Collection terminal for communication quality monitoring

Abstract

The invention relates to the technical field of communication quality, and particularly discloses an acquisition terminal for communication quality monitoring, which comprises an electric energy meter interface module for reading an instrument value from an electric energy meter; the time synchronization module is used for aligning the terminal clock to the coordinated world time and generating a time frame with 1s as granularity; the communication quality monitoring module is used for measuring a communication quality index and calculating a value at the 95% position within a preset statistical time period as a statistical result preset statistical time period; the data slicing and counting module is used for forming 10T0 fragments according to 10 fundamental wave periods in each time frame; the caching and out-of-order correction module is used for performing grouping, out-of-order rearrangement and missing compensation on instrument values and communication quality indexes according to the same time frame before uploading and completing framing in a waiting window; and the alarm and uploading module is used for outputting a local alarm by taking the value of the 95% position as a statistical result. According to the invention, the overall end-to-end time delay of the acquisition terminal is reduced.

Description

Technical Field

[0001] This invention relates to the field of communication quality monitoring technology, and more specifically, to a data acquisition terminal for communication quality monitoring. Background Technology

[0002] Distribution networks and electricity consumption information collection systems extensively utilize smart meters and data acquisition terminals across wide areas, transmitting meter readings to the master station's MDMS / SCADA / DMS platforms via public networks (cellular / Internet / VPN). Compared to dedicated networks, public network links suffer from uncontrollable scheduling and congestion, frequently experiencing jitter, queuing, and retransmissions, leading to out-of-order arrivals and end-to-end latency fluctuations. In monitoring services driven by second-level time sequences (such as second-by-second readings, event identification, and short-window statistics), this network uncertainty directly disrupts data aggregation within the same time frame: if data from different sources (electricity measurements, communication measurements from the same terminal, or multiple fields from the same frame) are misplaced in the master station's database, it will cause statistical drift, anomaly detection delays, or false alarms, and in severe cases, render control and alarm decisions baseless. In practice, the causes of public network jitter include wireless air interface scheduling, cross-cell / base station handover, network address translation / firewall state reconstruction, TCP retransmission and congestion control state transition, etc., which cause the delay distribution to exhibit long-tail characteristics and time correlation. Simply relying on average values ​​or instantaneous measurements is difficult to reflect the true arrival pattern of "the vast majority of data".

[0003] Existing publicly available literature 1 (Main Indicators and Test Methods of Distribution Network Phasor Data Concentrator, 2019) first introduces the typical structure of distribution network WAMS, such as Figure 2As shown, a hierarchical distributed architecture of "master station—PDC—µMPMU" is adopted: the top layer is the WAMS master station, which runs advanced applications such as phasor wide-area measurement, state estimation, early warning, model parameter identification, and wide-area protection based on satellite clock; the middle layer is the phasor data concentrator (PDC) deployed at nodes such as substations / switching stations, which is responsible for aggregating, aligning, and uploading phasor data from each access device; the communication backbone between the master station and each PDC is a fiber optic ring network composed of fiber optic SDH / MSTP switches; the access layer is the miniature multi-functional synchronous phasor measurement device (µMPMU) distributed in 10kV distribution line ring network cabinets, switching stations, or pole-mounted switches, which, in addition to phasor measurement and event recording, also integrates the "three remote" functions of distribution automation, fault diagnosis, and protection control; then, a reusable testing method and platform for distribution network PDCs are proposed. On the hardware side, a hybrid communication test bench is built using real µMPMUs from NARI / XJ, GPS timing, relay protection testers, and power line carrier (PLC) modules; on the software side, a self-developed PMU is proposed. An emulator (VC++, conforming to GB / T26865.2—2011, supporting up to ≤30 concurrent units, 1–200 frames / s, configurable phasor / analog / switching signals, and allowing manual injection of 10–500 ms cross-channel fixed delay) and a simulated master station are used. An automatic packet capture program based on libpcap is then run on the PDC under test to quantify the aggregation processing delay by subtracting the time difference between "receiving µMPMU message → forwarding to the master station". However, the reusable testing method for distribution network PDCs uses a fixed waiting window, reserving a constant timeout for each time frame to wait for data to "arrive". This approach is simple to implement, but insufficient waiting occurs when link conditions deteriorate, leading to an increased intra-frame field missing rate.

[0004] Therefore, a communication quality monitoring and buffer correction scheme that is application-layer oriented and directly aligned with time frames is needed: a scheme that can adaptively, with limited amplitude and hysteresis, dynamically adjust the waiting window based on link state measurement and judgment results, so that the "frame alignment rate" is maintained above the target threshold, while minimizing the overall end-to-end latency as much as possible. Summary of the Invention

[0005] To overcome the aforementioned deficiencies in the prior art, this invention provides a data acquisition terminal for communication quality monitoring. The terminal uses a 1-second time frame as a unified time axis, triggers a round-trip time stamp (RTT) message at the beginning of each second frame to complete RTT measurement and calculate jitter, packet loss, integrity anomaly counts, and the 95th percentile value of the RTT. Then, it constructs a quantile-oriented adaptive waiting mechanism, using the P95% of the RTT of the past N frames as the primary driving variable to adaptively adjust the waiting window, thereby solving the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: A data acquisition terminal for communication quality monitoring includes an energy meter interface module for reading meter values ​​from the energy meter and attaching a local time stamp, and further includes: The time synchronization module is used to align the terminal clock to Coordinated Universal Time and generate time frames with a 1-second granularity. The communication quality monitoring module is used to measure communication quality indicators and calculate the value at the 95th position within a preset statistical time period as the statistical result. The data slicing and statistics module is used to generate 10T0 segments based on 10 fundamental frequency periods within each time frame; The caching and out-of-order correction module is used to group, rearrange, and compensate for missing values ​​of instrument values ​​and communication quality indicators according to the same time frame before uploading, and complete the framing within the waiting window. The alarm and upload module is used to output the value at the 95th position as a statistical result to the local alarm. The caching and out-of-order correction module also constructs a quantile-oriented adaptive waiting mechanism. To reduce out-of-order and end-to-end latency caused by public network jitter, the 95th quantile of the round-trip time (RTT) of the last N frames is used as the main driving variable. During the cold start phase of the acquisition terminal, the waiting window of the first batch of acquired samples is initialized, and upper and lower boundaries and low and high thresholds for hysteresis are set. Subsequently, a sliding window evaluation cycle is entered, and the 95th quantile of the RTT is calculated. When the 95th quantile of the RTT is higher than the high threshold of the hysteresis threshold for M consecutive evaluation cycles, the waiting window is enlarged proportionally. When the 95th quantile of the RTT is lower than the low threshold of the hysteresis threshold for M consecutive evaluation cycles, the waiting window is shrunk proportionally.

[0007] Further, the calculation process for the 95th percentile of the round-trip delay is as follows: Within a sliding window, successfully completed round-trip measurement samples from this terminal are aggregated. Each sample is a round-trip delay obtained by pairing four timestamps, and passes message authentication and timestamp playback detection, is aligned with the current frame number, and has acceptable time quality. The samples are then cleaned, with unanswered probes recorded as "missing," used only for recording packet loss rate and not included in quantile statistics. If the number of valid samples is lower than a set lower limit, the collection is extended until the lower limit is reached or the labeling period result is deemed low confidence. After cleanup, all valid round-trip delays are sorted from smallest to largest, and values ​​falling near the 95th percentile of the total sample count are located. When this position falls exactly on a sample, the delay of that sample is directly taken. When it falls between two samples, a representative value between the two is interpolated proportionally and used as the 95th percentile of the round-trip delay for this period.

[0008] Furthermore, the electricity meter interface module includes physical access, session management, object reading and writing, and caching and verification units. It connects to the meter via RS-485, Ethernet, or photoelectric port, and uses DLMS / COSEM or equivalent protocols to read URMS, IRMS, active power, reactive power, frequency, and meter clock objects. It adds a UTC timestamp and TF(k) to each reading and writes it to an unlocked circular buffer. At the same time, it sets explicit quality flags for missing, timeout, and verification failure records.

[0009] Furthermore, the time synchronization module includes an NTP time synchronization unit and a GNSS time synchronization unit. The GNSS provides pulses per second for phase locking and frequency calibration of the local clock and generates TF(k). When the network is unstable, it enters a hold state and compensates according to historical frequency offset. During leap seconds, a smoothing strategy is adopted to make the application layer time monotonically increase.

[0010] Furthermore, the communication quality monitoring module completes a three-timestamp session via UDP, a custom application layer handshake, or a "service upload + receipt" mode, calculates the round-trip delay, defines jitter by the difference between adjacent frames, defines the packet loss rate by "number of unanswered requests / number of probes sent", and performs message authentication code and replay detection verification on the samples using a sliding window to remove extreme values ​​and median filtering.

[0011] Furthermore, the data slicing and statistics module selects target segments in multiple candidate 10T0 windows within each TF(k) according to a random or round-robin strategy, performs DC removal, amplitude limiting and window function processing on the data within the segment, and extracts the effective voltage value, frequency and total harmonic distortion. Within a preset statistical time period, it outputs the 95th percentile value of each indicator and the sample and rejection counts.

[0012] Furthermore, the communication quality monitoring module performs message authentication code verification and timestamp replay detection on the response. When a session fails due to network switching or network address translation reconstruction, it adopts exponential backoff retries no more than twice. When the time quality degradation exceeds the threshold, it adds a low time quality level mark to the frame record.

[0013] Furthermore, at the end of the preset statistical period, the alarm and transmission module jointly determines the 95th percentile of latency and the 95th percentile of packet loss rate. When both exceed the limit, a level 1 alarm is triggered and a relaxation strategy is implemented in conjunction with the adaptive waiting window. When only one exceeds the limit, a level 2 alarm is triggered. The alarm escalation and de-escalation are completed using the hysteresis consistency criterion of H consecutive intervals.

[0014] Furthermore, the measurement records output by the electricity meter interface module are packaged using a unified structure, including at least the device identifier, object identifier, measured value, unit, quality bit, UTC timestamp, TF(k), and signature digest. The data packets sent by the alarm and uploading module include at least the time frame identifier, preset statistical time period number, out-of-limit statistical snapshot, corresponding threshold, event label, link identifier, and time quality bit, and are retransmitted and played back in chronological order after network interruption recovery.

[0015] The technical effects and advantages of the data acquisition terminal for communication quality monitoring in this invention are as follows: The data acquisition terminal uses a 1-second time frame as a unified time axis, combines NTP / GNSS time synchronization with "triple timestamp" RTT measurement, and employs a P95% adaptive waiting window (with hysteresis and upper / lower limits) based on nearly N frame RTTs to achieve frame aggregation and out-of-order correction under public network jitter conditions. Simultaneously, it performs lightweight statistics on power and communication dual-domain data according to 10T0 segments and intervals at P95%, and transmits the data bound to the frame number. Combined with integrity verification, quality bits, and event tags, it achieves traceability and auditability. This significantly improves the intra-frame field integrity rate, reduces end-to-end latency and the risk of misplaced data entry / false alarms. Even during congestion or link switching, it ensures 1Hz acquisition of critical objects and controllable bandwidth. Alarm classification and hysteresis suppress alarm jitter, comprehensively improving data consistency, real-time performance, and reliability in scenarios such as power distribution information acquisition / WAMS. Attached Figure Description

[0016] Figure 1 This is a structural diagram of a data acquisition terminal for communication quality monitoring according to the present invention.

[0017] Figure 2 This is a schematic diagram of the structure of a data acquisition terminal using existing technology.

[0018] Figure 3 This is a schematic diagram illustrating the adaptive waiting window adjustment principle for quantiles in this invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] Example 1 This invention discloses a data acquisition terminal for monitoring communication quality, comprising: An energy meter interface module is used to read meter values ​​from an energy meter and attach a local time stamp. The time synchronization module is used to align the terminal clock to Coordinated Universal Time and generate time frames with a 1-second granularity. The communication quality monitoring module is used to measure communication quality indicators and calculate the value at the 95th position within a preset statistical time period as the statistical result. The data slicing and statistics module is used to generate 10T0 segments based on 10 fundamental frequency periods within each time frame; The caching and out-of-order correction module is used to group, rearrange, and compensate for missing values ​​of instrument values ​​and communication quality indicators according to the same time frame before uploading, and complete the framing within the waiting window. The alarm and upload module is used to output the value at the 95th position as a statistical result to the local alarm.

[0021] Furthermore, the electricity meter interface module, used to read meter values ​​from the electricity meter and attach a local time stamp, includes: the electricity meter interface module consists of a physical access unit, a session management unit, an object read / write unit, and a buffer and verification unit, and is connected to the meter via RS-485 / Ethernet / photoelectric port, adopting IEC 62056. After establishing an HDLC link and completing LLS / HLS authentication using DLMS / COSEM or an equivalent protocol, and negotiating LN references and transmission segmentation parameters, the system enters a stable reading state. Within the k-th time frame (TF(k)), a data acquisition cycle is triggered. At a frequency of at least 1Hz, URMS (1-0:32.7.0), IRMS (1-0:31.7.0), active power P (1-0:1.7.0), reactive power Q (1-0:3.7.0), frequency f (0-0:14.7.0), and the meter clock (0-0:1.0.0) are read according to the preset OBIS list. FCS verification and frame reassembly are performed at the driver layer, and APDU integrity verification and value range review are performed at the application layer. After successful parsing, a local UTC timestamp and TF(k) frame number are appended to each reading and written to an unlocked circular buffer. If the meter clock and local time synchronization deviation exceed a threshold, a "time sequence equals" error is added. The meter interface module is marked with a "level" for subsequent statistics. It supports segmented readings and windowed transmission. The number of retransmissions per frame and the interaction timeout are dynamically adjusted within a configurable range. When a timeout, verification failure, or object unavailability occurs, an exponential backoff and fast probing strategy is used to restore the session and explicitly mark missing readings to ensure that upper-layer out-of-order correction can be completed frame by frame. To reduce bandwidth usage, the module aggregates multiple object read requests within the same TF(k) and enables batch reading. It supports setting independent polling cycles and degradation thresholds for high-frequency objects. When the link is congested or the meter load increases, it automatically reduces the acquisition frequency of non-critical objects while maintaining 1Hz acquisition for critical objects such as URMS and f. All outputs are encapsulated in a unified measurement record structure, including device identifier, OBIS, measured value, unit, quality bit, timestamp, and signature digest fields, providing a consistent data entry for subsequent 10T0 segment calculation, P95% statistics, and same-frame binding.

[0022] Furthermore, the time synchronization module, used to align the terminal clock to Coordinated Universal Time and generate time frames with a 1-second granularity, includes: the time synchronization module consists of an NTP time synchronization unit, a GNSS time synchronization unit, a clock discipline unit, a time quality assessment unit, a time frame management unit, and an event recording unit, using a unified "1-second time frame" as the upper-level scheduling reference.

[0023] In this embodiment, publicly available hardware and software and reproducible network simulation parameters are used. The frame alignment rate (FCR) and end-to-end latency P95% are observed under different public network conditions by comparing three time frames: 0.5s, 1s, and 2s. The estimation error of power quality quantile statistics and system resource consumption are evaluated. The results show that in typical cellular / carrier jitter scenarios, the 1s frame maintains a high alignment rate while significantly lower latency than the 2s frame, and is significantly better than the 0.5s frame in terms of alignment rate and resource efficiency. The specific experimental conditions are as follows: 1. Terminal under test (DUT): x86 industrial computer (4C / 8G), Linux 5.x, Chrony (NTP client), DLMS / COSEM emulator, three-timestamp probe program (UDP response), implementing 10T0 segments, P95% statistics, and adaptive waiting window W (upper and lower bounds 0.2–2.0s, N=60, M=3, hysteresis low threshold T_low=300ms / high threshold T_high=700ms). 2. Time Synchronization: Chrony aligns with two local NTP servers (time residual is stable <10ms), GNSS is not used. 3. Network Simulation: Linux tc / netem injects latency / jitter / packet loss; three reproducible scenarios: S1 (Stable Ethernet): baseline latency 20ms, jitter 5ms, packet loss 0.1%; S2 (4G LTE Medium Jitter): baseline latency 120ms, jitter 60ms, packet loss 0.5%; S3 (NB-IoT / Carrier High Jitter): baseline latency 600ms, jitter 200ms, packet loss 1.0%. 4. Test Duration: Each frame length and scenario combination runs for 30 minutes (a preset statistical time period), repeated 3 times and the median is taken. 5. Sampling and Statistics: Power Quality: Select a 10T0 segment (50 Hz → 0.2 s) per frame, and calculate the P95% of URMS, f, and THD within 30 minutes; Communication Quality: Perform a three-timestamp round-trip measurement per frame, and calculate RTT_P95, JITTER_P95, and LOSS_P95; Reference Baseline: Use the P95% obtained by "full window / high frequency sampling" (0.2s continuous segment) in the same scenario as a reference, and compare the P95% estimation error of each frame length. 6. Index Definitions: FCR (Intra-frame Alignment Rate): The proportion of power segments and communication samples that are successfully grouped into the same TF(k) within the waiting window W; E2E P95: The 95th percentile of the end-to-end delay that can be uploaded from the start of the frame to the record that can be uploaded in the same frame; Resource Usage: Average CPU usage (%), upload bandwidth (kbit / s). Tables 1-3 show the test results: Table 1. In-frame alignment rate and end-to-end delay (median result over 30 min)

[0024] In S2 / S3 public network jitter scenarios, a 1-second frame significantly improves the FCR (approximately +2.6 to +6.3 percentage points) compared to a 0.5-second frame, while significantly reducing E2E P95 (approximately...) compared to a 2-second frame. (0.5~0.8s); In the stable S1 scenario, the difference between the three converges, but 2s still brings unnecessary waiting.

[0025] Table 2. Power Quality P95% Estimation Error and System Resource Utilization

[0026] In terms of P95% statistical precision, 0.5s has the lowest error due to its larger sample size, but at the cost of higher CPU / bandwidth and worse FCR; 2s has the smallest sample size, but the error is amplified; 1s strikes a balance between precision (≤ about 0.7%) and resource consumption, and together with the FCR / E2E advantages in Table 1, it constitutes the "engineering optimal trade-off".

[0027] Table 3. Adaptive convergence of the waiting window W

[0028] The test results show that, under the dual objectives of "achieving the same frame alignment rate and minimizing end-to-end latency", the 1-second time frame exhibits more robust alignment capability and lower latency cost compared to 0.5s and 2s, while maintaining sufficient quantile statistical accuracy and moderate resource consumption. It is the better choice of unified time granularity in this implementation.

[0029] The time synchronization module first initiates NTP initialization during power-on, establishes sessions with at least two authoritative time sources, supports NTS authentication or symmetric key verification, sets minpoll = 10 s and maxpoll = 120 s, and enables a multi-source selection strategy. It uses a sliding window and median filtering to suppress abnormal delays. When the deviation is less than 0.4 s, the PI servo uses a slew mode to drive the local oscillator at a rate of ≤500 ppm. When the deviation exceeds this threshold, a one-time step correction is performed. The GNSS time synchronization unit outputs 1 pulse per second and UTC messages. The clock discipline unit drives the kernel phase-locked discipline with 1 pulse per second, prioritizing the use of the 1 pulse per second phase to reduce the local clock jitter to the millisecond level and feeding back clockstats to the NTP to achieve external consistency. The time quality assessment unit calculates the NTP deviation, 1 pulse per second deviation, and two differential signals in real time, and performs consistency checks at a threshold of 0.1–0.5 s. When the two differential signals exceed the limit, event recording and alarms are triggered. The locking priority is "1 pulse per second locking > multi-source NTP consistency > single-source NTP".

[0030] When the network is unavailable or the upstream is unstable, it enters a hold state, compensating for drift based on the TCXO temperature drift model and historical frequency offset estimation, providing an accrual uncertainty boundary, and maintaining this state for more than 30-120 minutes. When the time quality flag degrades when the min or error upper bound approaches 1s, the time quality is downgraded. To ensure monotonically increasing at the application layer, the time synchronization module provides two strategies for leap second processing: stepping and smoothing. By default, 24-hour linear smoothing is performed within the leap second window, and a leap second flag is written into the event log. The time frame management unit generates a frame flag TF(k) and a frame start interrupt using a local clock aligned with UTC. The NTP / GNSS status and time quality level are broadcast with the frame to the acquisition, communication measurement, and statistics threads. All data records inserted into the upper layer have a UTC timestamp and TF(k) with an additional "time quality bit". The waiting window recommendation value is automatically and temporarily relaxed before and after NTP jitter increases or link switching to maintain the same frame alignment rate. After stabilization, convergence is performed by hysteresis. The event log unit persists the time source switching, two-way differential over-limit, leap second processing, entry / exit, authentication failure, etc., and provides query and upload interfaces to ensure the auditability and traceability of time synchronization.

[0031] Furthermore, the communication quality monitoring module is used to measure communication quality indicators and calculate the value at the 95th position within a preset statistical time period as the statistical result. The preset statistical time period includes: the communication quality monitoring module operates under the drive of a 1s time frame TF(k) provided by the time synchronization module, and includes a probe message generation unit, a three-timestamp session management unit, a timestamp capture and alignment unit, an RTT statistics unit, an integrity and security verification unit, an anomaly and congestion control unit, and an interval summary and uploading unit. In the first 10–50 ms of each TF(k) frame, the probe message generation unit constructs a light-load probe message, with fields including at least: frame number TF(k), terminal transmission time T1 (UTC), sequence number, session ID, and message authentication code (MAC), and sends it to the master station according to the configuration, selecting either UDP / custom application layer handshake or "application layer upload + acknowledgment" mode. When the master station receives the message, it records time T2 and sends back T1, T2, and the master station's transmitted time T3 in the response. When the terminal receives the response, it records time T4. The three-timestamp session management unit completes matching and out-of-order discarding. The timestamp capture and alignment unit calls the kernel high-precision timer to read T1 / T4 and aligns it with the UTC reference provided by the time synchronization module. The communication quality monitoring module calculates the round-trip time RTT = T4 accordingly. T1 (T3 T2), while using the difference between adjacent frames to define jitter, the packet loss rate is obtained by "number of times no response was received / number of times probes were sent" within a preset statistical time period.

[0032] To suppress occasional spikes, the RTT statistics unit uses a sliding window to remove extreme values ​​and median filtering for RTT and jitter. All measured samples, along with frame numbers and quality bits, are written to a buffer and grouped within a waiting window. The integrity and security verification unit performs MAC verification and timestamp replay detection on responses. Failed samples are marked as "integrity anomalies" and added to the anomaly count. When a session fails due to network switching, network address translation reconstruction, or increased main station load, session management retryes ≤2 times with exponential backoff (e.g., 200–1200ms) and marks the detection as missing in the current frame. The anomaly and congestion control unit uses the previous preset statistical time period RTT when probing and transmitting services on the shared link. The size of the probe message (48–256B) and the transmission lead time are appropriately relaxed for _P95 / LOSS_P95 to ensure that there is no perceptible impact on the main service. At the end of each preset statistical time period (15 / 30 / 60min), the interval summary uploading unit calculates RTT_P95, JITTER_P95, LOSS_P95 and integrity anomaly count respectively, and outputs them "frame-bound" together with the power quality P95% of the same interval. If the synchronization exceeds the limit, a "frame-linked anomaly" tag is generated. The communication quality monitoring module relies on the time synchronization module for time synchronization, but within a short time synchronization drift, the main station processing delay (T3) can still be offset by three timestamps. T2) and keep the measurement valid. When the time synchronization quality degrades beyond the threshold, a "low time sequence level" is automatically added to provide a basis for interpretation and traceability on the main station side.

[0033] Furthermore, the data slicing and statistics module is used to form 10T0 segments based on 10 fundamental frequency cycles within each time frame. This includes setting the analysis segment within each time frame to 10 fundamental frequency cycles (10 cycles = 0.2s for a 50Hz system; 12 cycles = 0.2s for a 60Hz system). This is based on the fact that the integer window length is consistent with international power quality measurement standards. IEC 61000-4-30 (Class A) specifies 10 / 12 power frequency cycles as the basic measurement time interval for voltage and frequency, etc. The 61000-4-7 uses a 0.2s analysis window for grouping and calculating the harmonic / interharmonic spectrum. Using an integer window avoids spectral leakage caused by non-integer truncation, ensuring comparability and reproducibility of fundamental and harmonic amplitudes, THD, and other indicators across different devices. The 0.2s segment's divisibility with the 1s time frame used in this embodiment allows for the selection of one segment per second from five candidate segments using a random or round-robin strategy, reducing computational and communication load and eliminating the periodic bias of fixed window selection. Using 10T0 (or 12T0 at 60Hz) as the segment length simultaneously meets multiple requirements for standard consistency, spectral accuracy, statistical robustness, and system real-time performance, directly reproducing and supporting the data processing link of this embodiment.

[0034] The data slicing and statistics module operates under the 1-second time frame TF(k) provided by the time synchronization module. It is built around a lightweight strategy of "analyzing only one 10T0 segment per second" and includes a window scheduling unit, a preprocessing unit, a feature extraction unit, and an interval statistics unit.

[0035] The window scheduling unit determines the target segment from multiple available 10T0 windows (0.2s for 50Hz) within TF(k) using a random or round-robin strategy. The random strategy uses the device identifier and TF(k) as seeds to ensure reproducibility, while the round-robin strategy rotates the windows sequentially according to their numbers to balance time-domain coverage. Subsequently, the preprocessing unit reads the URMS, IRMS, P, Q, f, or harmonic object values ​​of the same frame as the segment from the segment data obtained from the energy meter interface. If the meter supports instantaneous sampling or harmonic component objects, the samples within the segment are processed for DC removal, amplitude limiting, and window function processing. If only one weekly RMS / frequency sequence is available, the subsequence within the segment is used for approximate estimation, and the "estimation source" is marked. In the feature extraction unit, the data slicing and statistics module calculates the effective voltage value and frequency of the segments. If harmonic data exists, the total harmonic distortion (THD) or a subset thereof is obtained according to the 1st to 25th / 50th harmonics. The frequency can be given by the table or estimated by phase increment / zero-crossing interpolation. The segment results, along with TF(k), phase identifier and quality bit, are written into the circular buffer. Out-of-bounds or outlier values ​​are removed using the median / quartile box rule, and the original records are retained for traceability. The interval statistics unit uses 15min, 30min or 60min as the preset statistical time period, and calculates the P95th percentile values ​​of the effective voltage value, frequency and total harmonic distortion index of the segment results formed every second in the same interval, and outputs the statistical vector organized by phase and the sample count and removal count.

[0036] Furthermore, the caching and out-of-order correction module is used to group, rearrange, and compensate for missing data of meter values ​​and communication quality indicators according to the same time frame before uploading, and complete the framing within the waiting window. This includes: the caching and out-of-order correction module establishes a frame bucket in memory for each TF(k) and accepts records from the energy meter interface and communication quality monitoring in a circular buffer manner. Each record carries a Coordinated Universal Time (UTC) timestamp, frame number, quality bit, and signature digest. The caching and out-of-order correction module does not make any assumptions about the arrival order. It first sorts and removes duplicates based on the timestamp, and then uses "missing placeholders" to ensure the integrity of the frame structure. Fields that are still incomplete when the waiting window expires are entered into the database with a missing flag, while the original sample is retained for traceability.

[0037] To reduce out-of-order delivery and end-to-end latency caused by public network jitter, the caching and out-of-order correction module constructs a "quantile-oriented adaptive waiting mechanism," such as... Figure 3As shown, the 95th percentile value RTT_P95(t) is calculated using round-trip delay samples from the most recent period and compared with a pair of upper and lower threshold intervals with hysteresis. When RTT_P95(t) is above the upper threshold for M consecutive evaluation periods, the waiting window W(t) is increased by a step by a preset amplification factor (not exceeding W_max) to improve the alignment rate of the same frame. When RTT_P95(t) is below the lower threshold for M consecutive evaluation periods, W(t) is decreased by a step by a contraction factor (not lower than W_min) to reduce the end-to-end latency. If RTT_P95(t) falls within the threshold band or does not meet the condition of M consecutive periods, it remains unchanged. The left side of the figure shows a three-state machine (amplify / hold / shrink) that provides the decision logic. The right side uses two time curves to illustrate the relationship between the step adjustment of W(t) and the fluctuation of RTT_P95(t), and marks the upper and lower threshold bands and the boundaries of W_min / W_max. This reflects the dynamic trade-off and stable control between "intra-frame integrity rate" and "latency" through "quantile-driven + hysteresis + continuous periodic confirmation".

[0038] The specific process of the quantile-oriented adaptive waiting mechanism is as follows: The P95% of the round-trip time (RTT) of the last N frames (i.e., the most recent N consecutive 1-second time frames, where N is a configurable preset parameter) is used as the main driving variable. Multidimensional statistics and state machine control are introduced to continuously, finitely, and auditably dynamically adjust the waiting window. During the cold start phase, the acquisition terminal initializes the waiting window for the first batch of acquired samples (i.e., the same-frame identifier TF(k), terminal transmission time T1, master station reception recording time T2, master station feedback time T3, terminal reception acknowledgment recording time T4, and RTT) in the range of 0.8-1.2s, setting upper and lower boundaries [0.2s, 2.0s] and low and high hysteresis thresholds. Subsequently, a sliding window evaluation period is entered, with a window length of 30–120. During this period, extreme value removal and Ha operations are performed on the RTT samples. MPEL screening eliminates single-point anomalies by calculating the 95th percentile of round-trip time (RTT). The 95th percentile is compared to the previous window and combined with an absolute threshold to determine whether the link is in a "jitter increase" or "jitter decrease" state. When the 95th percentile of RTT is higher than the high threshold of the hysteresis threshold for M consecutive evaluation periods (M ranges from 2 to 5), the waiting window is proportionally enlarged with an amplification factor of 1.0-1.8 to improve intra-frame alignment and absorb out-of-order frames. The amplification action is constrained by an upper limit to avoid excessive waiting. When the 95th percentile of RTT is lower than the low threshold of the hysteresis threshold for M consecutive evaluation periods, the waiting window is proportionally shrunk with a shrinkage factor of 0.6-0.9 to reduce unnecessary waiting, compress end-to-end propagation delay, and is protected by a lower limit to avoid excessively narrow window cuts that could lead to numerous missing frames.

[0039] The calculation process for the 95th percentile of the round-trip delay is as follows: Using a 1-second time frame as the beat, successfully completed round-trip measurement samples from this terminal are aggregated within a sliding window. Each sample is a round-trip delay obtained by pairing four timestamps, and passes message authentication and timestamp playback detection, is aligned with the current frame number, and has acceptable time quality. The samples are then cleaned by removing individual points with protocol anomalies, obvious out-of-bounds errors, or those identified as outliers by rules. Probes that do not receive a response are recorded as "missing" and are only used to record the packet loss rate, not for quantile statistics. If the number of valid samples is lower than a set lower limit, the collection is extended until the lower limit is reached or the result for that period is marked as low confidence. After cleanup, all valid round-trip delays are sorted from smallest to largest, and values ​​falling near the 95th percentile of the total sample count are located: if this position falls exactly on a sample, the delay of that sample is directly taken; if it falls between two samples, a representative value between the two is interpolated proportionally and used as the 95th percentile of the round-trip delay for that period. The entire process runs independently along the link dimension and outputs the 95th percentile of the round-trip latency and the accompanying quality metadata at the end of the interval. This reflects the true end-to-end latency of "the vast majority of requests" while avoiding the misleading effect of a few abnormal spikes on the results.

[0040] Furthermore, the alarm and upload module is used to output the value at the 95th percentile as a statistical result for local alarm output. This includes: the alarm and upload module uses quantile statistics generated within a preset statistical time period as the trigger basis, and after each interval, it determines the communication latency P95%, packet loss rate P95%, and integrity anomaly count. When both the latency P95% and packet loss rate P95% exceed preset thresholds, a Level 1 alarm is triggered, immediately increasing the upload frequency to 1s–5s and simultaneously outputting a local audible / visual / relay alarm. When only one exceeds the limit, a Level 2 alarm is triggered, adjusting the upload frequency to 5s–30s and recording the event. To avoid alarm jitter, the module employs a consistency criterion and hysteresis strategy for H consecutive intervals. Only after the indicators continuously recover to within the threshold and remain within H intervals will the alarm automatically degrade until it is cleared. All uploaded data packets uniformly include a time frame identifier, a preset statistical time period number, a snapshot of out-of-limit statistics, a corresponding threshold, an event label (high latency, high packet loss, integrity anomaly, frame linkage anomaly, etc.), a link identifier, a time quality bit, and a message authentication code. During link jitter, the alarm and upload modules are linked with the waiting window. Under a level 1 alarm, the waiting window is widened to improve the frame aggregation rate. Under a level 2 alarm, upload is tightened at a limited rate to reduce link occupancy. When the network is briefly interrupted, local buffering and retransmission are enabled. After upload is restored, the data is replayed in chronological order to ensure auditability and data integrity on the main station side.

[0041] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

[0042] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A data acquisition terminal for communication quality monitoring, comprising an energy meter interface module for reading meter values ​​from an energy meter and attaching a local time stamp, characterized in that, Also includes: The time synchronization module is used to align the terminal clock to Coordinated Universal Time and generate time frames with a 1-second granularity. The communication quality monitoring module is used to measure communication quality indicators and calculate the value at the 95th position within a preset statistical time period as the statistical result. The data slicing and statistics module is used to generate 10T0 segments based on 10 fundamental frequency periods within each time frame; The caching and out-of-order correction module is used to group, rearrange, and compensate for missing values ​​of instrument values ​​and communication quality indicators according to the same time frame before uploading, and complete the framing within the waiting window. The alarm and upload module is used to output the value at the 95th position as a statistical result to the local alarm. The caching and out-of-order correction module also constructs a quantile-oriented adaptive waiting mechanism. To reduce out-of-order and end-to-end latency caused by public network jitter, the 95th quantile of the round-trip time (RTT) of the last N frames is used as the main driving variable. During the cold start phase of the acquisition terminal, the waiting window of the first batch of acquired samples is initialized, and upper and lower boundaries and low and high thresholds for hysteresis are set. Subsequently, a sliding window evaluation cycle is entered, and the 95th quantile of the RTT is calculated. When the 95th quantile of the RTT is higher than the high threshold of the hysteresis threshold for M consecutive evaluation cycles, the waiting window is enlarged proportionally. When the 95th quantile of the RTT is lower than the low threshold of the hysteresis threshold for M consecutive evaluation cycles, the waiting window is shrunk proportionally.

2. The data acquisition terminal for communication quality monitoring according to claim 1, characterized in that, The calculation process for the 95th percentile of the round-trip delay is as follows: Within a sliding window, successfully completed round-trip measurement samples from this terminal are aggregated. Each sample is a round-trip delay obtained by pairing four timestamps, and passes message authentication and timestamp playback detection, is aligned with the current frame number, and has acceptable time quality. The samples are then cleaned, with unanswered probes recorded as "missing," used only for recording packet loss rate and not included in quantile statistics. If the number of valid samples is below a set lower limit, the collection is extended until the lower limit is reached or the labeling period result is deemed low confidence. After cleanup, all valid round-trip delays are sorted from smallest to largest, and values ​​falling near the 95th percentile of the total sample count are identified. When this position falls on a specific sample, the delay of that sample is directly taken. When it falls between two samples, a representative value between the two is interpolated proportionally and used as the 95th percentile of the round-trip delay for this period.

3. The data acquisition terminal for communication quality monitoring according to claim 1, characterized in that, The electricity meter interface module includes physical access, session management, object reading and writing, and buffering and verification units. It connects to the meter via RS-485, Ethernet, or photoelectric port. It reads URMS, IRMS, active power, reactive power, frequency, and meter clock objects using DLMS / COSEM or equivalent protocols. It adds a UTC timestamp and TF(k) to each reading and writes it to an unlocked circular buffer. At the same time, it sets explicit quality flags for missing, timeout, and verification failure records.

4. The data acquisition terminal for communication quality monitoring according to claim 3, characterized in that, The time synchronization module includes an NTP time synchronization unit and a GNSS time synchronization unit. The GNSS provides pulses per second for phase locking and frequency calibration of the local clock and generates TF(k). When the network is unstable, it enters a hold state and compensates according to historical frequency offset. During leap seconds, a smoothing strategy is used to make the application layer time monotonically increase.

5. The data acquisition terminal for communication quality monitoring according to claim 1, characterized in that, The communication quality monitoring module completes a three-timestamp session via UDP, a custom application layer handshake, or a "service upload + receipt" mode, calculates the round-trip delay, defines jitter by the difference between adjacent frames, defines the packet loss rate by "number of unanswered requests / number of probes sent", and performs message authentication code and replay detection verification on the samples using a sliding window to remove extreme values ​​and median filtering.

6. The data acquisition terminal for communication quality monitoring according to claim 1, characterized in that, The data slicing and statistics module selects target segments in multiple candidate 10T0 windows within each TF(k) using a random or round-robin strategy. It performs DC removal, amplitude limiting, and window function processing on the data within the segments and extracts the effective voltage value, frequency, and total harmonic distortion. Within a preset statistical time period, it outputs the 95th percentile value of each indicator and the sample and rejection counts.

7. The data acquisition terminal for communication quality monitoring according to claim 1, characterized in that, The communication quality monitoring module performs message authentication code verification and timestamp playback detection on the response. When the session fails due to network switching or network address translation reconstruction, it adopts exponential backoff retries no more than 2 times. When the time quality degradation exceeds the threshold, it adds a low time quality level mark to the frame record.

8. The data acquisition terminal for communication quality monitoring according to claim 1, characterized in that, At the end of the preset statistical period, the alarm and transmission module jointly determines the 95th percentile of latency and the 95th percentile of packet loss rate. When both exceed the limit, a level 1 alarm is triggered and a relaxation strategy is implemented in conjunction with the adaptive waiting window. When only one exceeds the limit, a level 2 alarm is triggered. The alarm escalation and de-escalation are completed using the hysteresis consistency criterion of H consecutive intervals.

9. The data acquisition terminal for communication quality monitoring according to any one of claims 2-8, characterized in that, The measurement records output by the electricity meter interface module are packaged in a unified structure, including at least the device identifier, object identifier, measured value, unit, quality bits, UTC timestamp, TF(k), and signature digest. The data packets sent by the alarm and uploading module include at least the time frame identifier, preset statistical time period number, out-of-limit statistical snapshot, corresponding threshold, event label, link identifier, and time quality bits, and are retransmitted and played back in chronological order after network interruption recovery.

Images