as waveform peak misalignment, reversal of transient sequence, and false phase difference generated after splicing will occur, thereby affecting the credibility of work such as switching event, disturbance source location, and abnormal waveform review. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to provide a decentralized DTU station terminal system based on the linear interpolation method to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: A distributed DTU station terminal system based on linear interpolation includes multiple bay units and a common unit; The interval unit is configured to acquire bus voltage sampling sequences and feeder current sampling sequences carrying original sampling time tags and send them to the common unit; The common unit includes an event determination module, a time difference calculation module, and a data synchronization and reconstruction module; The event determination module is configured to identify the local switching event time of each interval unit based on the capacitor switching characteristics in the bus voltage sampling sequence. The time difference calculation module is configured to calculate the time alignment offset of each interval unit relative to the global reference time based on the local switching event time. The data synchronization and reconstruction module is configured to use the time alignment offset to shift the original sampling time label of each interval unit to obtain the aligned sampling point position of each interval unit, and merge the aligned sampling point positions of all interval units to construct a total set of sampling points. The data synchronization and reconstruction module is also configured to perform linear interpolation to fill in missing data at a certain time position of any interval unit if the missing data is at a certain time position of the total set of sampling points.

[0007] Preferably, the common unit is pre-configured with a switching characteristic frequency band library, which stores target frequency bands determined based on historical capacitor switching waveforms; the event determination module is configured to perform filtering processing on the bus voltage sampling sequence for the target frequency band to extract the energy envelope sequence, and mark the time corresponding to the sampling point with the largest value in the energy envelope sequence as the coarse event time.

[0008] Preferably, the event determination module is further configured to extract the energy envelope sequence near the coarse event time as a feature fingerprint fragment and generate a set of fine-tuning candidate offsets.

[0009] Preferably, the time difference calculation module is configured to designate one of the interval units as the reference interval unit, set its feature fingerprint segment as the reference waveform, and set the feature fingerprint segments of the remaining interval units as the waveforms to be tested; the time difference calculation module determines the fine alignment deviation of each interval unit by translating and matching the waveform to be tested relative to the reference waveform.

[0010] Preferably, the translation matching method is as follows: after applying each of the fine-tuning candidate offsets to the waveform to be tested, the waveform overlap between the translated waveform to be tested and the reference waveform is calculated, and the fine-tuning candidate offset corresponding to the optimal waveform overlap is determined as the fine alignment deviation.

[0011] Preferably, the waveform overlap is obtained by performing curve fitting on the reference waveform and the waveform to be tested respectively to construct a continuous reference feature curve and a waveform to be tested feature curve; calculating the definite integral of the absolute value of the difference between the reference feature curve and the translated waveform to be tested in the overlapping time region, wherein the optimal waveform overlap means that the calculation result of the definite integral is minimized.

[0012] Preferably, the local switching event time is obtained by adding the coarse event time and the fine alignment deviation; the time difference calculation module calculates the time alignment offset by calculating the difference between the local switching event time of the unit to be synchronized and the global reference time.

[0013] Preferably, the time difference calculation module determines the station reference time by directly setting the local switching event time of the reference interval unit as the station reference time.

[0014] Preferably, the method for determining the reference interval unit is as follows: calculate the pairwise optimal waveform overlap between the feature fingerprint segments of each interval unit in the entire station, and obtain the sum of the optimal waveform overlap of each interval unit with all other interval units; determine the interval unit with the largest sum of the optimal waveform overlap as the reference interval unit.

[0015] Preferably, the linear interpolation completion includes: finding the two nearest original sampling points before and after the missing data time as interpolation reference points; calculating the time distance between the missing data time and the previous interpolation reference point, and determining the ratio of the time distance to a fixed sampling period as a weighting coefficient; calculating the amplitude difference between the two interpolation reference points, multiplying the amplitude difference by the weighting coefficient to obtain an amplitude adjustment, and adding the amplitude adjustment to the amplitude of the previous interpolation reference point to obtain the sampled value at the missing data time. The fixed sampling period is the time difference between the two nearest original sampling points before and after the missing data time.

[0016] The advantage of this invention over existing technologies lies in its full utilization of the strong synchronicity and distinctive characteristics of capacitor switching events on the station bus side. Capacitor switching typically occurs at or near the bus, causing transient voltage fluctuations that propagate rapidly along the station's electrical connections. This results in multiple bays observing the same type of disturbance almost simultaneously in their respective bus voltage sampling sequences. These disturbances not only occur at concentrated times but also exhibit stable event fingerprint characteristics in terms of frequency distribution and energy envelope shape, showing significant differences from the gradual changes in daily load, power frequency background, and general noise. Therefore, they are more suitable as natural anchor points for station-wide alignment. Compared to methods relying on external time synchronization or manual experience for point selection, using the characteristics of switching fluctuations for event detection can obtain more stable event timing identification results under conditions of noise, amplitude differences, and measurement link inconsistencies, thus providing a reliable foundation for high-precision time alignment.

[0017] Based on this, the system identifies the local switching event time from the bus voltage sampling sequences transmitted from each bay through a common unit, and calculates the time alignment offset of each bay relative to the station's reference time. The original sampling time labels of each bay are then shifted, ensuring that the sampling point positions of multiple bays are uniformly projected onto the same time axis. Finally, the aligned sampling point positions are merged to construct a total set of sampling points. Since the alignment offset directly originates from the common observation results of the same physical event across multiple bays, the aligned bus voltage and feeder current sequences achieve higher consistency at key locations such as transient initiation, peak value, and attenuation. This significantly reduces problems such as false phase differences, misjudgment of sequence order, and waveform mismatches that occur after cross-bay splicing, improving the reliability of event replay, waveform splicing, and multi-source correlation analysis.

[0018] To balance engineering efficiency and alignment accuracy, the system employs a two-tiered strategy of "coarse-tuning positioning + fine-tuning micro-alignment." First, a pre-defined switching characteristic frequency band is used to selectively filter the bus voltage sequence and extract its energy envelope. This quickly identifies the sampling point with the most significant energy across the entire sequence as the coarse event moment, avoiding the costly point-by-point fine-matching on long-series data, thus ensuring processing efficiency and real-time performance. Then, a short window of characteristic fingerprint segments is extracted only near the coarse event moment, generating a set of fine-tuning candidate offsets. Within a small range, the waveform under test is translated relative to the reference waveform, and the fine alignment deviation is determined by selecting the best match based on waveform overlap. This approach concentrates computational resources on the locations where events are most likely to occur and the local segments requiring the most fine alignment, achieving high-precision alignment under controllable computational load. This allows the system to quickly complete full-station alignment while improving the overlap of multi-interval waveforms at the millisecond level or even finer granularity.

[0019] In actual substation operation, capacitor switching is a routine operational event in voltage and reactive power management. It can be triggered by automatic voltage control strategies or executed by maintenance personnel during voltage regulation, load fluctuations, or power factor adjustments. Therefore, it is not a low-probability fault but a natural event that occurs periodically during normal grid operation. Since this event typically occurs near the busbar and is visible to multiple bays, its transient fluctuations are highly synchronized in time and exhibit stable characteristics in form, making it naturally suitable as an anchor point for substation time alignment. Utilizing this natural event for alignment allows for the automatic extraction of alignment criteria directly from existing operational data without introducing additional manual time synchronization operations or relying on additional on-site time synchronization equipment modifications. This enables lower-cost, easier-to-deploy, high-precision synchronous alignment across multiple bays.

[0020] Furthermore, this scheme enhances usability in data reconstruction after alignment. By constructing a total set of sampling points and identifying missing data at certain time points in each interval, and then using linear interpolation to fill in the missing points, it avoids breakpoints in the time axis after merging, which could lead to subsequent algorithms being unable to calculate or making misjudgments. Simultaneously, the interpolation process, based on adjacent sampling points and a fixed sampling period, forms interpretable and easily reproducible repair results, facilitating engineering testing, traceability, and consistency verification. The determination of the reference interval unit incorporates a selection method based on the summarization of the pairwise optimal waveform overlap across the entire station. This automatically selects a more representative reference source when the signal-to-noise ratio and link conditions of multiple intervals are inconsistent, reducing instability caused by manual specification and further improving the accuracy and consistency of the alignment results across the entire station. Attached Figure Description

[0021] Figure 1 This is a schematic diagram of the overall physical architecture of the distributed DTU station terminal system provided in the embodiment of the present invention, showing the topology of multiple bay units connected to a common unit through a communication network; Figure 2 This is an overall logical flowchart of the system's data processing in this embodiment of the invention, covering the entire process from data acquisition and event recognition to data interpolation and completion; Figure 3 This is a schematic diagram illustrating the principle of extracting coarse event moments based on capacitor switching features in an embodiment of the present invention, showing the process from filtering the original waveform to extracting the energy envelope; Figure 4 This is a schematic diagram of the feature fingerprint fragment extraction method in an embodiment of the present invention, illustrating the process of determining a window on the energy envelope sequence and extracting feature data; Figure 5 This is a schematic diagram illustrating the principle of determining fine alignment deviation based on waveform translation matching in an embodiment of the present invention, showing the process of translating and comparing the waveform under test with the reference waveform; Figure 6This is a schematic diagram illustrating the formation process of the total set of sampling points in this invention; Figure 7 This is a schematic diagram of the geometric principle of using linear interpolation to fill in missing data in an embodiment of the present invention, which shows the mathematical logic of calculating the sampled value at the target time using the reference points before and after and their weights. Detailed Implementation

[0022] The specific embodiments of the present invention will now be described with reference to the accompanying drawings.

[0023] like Figure 1 As shown, the distributed DTU station terminal system provided by this invention includes multiple bay units and one common unit. Each bay unit is deployed at its corresponding feeder bay or measurement and control bay, used to collect electrical quantities locally and transmit them to the common unit via a communication network. The purpose of this distributed configuration is to complete local sampling close to the primary equipment, reducing interference from long-distance analog lead wires, while the common unit centrally performs station-wide data alignment, reconstruction, and external services, thus balancing on-site deployment convenience with station-wide consistent processing capabilities.

[0024] In one embodiment, the interval unit has at least a voltage sampling channel and a current sampling channel, and its sampling objects include a bus voltage sampling sequence and a feeder current sampling sequence carrying original sampling time tags. The original sampling time tag is used to describe the sampling time of each sampling point under the local clock of the interval unit. This tag can be generated by the internal timer of the interval unit, or it can be calculated by combining the sampling interruption count and the fixed sampling period. Since the local clock of each interval unit exists independently, even if the sampling period is consistent, the time axis between different intervals may be inconsistent due to start-up phase difference, timing drift, buffer queuing and communication jitter. Therefore, it is necessary to send the original sampling time tag along with the sampling data so that it can be uniformly corrected on the common unit side later.

[0025] like Figure 2 As shown, the common unit includes an event determination module, a time difference calculation module, and a data synchronization reconstruction module. The event determination module is used to identify capacitor switching events from the bus voltage sampling sequences sent by each bay unit and obtain the local switching event time. The reason for selecting capacitor switching as the alignment anchor point is that capacitor switching is a normal operating event in the station operation, often used for reactive power compensation and voltage support. Its occurrence location is usually close to the bus and can present a highly synchronized transient disturbance in the bus voltage sampling sequences of multiple bays. Therefore, multi-bay alignment can be completed directly by utilizing events that naturally occur during operation without adding manual time synchronization operations or relying on additional time synchronization hardware modifications.

[0026] like Figure 3As shown, in a further embodiment, the common unit has a pre-set switching characteristic frequency band library, which stores target frequency bands determined based on historical capacitor switching waveforms. The target frequency band is set to highlight the difference between switching disturbances and the power frequency background and random noise, making event identification more stable. The target frequency band can be extracted as follows: during the closing or opening operation of the station's switching device, transient segments of the bus voltage are extracted from at least 2 to 20 historical waveform recordings. Spectral analysis is performed on each segment, and the frequency band where the energy peak is located is statistically analyzed. Then, the intersection of multiple statistical results or the frequency band with an occurrence frequency higher than a threshold is taken as the target frequency band. The occurrence frequency threshold can be a ratio between 0.6 and 0.9 to ensure that the target frequency band includes the common components of most switching events without excessively introducing frequency bands unrelated to switching. If there are many types of capacitor banks in the station, sub-frequency band entries can also be established for different groups, and the corresponding entry can be selected according to the current switching object during operation. To accommodate different sampling rates, the target frequency band in the frequency band library can be represented by normalized frequency or by Hz and converted to discrete filter parameters according to the sampling rate.

[0027] In the implementation of the event determination module, the bus voltage sampling sequence can be filtered for the target frequency band to extract the energy envelope sequence, and the time corresponding to the sampling point with the largest value in the energy envelope sequence is marked as the coarse event time. The filter can be implemented using a finite impulse response filter or an infinite impulse response filter, and the filter order can be selected based on real-time requirements and filter sharpness, with a commonly used range of 16 to 256. The energy envelope sequence can be obtained by taking the absolute value of the filtered sequence and performing a moving mean square or moving integral. The sliding window length can be taken as 1 to 20 sampling periods to avoid misjudgment caused by occasional spikes, while maintaining sensitivity to switching transients. The coarse event time, as the result of the first step of localization, reduces the search problem across the entire sequence to a local fine alignment problem, thereby ensuring overall processing efficiency.

[0028] like Figure 4As shown, in a further embodiment, the event determination module also extracts the energy envelope sequence near the coarse event time as a feature fingerprint segment and generates a set of fine-tuning candidate offsets. The extraction window of the feature fingerprint segment can be set as a symmetrical window or an asymmetrical window around the coarse event time. A symmetrical window is suitable for scenarios where the morphology before and after the switching transient is relatively balanced, while an asymmetrical window is suitable for scenarios where the rising edge of the transient is more critical. The window length can be from 0.5 cycles to the number of sampling points corresponding to 5 cycles to ensure that the main energy change process of the switching disturbance is included, while avoiding the introduction of other operational disturbances by an excessively long window. The setting of the fine-tuning candidate offset is used to achieve a finite search for fine alignment. Its value range can cover the expected time offset between intervals. For example, when the sampling period is 1ms, the fine-tuning candidate offset can be set to a set of integer milliseconds from -20ms to +20ms, or a set of integer sampling points from -10 sampling points to +10 sampling points. If the on-site communication jitter is greater, it can be extended to -100ms to +100ms, but increasing the candidate set will increase the matching calculation workload. Therefore, it can be selected in combination with the scale of the station and the real-time requirements.

[0029] like Figure 5 As shown, in a further embodiment, the time difference calculation module designates one interval unit as the reference interval unit, sets its feature fingerprint segment as the reference waveform, and sets the feature fingerprint segments of the remaining interval units as the waveforms to be tested. The fine alignment deviation of each interval unit is determined by translating and matching the waveform to be tested relative to the reference waveform. The purpose of this design is to utilize the synchronicity of the switching disturbance in multiple intervals to transform the originally incomparable local time tag alignment problem into a waveform alignment problem, and to deduce the time deviation from the alignment result.

[0030] During translation matching, after applying each fine-tuning candidate offset to the waveform under test, the waveform overlap between the translated waveform under test and the reference waveform is calculated. The fine-tuning candidate offset corresponding to the optimal waveform overlap is determined as the fine alignment deviation. This finite candidate set search method avoids the instability caused by continuous domain optimization and also facilitates deterministic computation and time cost estimation on embedded common units.

[0031] In one embodiment, the waveform overlap is obtained by performing curve fitting on both the reference waveform and the waveform to be measured to construct continuous reference and measured feature curves. Then, the definite integral of the absolute value of the difference between the reference feature curve and the translated measured feature curve within the overlapping time region is calculated. In some embodiments, the waveform overlap can be represented by the negative value of the definite integral; therefore, the minimum definite integral corresponds to the optimal waveform overlap. As a convenient limiting case, if two waveforms completely overlap, then the difference between them is 0, and the corresponding definite integral is 0, thus indicating a high waveform overlap.

[0032] The purpose of curve fitting is to mitigate the jagged effect caused by the discreteness of sampling points, making the overlap evaluation focus more on the overall consistency of the shape rather than single-point errors. Curve fitting can be achieved using polynomial fitting, spline fitting, or piecewise linear fitting. To ensure real-time performance and numerical stability, the polynomial order can be 2 to 5, spline fitting can use quadratic or cubic splines, and piecewise linear fitting can directly connect sampling points to form a continuous curve. Definite integrals can be achieved through numerical integration, for example, by using the trapezoidal integral method to accumulate the absolute value of the difference point by point within the overlap time region, multiplying it by the time step. The overlap time region is determined by the common time coverage interval of the reference waveform and the shifted waveform under test, avoiding evaluation bias caused by missing boundaries.

[0033] In a further embodiment, the local switching event time is obtained by adding the coarse event time to the fine alignment deviation. This is because the coarse event time provides a rapid estimate of the event location, while the fine alignment deviation makes minor corrections to the coarse positioning result, ensuring the final event time remains stable and has higher accuracy. The time difference calculation module calculates the time alignment offset by calculating the difference between the local switching event time of the unit to be synchronized and the global reference time. The global reference time can be determined by directly setting the local switching event time of the reference interval unit as the global reference time. This avoids introducing additional external time sources, lowers the deployment threshold, and ensures that global alignment starts from the observation time of the same physical event at a certain reference interval.

[0034] In another embodiment, the reference interval unit is not fixed but automatically determined by the optimal waveform overlap between pairwise characteristic fingerprint segments of all interval units across the entire station. Specifically, the aforementioned fine-tuning candidate offset search is performed on the characteristic fingerprint segments of any two intervals to obtain the optimal waveform overlap between them. Then, the optimal waveform overlap of the same interval with all other intervals is summed, and the interval unit with the largest summation is determined as the reference interval unit. The purpose of this design is to prioritize the interval with the strongest morphological consistency with other intervals as the reference source, thereby reducing the impact of high noise in individual intervals, measurement link anomalies, or transient response distortions, making the overall station alignment results more robust.

[0035] The data synchronization and reconstruction module uses time alignment offsets to shift the original sampling time labels of each interval unit, thereby obtaining the aligned sampling point positions of each interval unit. It then merges the aligned sampling point positions of multiple interval units to construct a total set of sampling points. During the shifting process, the corresponding time alignment offset can be directly added to or subtracted from the time label of each original sampling point, mapping that sampling point to a unified time axis across the entire station. When the original sampling time labels are expressed using sampling point indices, the time alignment offset can be converted into a sampling point offset before shifting the sampling point index. The total set of sampling points is formed by summing the sampling point positions of each interval unit on a unified time axis. It can be sorted by time from smallest to largest and deduplicated to form a consistent time position set across the entire station, used for subsequent splicing and comparison of multi-interval data under the same time coordinate.

[0036] like Figure 6 As shown, it should be noted that the total set of sampling points is the union of the sampling point positions of multiple interval units on a unified time axis. It is not required that each interval unit naturally has a sampling point at every moment position in this union. Due to discrete differences in the sampling start point, clock offset correction, and translational landing position among different interval units, it is common for one interval unit to have a sampling point at a certain moment position, while another interval unit does not have a corresponding sampling point at that moment position. To facilitate multi-interval data alignment and comparison on the same time coordinate, the data synchronization and reconstruction module, for any interval unit, when it lacks a corresponding sampling value at a certain moment position in the total set of sampling points, performs linear interpolation to fill in the missing value using the adjacent original sampling data of that interval unit, thereby generating a sampling value for comparison at that moment position.

[0037] like Figure 7As shown, the linear interpolation completion is implemented by finding the two closest original sampling points before and after the missing data moment in the original sampled data of the interval unit as interpolation reference points. The time distance between the missing data moment and the previous interpolation reference point is calculated, and the ratio of this time distance to a fixed sampling period is determined as a weighting coefficient. The amplitude difference between the two interpolation reference points is then calculated, multiplied by the weighting coefficient to obtain the amplitude adjustment amount, and added to the amplitude of the previous interpolation reference point to obtain the sampled value at the missing data moment. The fixed sampling period is given by the interval unit sampling configuration, and a common range is 0.25ms to 5ms. To avoid the interpolation result deviating from the true waveform when the span between adjacent sampling points is too large, in one embodiment, the maximum span of the interpolation can be limited. For example, if the time span between the previous and next interpolation reference points exceeds 5 sampling periods, no interpolation result is generated and it is marked as uncompleted. By using the above method, without changing the original sampling mechanism of each interval unit, the data from multiple intervals can have a consistent data form that can be directly compared at the same time position corresponding to the total set of sampling points.

[0038] Through the above implementation methods, this invention achieves high-precision alignment of multi-interval sampling data by utilizing the transient event of capacitor switching, which occurs naturally during normal operation and can be observed at multiple intervals, without relying on additional manual time synchronization. After alignment, the continuity and availability of all station data on a unified time axis are ensured by constructing a total set of sampling points and completing the data through linear interpolation. This provides a stable and reliable data foundation for station-level event review, waveform splicing, and multi-source joint analysis.

[0039] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A distributed DTU station terminal system based on linear interpolation, characterized in that, It includes multiple interval units and one common unit; The interval unit is configured to acquire bus voltage sampling sequences and feeder current sampling sequences carrying original sampling time tags and send them to the common unit; The common unit includes an event determination module, a time difference calculation module, and a data synchronization and reconstruction module; The event determination module is configured to identify the local switching event time of each interval unit based on the capacitor switching characteristics in the bus voltage sampling sequence. The time difference calculation module is configured to calculate the time alignment offset of each interval unit relative to the global reference time based on the local switching event time. The data synchronization and reconstruction module is configured to use the time alignment offset to shift the original sampling time label of each interval unit to obtain the aligned sampling point position of each interval unit, and merge the aligned sampling point positions of all interval units to construct a total set of sampling points. The data synchronization and reconstruction module is also configured to perform linear interpolation to fill in missing data at a certain time position of any interval unit if the missing data is at a certain time position of the total set of sampling points.

2. The system according to claim 1, characterized in that, The common unit is pre-set with a switching characteristic frequency band library, which stores target frequency bands determined based on historical capacitor switching waveforms; the event determination module is configured to perform filtering processing on the bus voltage sampling sequence for the target frequency band to extract the energy envelope sequence, and mark the time corresponding to the sampling point with the largest value in the energy envelope sequence as the coarse event time.

3. The system according to claim 2, characterized in that, The event determination module is also configured to extract the energy envelope sequence near the coarse event time as a feature fingerprint fragment and generate a set of fine-tuning candidate offsets.

4. The system according to claim 3, characterized in that, The time difference calculation module is configured to designate one of the interval units as the reference interval unit, set its feature fingerprint segment as the reference waveform, and set the feature fingerprint segments of the remaining interval units as the waveforms to be tested; the time difference calculation module determines the fine alignment deviation of each interval unit by translating and matching the waveform to be tested relative to the reference waveform.

5. The system according to claim 4, characterized in that, The specific method of translation matching is as follows: after applying each of the fine-tuning candidate offsets to the waveform to be tested, the waveform overlap between the translated waveform to be tested and the reference waveform is calculated, and the fine-tuning candidate offset corresponding to the optimal waveform overlap is determined as the fine alignment deviation.

6. The system according to claim 5, characterized in that, The waveform overlap is obtained by performing curve fitting on the reference waveform and the waveform to be tested respectively to construct a continuous reference feature curve and a waveform to be tested feature curve; calculating the definite integral of the absolute value of the difference between the reference feature curve and the translated waveform to be tested in the overlapping time region, wherein the optimal waveform overlap means that the calculation result of the definite integral is minimized.

7. The system according to claim 4, characterized in that, The local switching event time is obtained by adding the coarse event time and the fine alignment deviation; the time difference calculation module calculates the time alignment offset by calculating the difference between the local switching event time of the unit to be synchronized and the global reference time.

8. The system according to claim 4, characterized in that, The time difference calculation module determines the station reference time by directly setting the local switching event time of the reference interval unit as the station reference time.

9. The system according to claim 6, characterized in that, The method for determining the reference interval unit is as follows: calculate the pairwise optimal waveform overlap between the characteristic fingerprint segments of each interval unit in the entire station, and obtain the sum of the optimal waveform overlap of each interval unit with all other interval units; the interval unit with the largest sum of the optimal waveform overlap is determined as the reference interval unit.

10. The system according to claim 1, characterized in that, The linear interpolation completion includes: finding the two closest original sampling points before and after the missing data time as interpolation reference points; calculating the time distance between the missing data time and the previous interpolation reference point, and determining the ratio of the time distance to a fixed sampling period as a weighting coefficient; calculating the amplitude difference between the two interpolation reference points, multiplying the amplitude difference by the weighting coefficient to obtain the amplitude adjustment amount, and adding the amplitude adjustment amount to the amplitude of the previous interpolation reference point to obtain the sampling value at the missing data time.

Images