A collection terminal for communication quality monitoring

Abstract

The application relates to the technical field of communication quality, and particularly discloses a collection terminal for communication quality monitoring, which comprises the following modules: an electric energy meter interface module, which reads instrument values from an electric energy meter; a time synchronization module, which is used for aligning a terminal clock to coordinated universal time and generating a time frame with 1s granularity; a communication quality monitoring module, which is used for measuring a communication quality index and calculating a value at the 95% position in a preset statistical time period as a statistical result; a data slicing and statistical module, which is used for forming 10T0 segments according to 10 fundamental wave periods in each time frame; a cache and out-of-order correction module, which is used for grouping, out-of-order rearranging and loss compensation of the instrument values and the communication quality index according to the same time frame before uploading and completing framing in a waiting window; and an alarm and uploading module, which is used for outputting the value at the 95% position as the statistical result to a local alarm. The application reduces the overall end-to-end time delay of the collection terminal.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of communication quality monitoring, more particularly, to a communication quality monitoring acquisition terminal. BACKGROUND

[0002] The power distribution network and the power consumption information acquisition system use a large number of smart meters and acquisition terminals in a wide range, and the meter values are uploaded to the MDMS / SCADA / DMS platform of the main station through the public network (cellular / internet / VPN). Compared with the private network, the scheduling and congestion of the public network link are uncontrollable, and frequent jitter, queuing and retransmission occur, thereby causing out-of-order arrival and end-to-end delay fluctuation. In the monitoring business driven by the second-level time sequence (such as second-by-second reading, event identification, and short window statistics), this network uncertainty will directly destroy the data aggregation of the "same time frame": if different sources (electric energy measurement of the same terminal, communication measurement, or multiple fields of the same frame) are misaligned in the main station end, it will cause statistical caliber drift, abnormal detection delay or false alarm, and even make the control and alarm decision lose basis. In practice, the causes of public network jitter include wireless air interface scheduling, cross-cell / base station switching, network address translation / firewall state reconstruction, TCP retransmission and congestion control state transition, etc., resulting in long tail characteristics and time correlation of delay distribution, and it is difficult to reflect the real arrival law of "most data" by simply relying on the average value or instantaneous measurement.

[0003] The existing public document 1 (main indicators and test methods of power distribution network phasor data concentrator, 2019) first introduces the typical structure of the power distribution network WAMS, such as Figure 2As shown, a "master station-PDC-µMPMU" hierarchical distributed architecture is adopted: the top layer is a WAMS master station, which runs advanced applications such as satellite clock-based wide-area phasor measurement, state estimation, early warning, model parameter identification and wide-area protection; the middle layer is a phasor data concentrator (PDC) deployed at nodes such as substations and switching stations, which is responsible for gathering, aligning and uploading phasor data from each access device; the backbone communication between the master station and each PDC is an optical fiber ring network accessed by optical fiber SDH / MSTP switches; the access layer is a micro multi-functional synchronous phasor measurement device (µMPMU) distributed at 10kV distribution line ring main units, switching stations or pole-mounted switches, which, in addition to phasor measurement and event recording, also integrates distribution automation "three remote", fault diagnosis and protection control functions; then a reusable test method and platform for distribution network PDC are proposed, on the hardware side a hybrid communication bench is built with real µMPMU from NARI / XJ, GPS timing, relay protection tester and power line carrier (PLC) module, on the software side a PMU Emulator (VC++, complying with GB / T26865.2-2011, up to ≤30 concurrent, supporting 1-200 frames / s, phasor / analogue / switching quantity configurable, and can artificially inject 10-500 ms cross-channel fixed delay) and a simulated master station are self-developed, and then an automatic packet capturing program based on libpcap is run on the measured PDC to quantify the gathering processing delay by taking the difference between the two time instants of "receiving µMPMU packet→forwarding to master station". However, the reusable test method for distribution network PDC adopts a fixed waiting window, which reserves a constant timeout for each time frame to wait for the data to "arrive", which is simple to implement, but when the link condition deteriorates, the waiting is insufficient and the frame field missing rate rises.

[0004] Therefore, an application layer-oriented, time frame directly aligned communication quality monitoring and buffer correction scheme is needed: to make adaptive, limited amplitude, hysteresis dynamic adjustment of the waiting window based on link state metrics and decision results, so that the "same frame alignment rate" is maintained above the target threshold, while the overall end-to-end delay is as low as possible. SUMMARY

[0005] In order to overcome the above-mentioned defects of the prior art, the application provides a communication quality monitoring acquisition terminal, which takes 1s time frame as a unified time axis, triggers a round-trip timestamp message at the beginning of each second to complete RTT measurement and calculate jitter, packet loss, integrity anomaly count and 95% quantile value of round-trip time delay, and then constructs an adaptive waiting mechanism oriented to quantile number, taking the P95% of nearly N frame round-trip time delay RTT as the main driving quantity to adaptively adjust the waiting window, so as to solve the problems raised in the background art.

[0006] To achieve the above object, the application provides the following technical scheme:

[0007] A collection terminal for communication quality monitoring, comprising an electric energy meter interface module for reading meter values from an electric energy meter and attaching a local time label, further comprising:

[0008] a time synchronization module for aligning the terminal clock to Coordinated Universal Time and generating time frames with a granularity of 1s;

[0009] a communication quality monitoring module for measuring communication quality indicators and calculating the value at the 95th percentile within a preset statistical time period as a statistical result;

[0010] a data slicing and statistics module for forming 10T0 segments within each time frame according to 10 fundamental wave periods;

[0011] a cache and out-of-order correction module for grouping, out-of-order rearranging and missing compensation of meter values and communication quality indicators according to the same time frame before uploading and completing framing within a waiting window;

[0012] an alarm and uploading module for outputting a local alarm with the value at the 95th percentile as a statistical result;

[0013] The cache and out-of-order correction module further constructs an adaptive waiting mechanism for quantile to reduce out-of-order and end-to-end delay caused by public network jitter, taking the 95th percentile value of near N-frame round-trip time RTT as the main driving force, initializing the first batch of collection samples to the waiting window in the cold start stage of the collection terminal, setting the upper and lower boundaries and the hysteresis threshold low threshold and high threshold; then entering the sliding window evaluation period, calculating the 95th percentile value of the round-trip time, when the 95th percentile value of the round-trip time is higher than the hysteresis threshold high threshold for continuous M evaluation periods, the waiting window is enlarged by a certain proportion; when the 95th percentile value of the round-trip time is lower than the hysteresis threshold low threshold for continuous M evaluation periods, the waiting window is contracted by a certain proportion.

[0014] Further, the calculation process of the 95th percentile value of the round-trip time is as follows: in the sliding window, the successful round-trip measurement samples of the terminal are converged, each sample is obtained by pairing four time stamps, and the round-trip time is obtained, and the message authentication and time stamp playback detection are aligned with the current frame number and the time quality is qualified; then purify the samples, mark the unresponded probes as "missing" and only use them to record the packet loss rate without entering the quantile statistics; if the number of valid samples is lower than the set lower limit, extend the collection until the lower limit is reached or the period result is low confidence; after purifying, sort all valid round-trip times from small to large, locate the value near the 95th percentile of the overall sample number: when the position falls on a certain sample, directly take the time delay of the sample; when it falls between two samples, a representative value between the two is obtained by proportional interpolation, which is taken as the 95th percentile value of the round-trip time of the period.

[0015] Further, the electric energy meter interface module includes physical access, session management, object read-write, and cache and verification units, connects with the meter through RS-485, Ethernet or optical-electric port, reads URMS, IRMS, active power, reactive power, frequency and meter clock objects using DLMS / COSEM or equivalent protocol, adds UTC timestamp and TF(k) to each reading and writes into a non-locking ring buffer, and sets explicit quality marks for missing, timeout and verification failure records.

[0016] Further, the time synchronization module includes an NTP time synchronization unit and a GNSS time synchronization unit, the GNSS provides a pulse per second for phase locking and frequency calibration of the local clock and generates TF(k), enters a hold state when the network is unstable and compensates according to the historical frequency deviation, and adopts a smoothing strategy during a leap second to make the application layer time monotonically increasing.

[0017] Further, the communication quality monitoring module completes a three-time stamp session through UDP, a custom application layer handshake or a "service submission + receipt" mode, calculates the round-trip time, defines the jitter by the difference between adjacent frames, defines the packet loss rate by the number of unanswered times / the number of sent probes, and performs message authentication code and playback detection verification on samples using sliding window de-extreme value and median filtering.

[0018] Further, the data slicing and statistical module selects a target segment in a plurality of 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 in the segment, and extracts the voltage effective value, frequency and total harmonic distortion, and outputs the 95% quantile value of each index and the sample and rejection count within a preset statistical time period.

[0019] Further, the communication quality monitoring module performs message authentication code verification and timestamp playback detection on the response, retries no more than twice using exponential backoff when network switching or network address translation reconstruction causes session failure, and adds a low time quality level mark to the frame record when the time synchronization quality degradation exceeds the threshold.

[0020] Further, the alarm and submission module jointly determines the delay 95% quantile value and the packet loss rate 95% quantile value at the end of the preset statistical time period, enters a first level alarm when both exceed the limit and performs a relaxation strategy in conjunction with an adaptive waiting window, enters a second level alarm when only one exceeds the limit, and uses a continuous H interval hysteresis consistency criterion to complete alarm escalation and degradation.

[0021] Further, the measurement record output by the electric energy meter interface module is packaged in a unified structure, and at least includes device identification, object identification, measured value, unit, quality bit, UTC timestamp, TF(k) and signature digest, while the upload data packet of the alarm and upload module at least includes time frame identification, preset statistical time period number, out-of-limit statistical quantity snapshot, corresponding threshold, event label, link identification and time quality bit, and after network interruption is restored, the playback is supplemented in time sequence.

[0022] The technical effect and advantages of the collection terminal for communication quality monitoring of the application are as follows: the collection terminal of the application takes 1s time frame as a unified time axis, combines NTP / GNSS time setting and "three time stamp" RTT measurement, and realizes frame alignment and out-of-order correction under public network jitter condition by using P95% adaptive waiting window (with hysteresis and upper and lower limits) based on nearly N frames of RTT; at the same time, the electric energy and communication dual-domain data are statistically lightened according to 10T0 segments and interval P95%, and are uploaded in combination with frame number binding, integrity check, quality bit and event label to realize traceability and auditability. Therefore, the field completeness rate in the frame is significantly improved, the end-to-end delay and misplacement storage / misreporting risk are reduced; when congestion or link switching occurs, the 1Hz collection of key objects and controllable bandwidth are still guaranteed, alarm grading + hysteresis is used to suppress alarm jitter, and the data consistency, real-time performance and reliability of the power distribution information collection / WAMS and other scenes are overall improved. BRIEF DESCRIPTION OF DRAWINGS

[0023] Figure 1 The structure diagram of the collection terminal for communication quality monitoring of the application.

[0024] Figure 2 The structure diagram of the collection terminal for communication quality monitoring of the application.

[0025] Figure 3 The principle diagram of the adaptive waiting window adjustment for the application. DETAILED DESCRIPTION

[0026] The technical solutions in the embodiments of the application will be clearly and completely described below with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are only part of the embodiments of the application, rather than all the embodiments of the application. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the application.

[0027] Embodiment 1

[0028] The collection terminal for communication quality monitoring of the application comprises:

[0029] The electric energy meter interface module is configured to read the instrument value from the electric energy meter and attach a local time label.

[0030] a time synchronization module for aligning the terminal clock to the coordinated universal time and generating a time frame with a granularity of 1s;

[0031] a communication quality monitoring module for measuring a communication quality indicator and calculating a value at the 95th percentile as a statistical result in a preset statistical time period;

[0032] a data slicing and statistics module for forming 10T0 segments according to 10 fundamental periods in each time frame;

[0033] a buffering and out-of-order correction module for grouping, out-of-order rearranging and missing compensation of the meter values and the communication quality indicators according to the same time frame before uploading and completing the framing in a waiting window;

[0034] an alarm and uploading module for outputting a local alarm with the value at the 95th percentile as the statistical result.

[0035] 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.

[0036] 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.

[0037] In this embodiment, publicly available software and hardware are used with reproducible network simulation parameters. The frame alignment rate (FCR) and end-to-end latency P95% are observed under different public network conditions for time frames of 0.5s, 1s, and 2s. The estimation error of the power quality percentile statistics and the system resource occupation are evaluated. The results show that in a typical cellular / carrier jitter scenario, the 1s frame has a significantly lower latency than the 2s frame while maintaining a high alignment rate, and significantly outperforms the 0.5s frame in terms of alignment rate and resource cost-effectiveness. The experimental conditions are as follows: 1. DUT: x86 industrial computer (4C / 8G), Linux 5.x, Chrony (NTP client), DLMS / COSEM simulation reader, three-time-stamp detection program (UDP response), 10T0 segment, 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 two local NTP servers (time residual stable <10ms), without GNSS. 3. Network simulation: Linux tc / netem injects delay / jitter / loss; three reproducible scenarios: S1 (Ethernet stable): baseline delay 20ms, jitter 5ms, loss 0.1%; S2 (4G LTE moderate jitter): baseline delay 120ms, jitter 60ms, loss 0.5%; S3 (NB-IoT / carrier high jitter): baseline delay 600ms, jitter 200ms, loss 1.0%. 4. Test duration: each frame length and scenario combination runs for 30min (a pre-set statistical time period), repeated 3 times to take the median. 5. Sampling and statistics: power quality: select one 10T0 segment (50Hz→0.2s) per frame, calculate URMS, f, THD P95% within 30min; communication quality: three-time-stamp round-trip measurement per frame, calculate RTT_P95, JITTER_P95, LOSS_P95; reference baseline: P95% obtained under the same scenario with "full window / high-frequency sampling" (0.2s continuous segment) as the reference, compare the P95% estimation error of each frame length. 6. Index definition: FCR (frame alignment rate): within the waiting window W, the proportion of power segments and communication samples successfully grouped according to the same TF(k); E2E P95: the 95th percentile of end-to-end latency from the start of the frame to the record that can be uploaded in the same frame; resource occupation: CPU average occupation (%), upload bandwidth (kbit / s). Tables 1-3 are the test results:

[0038] Table 1 Frame alignment rate and end-to-end latency (30min median results)

[0039]

[0040] In S2 / S3 public network jitter scenarios, 1s frame significantly improves FCR (about +2.6~6.3 percentage points) relative to 0.5s, and significantly reduces E2E P95 (about 0.5~0.8s) relative to 2s; in S1 stable scenarios, the difference between the three converges, but 2s still brings unnecessary waiting.

[0041] Table 2: Power quality P95% estimation error and system resource occupation

[0042]

[0043] In terms of P95% statistical accuracy, 0.5s has the lowest error due to larger sample size, but at the cost of higher CPU / bandwidth and worse FCR; 2s has the smallest sample size and amplifies the error; 1s balances between accuracy (≤ about 0.7%) and resource occupation, and cooperates with the FCR / E2E advantages in Table 1 to form an "engineering optimal compromise".

[0044] Table 3: Adaptive convergence of waiting window W

[0045]

[0046] From the test results, it can be seen that under the dual goals of "same frame alignment rate meets the standard + as low as possible end-to-end delay", 1s time frame shows more robust alignment capability and lower delay cost relative to 0.5s and 2s, and maintains sufficient statistical accuracy of quantile and moderate resource occupation, which is a more optimal unified time granularity selection in the embodiment.

[0047] The time synchronization module first initiates NTP initialization in the power-on stage, establishes a session with no less than two authoritative time sources, supports NTS authentication or symmetric key verification, sets minpoll = 10 s, maxpoll = 120 s and enables the multi-source optimization strategy, uses a sliding window and median filtering to suppress abnormal delay, and uses the PI servo to pull the local oscillator at a rate of ≤500 ppm in the slew mode when the deviation is less than 0.4 s, and performs a one-time step correction when the deviation is greater than the threshold; the GNSS time unit outputs 1 pulse per second and UTC message, the clock discipline unit drives the kernel phase-locked discipline with a pulse per second, and preferentially uses the pulse per second phase to reduce the local clock jitter to the millisecond level and feedback clockstats to the NTP to achieve external consistency; the time quality assessment unit calculates the NTP deviation, pulse per second deviation and two-way difference in real time, and performs consistency check according to the threshold of 0.1-0.5 s, and triggers event recording and alarm when the two-way difference exceeds the limit, and the lock priority is "pulse per second lock > multi-source NTP consistency > single-source NTP".

[0048] When the network is unavailable or upstream is unstable, enter the holdover state, estimate and compensate the drift based on the TCXO temperature drift model and historical frequency offset, give a growing uncertainty boundary and downgrade the time quality mark when exceeding the set time length of 30-120 min or the error upper bound approaches 1 s; to ensure the monotonicity of the application layer, the time synchronization module provides two strategies of step and smoothing for leap second processing, and by default performs 24h linear smoothing within the leap second window and writes the leap second identifier in the event record; the time frame management unit generates frame identifier TF(k) and frame start interrupt with the local clock aligned with UTC, broadcasts the NTP / GNSS state and time quality level with the frame to the acquisition, communication measurement and statistical threads, and all data records inserted into the upper layer are provided with UTC timestamp and TF(k) and "time quality bit"; temporarily relax the waiting window recommendation value before and after the NTP jitter rises or the link switches to maintain the frame alignment rate, and converge according to the hysteresis after recovery; the event recording unit persists the time source switching, two-way difference exceeding the limit, leap second processing, holdover entry / exit, authentication failure, etc., provides query and upload interfaces, and ensures the auditability and traceability of time synchronization.

[0049] Further, the communication quality monitoring module is used for measuring the communication quality index and calculating the value at the 95% position in a preset statistical time period as a statistical result. The preset statistical time period comprises: the communication quality monitoring module working under the driving of a 1s time frame TF(k) provided by the time synchronization module, which comprises a probe packet generation unit, a three-time stamp session management unit, a time stamp capturing and aligning unit, an RTT statistical unit, an integrity and security checking unit, an exception and congestion control unit and an interval summary sending unit. In the first 10-50 ms of each TF(k), the probe packet generation unit constructs a light load probe packet, and the fields at least include: frame number TF(k), terminal sending time T1 (UTC), sequence number, session ID and message authentication code (MAC), and is sent to the master station in the UDP / custom application layer handshake or "application layer sending + reply" mode according to the configuration selection; the master station records the time T2 when receiving, and returns T1, T2 and the master station sending time T3 in the response, the terminal records the time T4 when receiving the response, the three-time stamp session management unit completes matching and out-of-order discarding, and the time stamp capturing and aligning unit calls the kernel high-precision timer to read T1 / T4 and aligns with the UTC reference provided by the time synchronization module; the communication quality monitoring module calculates the round trip time RTT=T4 T1 (T3 T2) accordingly, and simultaneously defines the jitter by using the adjacent frame difference, and the packet loss rate is obtained according to the number of unresponded times / the number of sent probes in the preset statistical time period.

[0050] In order to suppress the occasional peak, the RTT statistical unit adopts the sliding window de-extreme value and median filtering for the RTT and jitter, all measured samples are written into the buffer with the frame number and quality bit and accept the waiting window; the integrity and security checking unit performs the MAC verification and time stamp playback detection on the response, and the failed sample is recorded as "integrity exception" and enters the exception counting; when the network switching, network address translation reconstruction or master station load rising causes the session failure, the session management retries ≤2 times with the exponential backoff (such as 200-1200 ms) and marks the probe missing in the current frame; the exception and congestion control unit moderately relaxes the probe packet size (48-256B) and sending advance according to the last preset statistical time period RTT_P95 / LOSS_P95 when sending the probe and the service on the common link, so as to ensure that it does not constitute a perceptible impact on the main service; at the end of each preset statistical time period (15 / 30 / 60 min), the interval summary sending unit calculates the RTT_P95, JITTER_P95, LOSS_P95 and integrity exception count respectively, and outputs them together with the power quality P95% in the same interval in the form of "frame binding", and generates the "frame linkage exception" label if the limits are exceeded simultaneously; the communication quality monitoring module depends on the time synchronization module, but in the short time, the three-time stamp can still offset the master station processing delay (T3 T2) and keep the measurement valid, automatically attach a "low time quality level" when the time quality degradation exceeds the threshold, provide the basis for the main station side interpretation and traceability.

[0051] Further, the data slicing and statistics module is used to form 10T0 segments in each time frame according to 10 fundamental periods, including: setting the analysis segment in each time frame to 10 fundamental periods (50Hz system: 10 periods = 0.2s; 60Hz system: 12 periods = 0.2s), which is based on the fact that the whole window length is consistent with the international power quality measurement standard, IEC 61000-4-30 (Class A) specifies 10 / 12 power frequency periods as the basic measurement time interval of voltage, frequency, etc. EC 61000-4-7 uses a 0.2s analysis window to group and calculate the harmonic / superharmonic spectrum, using the whole window can avoid the spectral leakage caused by non-integer window truncation, making the fundamental and each order harmonic amplitude, THD and other indicators comparable and reproducible between different devices; The 0.2s segment and the 1s time frame used in the embodiment have an integer division relationship, so that one can be selected from five candidate segments per second according to a random or round-robin strategy, which not only reduces the calculation and communication load, but also eliminates the periodic bias of fixed window selection. Using 10T0 (or 12T0 under 60Hz) as the segment length meets the multiple requirements of standard consistency, spectral accuracy, statistical robustness and system real-time, which can directly reproduce and support the data processing link of the embodiment.

[0052] The data slicing and statistics module runs under the 1s time frame TF(k) given by the time synchronization module, and is constructed around the lightweight strategy of "analyzing only one 10T0 segment per second", which includes a window scheduling unit, a preprocessing unit, a feature extraction unit and an interval statistics unit.

[0053] The window scheduling unit determines a target segment from a plurality of 10T0 windows (0.2s at 50Hz) available within TF(k) according to a random or round-robin strategy, the random strategy ensures reproducibility with device identity and TF(k) as seed, the round-robin strategy rotates in turn according to window sequence number to balance time domain coverage; subsequently, the preprocessing unit reads URMS, IRMS, P, Q, f or harmonic object values in the same frame as the segment from the segment data obtained from the electric energy meter interface, if the meter supports instantaneous sampling or harmonic component objects, perform DC removal, amplitude limiting and window function processing on the samples within the segment, if only per-cycle RMS / frequency sequence is available, approximate estimation is performed on the sub-sequence within the segment and the "estimation source" is marked. In the feature extraction unit, the data slicing and statistics module calculates the voltage effective value and frequency of the segment, if there is harmonic data, the total harmonic distortion THD or its subset is calculated according to the 1st to 25th / 50th harmonic, the frequency can be given by the meter or estimated by phase increment / zero-crossing interpolation; the segment result is written into the ring buffer together with TF(k), phase identification and quality bit, outliers are removed by median / quartile box rule and the original record is retained for traceability; the interval statistics unit takes 15min, 30min or 60min as the preset statistical time period, calculates the P95% quantile value of the voltage effective value, frequency and total harmonic distortion index of the segment formed every second in the same interval, and outputs the statistical vector organized by phase and sample count, rejection count.

[0054] Further, a cache and out-of-order correction module is used to group, rearrange out of order and compensate for missing values of the instrument values and communication quality indicators according to the same time frame before sending, and complete the framing within the waiting window, including: the cache and out-of-order correction module establishes a frame bucket in memory for each TF(k) and receives records from the electric energy meter interface and communication quality monitoring in a ring buffer manner, each record carries a coordinated universal time (UTC) timestamp, frame number, quality bit and signature digest; the cache and out-of-order correction module does not make assumptions about the arrival order, first sorts and removes duplicates according to the timestamp, then ensures the integrity of the frame structure with "missing placeholder", when the waiting window arrives, the fields that are still not complete are stored with a missing flag, while the original samples are retained for traceability.

[0055] In order to reduce the out-of-order and end-to-end delay caused by public network jitter, the cache and out-of-order correction module constructs an "adaptive waiting mechanism based on quantile number", such as Figure 3As shown, the 95% quantile value RTT_P95(t) of the round trip delay samples in the latest period of time is calculated and compared with a pair of upper and lower threshold intervals with hysteresis; when RTT_P95(t) is above the upper threshold for continuous M evaluation periods, the waiting window W(t) is increased by a preset amplification factor (not more than W_max) to improve the frame alignment rate; when RTT_P95(t) is below the lower threshold for continuous M evaluation periods, W(t) is decreased by a contraction factor (not less than W_min) to reduce the end-to-end delay; if RTT_P95(t) falls within the threshold band or does not meet the continuous M period condition, it remains unchanged. The left three-state state machine (amplification / hold / contraction) in the figure gives the decision logic, and the right two time curves show the stepwise adjustment of W(t) and the fluctuation of RTT_P95(t), and mark the upper and lower threshold bands and the W_min / W_max boundaries, which reflect the dynamic compromise and stable control between "frame completeness rate" and "delay" through "quantile driven + hysteresis + continuous period confirmation".

[0056] The specific process of the quantile-oriented adaptive waiting mechanism is as follows: the P95% of the round trip delay RTT of the last N frames (i.e. the last continuous N 1-second time frames, N being a configurable preset parameter) is taken as the main driving quantity, and multi-dimensional statistics and state machine control are introduced to continuously, limit, and auditable dynamically adjust the waiting window: the terminal initializes the first batch of collected samples (i.e. same frame identifier TF(k), terminal sending time T1, master station receiving time record T2, master station sending time T3, terminal receiving response time record T4, and RTT) in the cold start phase to the interval of 0.8-1.2s, and sets the upper and lower boundaries [0.2s, 2.0s] and the hysteresis thresholds low and high; then enters the sliding window evaluation period, the window length is 30-120, and the RTT samples in the period are subjected to de-extreme value and Hampel screening to eliminate single-point anomalies, and the 95% quantile value of the round trip delay (RTT) is calculated. The 95% quantile value of the round trip delay is compared with the last window and combined with the absolute threshold to judge whether the link is in the "dithering rising" or "dithering falling" state: when the 95% quantile value of the round trip delay is higher than the hysteresis threshold high for continuous M evaluation periods (M is 2-5), the waiting window is enlarged by a certain proportion, and the amplification factor is 1.0-1.8, to improve the frame alignment rate and absorb out-of-order, and the amplification action is limited by the upper limit to avoid excessive waiting; when the 95% quantile value of the round trip delay is lower than the hysteresis threshold low for continuous M evaluation periods, the waiting window is contracted by a certain proportion, and the contraction factor is 0.6-0.9, to reduce unnecessary waiting and compress the end-to-end propagation delay, and the contraction action is protected by the lower limit to avoid cutting too narrow and causing a large number of missing.

[0057] The calculation process of the 95% quantile value of the round-trip delay is as follows: taking 1 second time frame as the beat, gathering the successfully completed round-trip measurement samples of the terminal in the sliding window, each sample being obtained from the round-trip delay of four timestamp pairs, and being detected by message authentication and timestamp playback, being aligned with the current frame number and being qualified for time quality bit; then purifying the samples, removing individual points that are protocol exceptions, obviously out of range or identified as outliers by rules, and recording the unresponded probes as "missing" for recording the packet loss rate only and not entering the quantile statistics; if the number of valid samples is lower than the set lower limit, the collection is extended until the lower limit is reached or the period result is marked as low confidence. After purification, all valid round-trip delays are sorted from small to large, and the value near the 95% position of the overall sample number is located: when the position falls on a certain sample, the delay of the sample is directly taken; when it falls between two samples, a representative value between the two is obtained by proportional interpolation, which is taken as the 95% quantile value of the round-trip delay of the period. The whole process is independently run by link dimension and outputs the 95% quantile value of the round-trip delay and the quality metadata at the end of the interval, so as to reflect the end-to-end real delay of "the majority of requests" and avoid the misleading of a small number of abnormal peaks to the result.

[0058] Further, an alarm and uploading module is configured to output the value of the 95% position as a statistical result to local alarm, comprising: the alarm and uploading module taking the quantile statistics produced by the preset statistical time period as the trigger basis, and determining the communication delay P95%, the packet loss rate P95% and the integrity anomaly count at the end of each interval: when the delay P95% exceeds the preset threshold and the packet loss rate P95% exceeds the preset threshold, entering the first level alarm, immediately increasing the uploading frequency to 1s-5s and outputting the local sound / light relay alarm at the same time; when only one of them exceeds the threshold, entering the second level alarm, adjusting the uploading frequency to 5s-30s and recording the event. In order to avoid alarm jitter, the module uses the consistency criterion and hysteresis strategy of continuous H intervals, and only when the indicators continuously recover to the threshold and maintain for H intervals, the alarm is automatically downgraded until it is cleared. All uploading data packets uniformly contain time frame identifier, preset statistical time period number, out-of-limit statistics snapshot, corresponding threshold, event label (high delay, high packet loss, integrity anomaly, same frame linkage anomaly, etc.), link identifier, time quality bit and message authentication code; during link jitter, the alarm and uploading module is linked with the waiting window, the waiting window is preferentially relaxed under the first level alarm to improve the same frame alignment rate, and the uploading is tightened under the second level alarm to reduce link occupation; when the network is interrupted for a short time, local buffering and retransmission are enabled, and after the uploading is restored, the playback is performed in time sequence, ensuring that the master station side can be audited and the data is complete.

[0059] The above merely describes specific embodiments of the present application, but the protection scope of the present application is not limited thereto, and any modification or substitution within the technical range disclosed by the present application can be easily thought of by those skilled in the art, and should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

[0060] Finally, the above merely describes preferred embodiments of the present application, and is not used to limit the present application, and any modification, equivalent substitution, improvement, etc. within the spirit and principles of the present application should be included in the protection scope of the present application.

Claims

1. A collection terminal for communication quality monitoring, comprising an electricity meter interface module for reading meter values from an electricity meter and attaching a local time stamp, characterized in that, Also comprising: a time synchronization module for aligning the terminal clock to the coordinated universal time and generating a time frame with 1s granularity; a communication quality monitoring module for measuring the communication quality indicators and calculating the value at the 95% quantile within a preset statistical time period as the statistical result; a data slicing and statistics module for forming 10T0 segments within each time frame according to 10 fundamental wave periods, setting the analysis segment within each time frame to 10 fundamental wave periods; the 0.2s segment and the adopted 1s time frame are in a whole division relationship, so that 1 out of 5 candidate segments is selected per second according to a random or rotation strategy; the data slicing and statistics module operates under the driving of the 1s time frame given by the time synchronization module, and includes a window scheduling unit, a preprocessing unit, a feature extraction unit and an interval statistics unit; the window scheduling unit determines the target segment from multiple 10T0 windows within the 1s time frame according to a random or rotation strategy; the feature extraction unit calculates the voltage effective value and frequency of the segment; the interval statistics unit calculates the 95% quantile of the voltage effective value, frequency and total harmonic distortion indicator of the segment results formed per second within the same interval; a cache and out-of-order correction module for grouping, out-of-order rearranging and missing compensation of the instrument values and communication quality indicators according to the same time frame before uploading, and completing the framing within the waiting window; an alarm and uploading module for outputting the local alarm with the value at the 95% quantile as the statistical result; the cache and out-of-order correction module also constructs an adaptive waiting mechanism for quantile to reduce the out-of-order and end-to-end delay caused by public network jitter, takes the 95% quantile of the near N-frame round-trip time RTT as the main driving quantity, initializes the first batch of collected samples to the waiting window during the collection terminal cold start stage, sets the upper and lower boundaries and the hysteresis threshold low threshold and high threshold; then enters the sliding window evaluation period, calculates the 95% quantile of the round-trip time, and when the 95% quantile of the round-trip time of the continuous M evaluation periods is higher than the hysteresis threshold high threshold, the waiting window is enlarged in proportion; when the 95% quantile of the round-trip time of the continuous M evaluation periods is lower than the hysteresis threshold low threshold, the waiting window is contracted in proportion.

2. The collection terminal for communication quality monitoring according to claim 1, characterized by, The calculation process of the 95% quantile of the round-trip time is as follows: in the sliding window, the successful round-trip measurement samples of the terminal are converged, each sample is obtained by pairing four time stamps to get the round-trip time, and is aligned with the current frame number and has qualified time quality through message authentication and time stamp playback detection; then the samples are purified, the unresponded probes are marked as "missing" and only used to record the packet loss rate without entering the quantile statistics; if the number of valid samples is lower than the set lower limit, the collection is extended until the lower limit is reached or the period result is marked as low confidence; after the purification, all valid round-trip times are sorted from small to large, and the value near the 95% quantile of the overall sample number is located: when the position falls on a certain sample, the time delay of the sample is directly taken; when it falls between two samples, a representative value between the two is obtained by proportional interpolation, which is taken as the 95% quantile of the round-trip time of the period.

3. The collection terminal for communication quality monitoring according to claim 1, wherein The interface module of the electric energy meter includes physical access, session management, object read-write, and cache and verification units. It is connected with the meter through RS-485, Ethernet or optical-electric port, reads URMS, IRMS, active power, reactive power, frequency and meter clock objects by using DLMS / COSEM or equivalent protocol, adds UTC timestamp and TF(k) for each reading, and writes into a non-locking ring buffer, and sets explicit quality marks for missing, timeout and verification failure records.

4. The collection terminal for communication quality monitoring according to claim 3, wherein The time synchronization module includes NTP time synchronization unit and GNSS time synchronization unit. GNSS provides one pulse per second for phase locking and frequency calibration of the local clock to generate TF(k). When the network is unstable, it enters a holding state and compensates according to the historical frequency deviation. During the leap second, a smoothing strategy is used to make the application layer time monotonically increasing.

5. The collection terminal for communication quality monitoring according to claim 1, wherein The communication quality monitoring module completes the three-time stamp session through UDP, custom application layer handshake or "service upload + receipt" mode, calculates the round-trip delay, defines the jitter by adjacent frame difference, defines the packet loss rate by "unanswered number / sent probe number", and performs message authentication code and playback detection verification on the sample by using sliding window de-extreme value and median filtering.

6. The collection terminal for communication quality monitoring according to claim 1, wherein The data slicing and statistical module selects the target segment 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 in the segment, and extracts the voltage effective value, frequency and total harmonic distortion. In the preset statistical time period, the 95% quantile value of each index and the sample and rejection count are output.

7. The collection terminal for communication quality monitoring according to claim 1, wherein The communication quality monitoring module performs message authentication code verification and timestamp playback detection on the response. When network switching or network address translation reconstruction causes session failure, it uses exponential backoff retry not more than 2 times, and when the time synchronization quality degradation exceeds the threshold, it adds a low time quality level mark to the frame record.

8. A collection terminal for communication quality monitoring according to any of claims 2-7, characterized in that, The measurement records output by the electric energy meter interface module are packaged in a unified structure, including at least device identification, object identification, measured value, unit, quality bit, UTC timestamp, TF(k) and signature digest. The upload data packet of the alarm and upload module includes at least time frame identification, preset statistical time period number, out-of-limit statistical quantity snapshot, corresponding threshold, event label, link identification and time quality bit, and after network interruption is restored, it is played back in time sequence.

Images