A method and system for master-slave clock synchronization in a distribution network based on cellular networks

Abstract

This invention discloses a master-slave clock synchronization method and system for power distribution networks based on cellular networks. Using 4G / 5G cellular networks as the PTP synchronization message transmission carrier, it eliminates the need for satellite signal reception, fundamentally overcoming the inherent limitations of satellite signals being unable to penetrate in obstructed scenarios such as underground tunnels and indoor ring main units, thus avoiding time synchronization interruptions or failures. Simultaneously, cellular networks can extend signal coverage through indoor distribution systems and micro base stations, offering stronger scenario adaptability compared to satellite time synchronization. This invention eliminates the need for laying dedicated fiber optic lines, directly reusing the cellular infrastructure already deployed on a large scale by operators. Communication distances range from several kilometers to tens of kilometers, meeting the long-distance synchronization needs of ring main units in cross-regional and county-level power distribution networks, significantly reducing on-site construction difficulty and renovation costs.

Description

Technical Field

[0001] This invention relates to the field of power system time synchronization technology, and in particular to a master-slave clock synchronization method and system for distribution networks based on cellular networks. Background Technology

[0002] Synchronization detection in power systems is a core foundational technology for ensuring the safe and stable operation of urban distribution networks and rural power grids in counties. Distribution networks have numerous ring main units distributed in areas such as streets, underground cable tunnels, and building mezzanines. Before performing ring main unit closure or cross-regional line connection operations, it is essential to synchronously collect the voltage amplitude, frequency, and phase parameters of the lines on both the main grid side and the load side. Only when the synchronous sampling error of electrical quantities on both sides is controlled within the microsecond range can the phase detection results be truly reliable. If the clock synchronization deviation between the two sides' acquisition units is too large, it will cause phase measurement distortion, misjudgment of synchronization conditions, and generate huge circulating currents and short-circuit faults, which in severe cases can lead to the disconnection of the regional power grid.

[0003] Synchronization testing of power systems is a core foundational technology for ensuring the safe and stable operation of urban distribution networks and rural power grids in counties. A large number of ring main units (RNBs) in distribution networks are distributed in areas such as streets, underground cable tunnels, building mezzanines, and distribution rooms. Before performing RNB closure or cross-regional line connection operations, it is essential to synchronously collect the voltage amplitude, frequency, and phase parameters of the lines on both the main grid side and the load side. The accuracy of time synchronization directly determines the reliability of synchronization testing. During distribution network synchronization testing, the synchronous sampling error of electrical quantities on both sides must be controlled within the microsecond range to ensure accurate and reliable phase difference measurements. If the clock synchronization deviation between the two acquisition units exceeds the microsecond tolerance range, it will directly lead to phase measurement distortion, misjudgment of synchronization conditions, and the generation of huge circulating currents and short-circuit surge currents at the moment of RNB closure. This can cause damage to switching equipment or, in severe cases, lead to the disconnection of the regional power grid, resulting in large-scale power outages. Therefore, achieving high-precision time synchronization between the RNB and the main substation is a prerequisite for ensuring the safe and stable operation of the distribution network.

[0004] Currently, the mainstream time synchronization solutions for remote ring main units in the industry are divided into three categories, all of which have technical shortcomings that cannot simultaneously meet the requirements of long-distance operation, no satellite dependence, microsecond-level synchronization, and suppression of cellular asymmetric time delay errors:

[0005] The first category is the GPS / BeiDou satellite timing solution. This solution relies on open outdoor environments to receive satellite radio frequency signals. However, satellite signals are severely attenuated in underground cable tunnels, indoor ring main units, and areas blocked by high-rise buildings, resulting in frequent satellite search failures and timing interruptions. Adding an outdoor signal amplification antenna would significantly increase on-site construction costs, and it is completely unfeasible in enclosed tunnel scenarios, failing to cover the simultaneous monitoring needs of the entire power distribution network.

[0006] The second category includes short-range wireless PTP time synchronization solutions such as LoRa and WIFI. The theoretical effective communication distance of this type of communication medium is only tens to hundreds of meters, which cannot meet the requirements for synchronous detection of remote ring network cabinets across streets, districts, and distances of several kilometers. At the same time, LoRa spread spectrum modulation naturally has a large time delay jitter, and the traditional PTP protocol assumes that the uplink and downlink transmission delays are completely symmetrical. Faced with the dynamic time delay difference caused by wireless multipath fading and RF temperature drift, the synchronization offset can reach tens of microseconds, which cannot achieve the microsecond-level accuracy target for power synchronization detection.

[0007] The third category is the traditional 4G / 5G cellular standard PTP time synchronization scheme. Existing schemes simply reuse the IEEE 1588 standard clock offset calculation formula, forcibly assuming equal uplink and downlink transmission delays, without establishing compensation models for error sources specific to cellular networks. Factors such as 4G uplink random access backoff, uneven uplink and downlink scheduling time slots in 5G, dynamic fluctuations in cell load, and instantaneous delay jumps during cell handover continuously introduce time-varying asymmetric delay differences. Traditional schemes cannot distinguish between inherent hardware delays, static cellular scheduling delays, and dynamic channel disturbance delays. Synchronization accuracy drifts sharply with increasing network load, with long-term synchronization jitter exceeding 10μs, directly causing distortion in synchronization phase detection and posing significant network security risks.

[0008] In summary, existing time synchronization and wireless time synchronization technologies cannot simultaneously solve the four major engineering pain points: satellite signal blockage failure, limited short-range communication coverage, dynamic asymmetric delay error in cellular networks, and excessive long-term synchronization jitter. There is an urgent need for a high-precision wireless synchronization detection method that is compatible with 4G / 5G long-distance cellular communication and can suppress all types of asymmetric delay errors in a layered manner. Summary of the Invention

[0009] The purpose of this invention is to solve the technical problem that existing power distribution network time synchronization schemes cannot simultaneously address long-distance coverage, lack of satellite signal dependence, and asymmetric delay suppression of cellular networks, resulting in synchronization accuracy failing to meet the microsecond-level requirements of wireless synchronous detection.

[0010] To solve the above technical problems, the technical solution of the present invention is as follows:

[0011] A master-slave clock synchronization method for a distribution network based on a cellular network is used to synchronize the master clock unit of the distribution network master station and the slave unit of the remote ring network cabinet through a 4G / 5G cellular communication network; characterized in that it includes:

[0012] S1. Obtain the operating temperature of the master clock unit and slave unit, and obtain the current hardware latency compensation amount of the master clock unit and slave unit;

[0013] S2. Construct a sliding timestamp buffer window, collect timestamp information corresponding to multiple sets of master-slave clock message interactions based on the PTP protocol, and construct a multi-period observation sample set;

[0014] S3. Based on a multi-period observation sample set, an asymmetric delay PTP observation equation set is introduced. The static delay difference of the cellular network is solved by least squares batch identification, and the global static delay compensation is obtained by superimposing it with the basic compensation amount from the unit.

[0015] S4. Obtain the clock deviation residual of each round of PTP interaction as the observation value, update it using an adaptive extended Kalman filter, obtain the dynamic delay disturbance at the current moment, and superimpose the dynamic delay disturbance with the global static delay compensation to obtain the real-time complete uplink and downlink total delay difference.

[0016] S5. The clock offset estimate is corrected using the real-time complete uplink and downlink total delay difference, and the slave unit clock is corrected using clock servo control.

[0017] Furthermore, the specific method for obtaining the basic hardware latency compensation amount in step S1 includes:

[0018] Collect the 4G / 5G RF transmit path delay Ttx and RF receive path delay Trx of the master clock unit and slave unit within the temperature range of -40℃ to 70℃, and construct a temperature-RF delay lookup table;

[0019] Obtain the real-time operating temperature of the master clock unit and slave unit, look up the table to obtain the RF transmission path delay and RF reception path delay of the master clock unit and slave unit at the current temperature, and calculate the basic hardware delay compensation amount for each unit: Basic hardware delay compensation amount = RF transmission path delay of the corresponding unit - RF reception path delay of the corresponding unit.

[0020] Furthermore, the construction of the multi-period observation sample set is specifically as follows: 30 to 100 complete quaternary timestamps are continuously collected as the multi-period observation sample set; the quaternary timestamps include the master clock unit periodically sending Sync messages, FollowUp messages, MAC hardware capture message sending timestamp t1, slave unit MAC hardware capture Sync message receiving timestamp t2, slave unit sending DelayReq messages local sending timestamp t3, and master clock MAC hardware capture DelayResp message receiving timestamp t4.

[0021] Furthermore, it also includes removing abnormal samples from the constructed multi-period observation sample set: the uplink round-trip delay of each DelayReq message group is counted, the uplink round-trip delay is compared with the preset uplink delay threshold, and if it exceeds the threshold, it is determined to be an abnormal sample and removed from the multi-period observation sample set.

[0022] Furthermore, the state vector of the adaptive extended Kalman filter AEKF in step S4 is:

[0023]

[0024] in, For clock offset, For clock frequency offset, The static delay difference of the cellular network is obtained by solving step S3. For dynamic time delay perturbations;

[0025] Establish state and observation equations adapted to the drift characteristics of the power crystal oscillator and the evolution law of cellular time delay, and set the PTP synchronization iteration period to be [value missing]. Construct a 4×4 state transition matrix:

[0026]

[0027] The state prediction equation is

[0028]

[0029] in, The prior state estimate at time k; For process noise, , It follows a normal distribution; The process noise covariance matrix;

[0030] Covariance prediction is performed based on the state transition matrix. The covariance prediction formula is as follows:

[0031]

[0032] in, The covariance matrix is ​​estimated for the prior state. This is the state estimation filtering covariance matrix after the previous iteration, used to characterize the uncertainty of the estimation error of each component of the four-dimensional state vector; Let F be the transpose of the state transition matrix F;

[0033] The observation equation is

[0034]

[0035] in, PTP clock bias residuals; observation matrix , To observe the noise, To observe the noise covariance matrix;

[0036] Calculate the Kalman gain based on the observation matrix and the observation noise covariance matrix:

[0037]

[0038] in, The Kalman gain matrix; Let H be the transpose of the observation matrix;

[0039] The state vector is corrected using Kalman gain, and the updated state estimation filter covariance matrix is ​​obtained:

[0040]

[0041] in, It is a 4×4 identity matrix;

[0042] The state update is completed by correcting the prior state variables using observed residuals.

[0043]

[0044] From the output state vector Extracting dynamic delay perturbation components from cellular networks .

[0045] Furthermore, the process noise covariance matrix Q is based on the mean of round-trip delay jitter. To make adaptive adjustments, specifically:

[0046]

[0047] in, is the round-trip delay of the i-th PTP round; N is the number of quaternary timestamp sample groups collected within the sliding timestamp buffer window; The historical average round-trip latency within the sliding timestamp cache window; if Greater than the preset threshold Then the process noise matrix is ​​adjusted to Amplify the process noise matrix; otherwise, adjust the process noise matrix to... This reduces the noise matrix during the process.

[0048] Furthermore, it also includes cell handover processing: real-time monitoring of cell handover signaling reported from the 4G / 5G communication module of the unit; when a cell handover action is detected, clearing the sliding timestamp buffer window and resetting the state estimation filter covariance matrix.

[0049] Furthermore, when using 5G cellular networks for transmission, it also includes time slot scheduling pre-compensation for TDD standards, specifically:

[0050] Obtain the time slot allocation parameters issued by the 5G cellular network to determine the number of downlink time slots within a single scheduling cycle. Uplink time slots and special time slot numbers ;

[0051] Setting single-slot reference transmission delay and special time slot conversion delay ;

[0052] Calculate downlink average scheduling delay Uplink average scheduling delay ;

[0053] Calculate the inherent time delay difference of time slot scheduling ;

[0054] Will The global static latency compensation amount superimposed on the 5G cellular network.

[0055] Furthermore, the specific formula for correcting the clock offset estimate using the real-time complete uplink and downlink total delay difference in step S5 is as follows:

[0056]

[0057] in, Estimate the actual clock offset; The master clock unit periodically sends Sync messages, FollowUp messages, and MAC hardware capture messages with timestamps. To capture the timestamp of the Sync message reception from the unit MAC hardware; For the local transmission timestamp of the DelayReq message sent from the unit; The master clock MAC hardware captures the DelayResp message reception timestamp; This represents the real-time, complete total uplink and downlink latency difference.

[0058] The clock servo control is a feedforward PI clock servo; the real clock offset estimate is input into the PI controller, and the real-time complete uplink and downlink total delay difference is directly injected into the servo output as the feedforward correction amount. The PI controller outputs the clock adjustment amount, and the servo output adjustment amount acts on the TCXO temperature control crystal oscillator of the slave unit.

[0059] The present invention also provides a master-slave clock synchronization system for a distribution network based on a cellular network, for executing the aforementioned master-slave clock synchronization method for a distribution network based on a cellular network, characterized in that it includes: a PTP master clock unit of the distribution network master station, a 4G / 5G cellular communication network, and at least one synchronization detection slave unit of a remote ring network cabinet;

[0060] The master clock unit includes an industrial-grade 4G / 5G communication module, a MAC layer hardware timestamp capture module, a TCXO temperature-controlled crystal oscillator clock source, an RF temperature acquisition sensor, and local non-volatile memory. The master clock unit captures the timestamp of the PTP message through its MAC layer hardware timestamp capture module, and corrects the hardware delay deviation of its own timestamp through the temperature-delay lookup table in the RF temperature acquisition sensor and local non-volatile memory.

[0061] The unit includes an industrial-grade 4G / 5G communication module, a MAC layer hardware timestamp capture module, a TCXO temperature-controlled crystal oscillator clock source, an RF temperature acquisition sensor, local non-volatile memory, an AEKF algorithm processing unit, a cellular-specific correction logic processing unit, and a feedforward PI clock servo unit.

[0062] Temperature data acquired from the unit's RF temperature acquisition sensor is used to obtain the hardware delay base compensation amount through the temperature-delay lookup table in the local non-volatile memory. The MAC layer hardware timestamp capture module captures the timestamp of the PTP message and inputs it to the AEKF algorithm operation unit and the cellular dedicated correction logic processing unit. The outputs of the AEKF algorithm operation unit and the cellular dedicated correction logic processing unit are superimposed and input to the feedforward PI clock servo unit. The output adjustment amount of the feedforward PI clock servo unit is applied to the TCXO temperature control crystal oscillator clock source.

[0063] The master clock unit and the slave unit exchange PTP messages through the 4G / 5G cellular communication network.

[0064] The present invention has the following beneficial effects:

[0065] This invention uses 4G / 5G cellular networks as the PTP synchronization message transmission carrier, eliminating the need for satellite signal reception. This fundamentally overcomes the inherent limitation of satellite signals being unable to penetrate obstructed environments such as underground tunnels and indoor ring main units, preventing timing interruptions or failures. Simultaneously, cellular networks can extend signal coverage through indoor distribution systems and micro base stations, offering greater adaptability to different scenarios compared to satellite timing. This invention eliminates the need for dedicated fiber optic lines, directly reusing the large-scale cellular infrastructure already deployed by operators. Communication distances range from several kilometers to tens of kilometers, meeting the long-distance synchronization needs of cross-regional and county-level distribution network ring main units, significantly reducing on-site construction difficulty and upgrade costs.

[0066] This invention calculates the basic RF hardware delay compensation for the master clock unit and slave unit, as well as the static delay difference in the cellular network. It then utilizes an adaptive extended Kalman filter to track dynamic delay disturbances in real time, employing a three-layer joint compensation architecture to obtain the real-time complete uplink and downlink total delay difference. This total delay difference is then used to correct the estimated clock offset, and a feedforward PI clock servo is used to correct the slave unit clock. The total delay difference is directly injected into the servo output as a feedforward correction, decoupling the delay correction from the feedback loop and significantly reducing the response lag to network time-varying disturbances. This achieves complete suppression of asymmetric delay across all scenarios, controlling the static synchronization error to within 0.4μs, and reducing dynamic jitter to no more than 2μs under high network load. Compared to the tens of microseconds-level offset of existing 4G / 5G PTP solutions, this represents an accuracy improvement of more than an order of magnitude. Synchronization jitter is reduced by more than 60% under high load and drastic fluctuation scenarios, meeting the requirements for synchronous detection in power distribution networks.

[0067] This invention also eliminates the latency jumps and long-term drifts unique to cellular mobile communication by removing abnormal samples caused by uplink random access backoff, pre-calculating the inherent latency difference of 5G TDD time slot scheduling, and clearing the buffer window and resetting the filter covariance matrix during cell handover, thus ensuring the long-term stability of synchronization accuracy in the complex time-varying environment of cellular networks. Attached Figure Description

[0068] Figure 1 This is a flowchart of a master-slave clock synchronization method for a distribution network based on a cellular network, according to the present invention.

[0069] Figure 2 This is a structural diagram of a master-slave clock synchronization system for a power distribution network based on a cellular network, according to the present invention. Detailed Implementation

[0070] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0071] Example

[0072] like Figure 1 As shown in the figure, this embodiment presents a master-slave clock synchronization method for a distribution network based on a cellular network, used to synchronize the master clock unit of the distribution network master station and the slave unit of the remote ring network cabinet through a 4G / 5G cellular communication network; it includes the following steps.

[0073] S1. Obtain the operating temperature of the master clock unit and slave unit, and obtain the current hardware latency compensation amount of the master clock unit and slave unit.

[0074] Specific methods for obtaining the basic hardware latency compensation amount include:

[0075] Collect the 4G / 5G RF transmission path latency of the master clock unit and slave unit within the temperature range of -40℃ to 70℃. and RF receiving path delay A temperature-RF delay lookup table is constructed and stored in the corresponding local non-volatile memory Flash.

[0076] Obtain the real-time operating temperature of the master clock unit and slave units. Based on the real-time operating temperature, retrieve the RF transmit path delay and RF receive path delay of the master clock unit and slave unit corresponding to that real-time temperature from the corresponding temperature-RF delay two-dimensional lookup table, and calculate the current hardware delay base compensation amount for each unit: .

[0077] If the real-time temperature is between two adjacent calibration temperature points, retrieve the two calibration temperature points adjacent to the real-time temperature and their corresponding RF transmit path delay and RF receive path delay from the temperature-RF delay two-dimensional lookup table; then perform linear interpolation on the RF transmit path delay and RF receive path delay to obtain the result at the current temperature. and ; Calculate the basic hardware latency compensation for each unit.

[0078] The hardware latency baseline compensation of the master clock unit is used to correct the inherent hardware latency deviation of its own transmit and receive timestamps.

[0079] S2. Construct a sliding timestamp buffer window, collect timestamp information corresponding to multiple sets of master-slave clock message interactions based on the PTP protocol, and construct a multi-period observation sample set.

[0080] Construct a sliding timestamp buffer window and continuously collect 30 to 100 complete quaternary timestamps. This forms a multi-period observation sample set; the window value range is selected based on the average load fluctuation characteristics of 4G / 5G cellular networks, with the upper limit of 100 groups when the network is under high load and the lower limit of 30 groups when the network is under low load.

[0081] The quaternary timestamp includes the timestamps for sending Sync messages, FollowUp messages, and MAC hardware capture messages periodically issued by the master clock unit. Capture Sync message receive timestamp from unit MAC hardware Local transmission timestamp of the DelayReq message sent from the unit Master clock MAC hardware capture of DelayResp message receive timestamp .

[0082] In the process of constructing a multi-period observation sample set, abnormal samples are also removed from the constructed multi-period observation sample set: the uplink round-trip delay of each DelayReq message is calculated, and the uplink round-trip delay is compared with the preset uplink delay threshold. If it exceeds the threshold, it is determined to be an abnormal sample and removed from the multi-period observation sample set.

[0083] S3. Based on a multi-period observation sample set, an asymmetric delay PTP observation equation set is introduced. The static delay difference of the cellular network is solved by least squares batch identification, and the global static delay compensation is obtained by superimposing it with the basic compensation amount of the unit.

[0084] Asymmetric time delay PTP observation equations:

[0085]

[0086] in, The actual clock offset of the slave cell relative to the master clock; Total downlink transmission latency for 4G / 5G; This refers to the total uplink transmission latency for 4G / 5G.

[0087] The total uplink and downlink transmission delay is divided into static components and dynamic disturbance components:

[0088]

[0089]

[0090] in, The static latency component caused by cellular base station scheduling and buffering; , This refers to the dynamic time delay disturbance component caused by multipath and changes in cell load.

[0091] Define the uplink and downlink static delay difference as follows:

[0092]

[0093] Construct a system of least squares linear equations for all groups of samples within the multi-period observation sample, in order to As one of the parameters to be estimated (while eliminating the actual clock offset of the cell relative to the master clock), (Impact), batch solution to obtain the estimated value of static delay difference of cellular network. .

[0094] Estimate the static delay difference of cellular networks The hardware delay compensation amount of the slave unit output in step S1 By superimposing these values, the global static delay compensation amount is obtained.

[0095] S4. Obtain the clock deviation residual of each PTP interaction as the observation value, update it using the adaptive extended Kalman filter (AEKF), obtain the dynamic delay disturbance at the current moment, and superimpose the dynamic delay disturbance with the global static delay compensation to obtain the real-time complete uplink and downlink total delay difference.

[0096] In step S4, the state vector of the adaptive extended Kalman filter (AEKF) is:

[0097]

[0098] in, For clock offset, For clock frequency offset, The static delay difference of the cellular network is obtained in step S3. This is a dynamic time delay disturbance.

[0099] State transition matrix for:

[0100]

[0101] in, This indicates the PTP synchronization period, which is the time interval between two Sync messages sent by the master clock.

[0102] Set the PTP synchronization iteration period to , This refers to the time interval from when a unit completes a full round of PTP four-timestamp acquisition and performs one filtering iteration operation. In an ideal packet-loss-free link, this interval is equal to the interval between the transmission of two consecutive Sync messages from the master clock.

[0103] The state prediction equation is

[0104]

[0105] in, The prior state estimate at time k; For process noise, , It follows a normal distribution; The process noise covariance matrix;

[0106] Covariance prediction is performed based on the state transition matrix. The covariance prediction formula is as follows:

[0107]

[0108] in, The covariance matrix is ​​estimated for the prior state. This is the state estimation filtering covariance matrix after the previous iteration, used to characterize the uncertainty of the estimation error of each component of the four-dimensional state vector; Let F be the transpose of the state transition matrix F;

[0109] The observation equation is

[0110]

[0111]

[0112] in, This refers to the PTP clock offset residual. H represents observation noise; H is the observation matrix. ; To observe the noise covariance matrix.

[0113] Calculate the Kalman gain based on the observation matrix and the observation noise covariance matrix:

[0114]

[0115] in, The Kalman gain matrix; Let H be the transpose of the observation matrix;

[0116] The state vector is corrected using Kalman gain, and the updated state estimation filter covariance matrix is ​​obtained:

[0117]

[0118] in, It is a 4×4 identity matrix;

[0119] The state update is completed by correcting the prior state variables using observed residuals.

[0120]

[0121] From the output state vector Extracting dynamic delay perturbation components from cellular networks .

[0122] The process noise covariance matrix Q is based on the mean of round-trip delay jitter. To make adaptive adjustments, specifically:

[0123] Obtain the actual round-trip time (RTT) for each PTP interaction within the window, and combine it with the long-term steady-state historical average RTT within the sliding timestamp cache window. Calculate the absolute vibration deviation of a single wheel The arithmetic mean of N absolute jitter deviation samples within the sliding timestamp buffer window is used to obtain the mean delay jitter, which characterizes the current cellular network disturbance intensity.

[0124]

[0125] in, is the round-trip delay of the i-th PTP round; N is the number of quaternary timestamp sample groups collected within the sliding timestamp buffer window; The historical average round-trip latency within the sliding timestamp cache window.

[0126] Average latency jitter With preset threshold Comparison: If This indicates that the current cellular network is experiencing strong disturbances, so the process noise covariance matrix Q is adjusted to... To improve the filter tracking speed; if This indicates that the current cellular network disturbance is relatively weak, so Q is adjusted to... This is done to suppress random noise in the timestamps and ensure stable synchronous output. That is:

[0127]

[0128] in, and For a pre-defined diagonal matrix, The values ​​of each diagonal element are greater than The corresponding elements correspond to the process noise covariance configurations for high-disturbance and low-disturbance scenarios, respectively.

[0129] The above adaptive adjustment is performed after each round of PTP message exchange is completed.

[0130] After each round of PTP message exchange, a complete filtering iteration is performed, outputting the dynamic delay perturbation difference. The dynamic disturbance difference is superimposed with the global static delay compensation to obtain the real-time complete uplink and downlink total delay difference:

[0131]

[0132] in, The basic compensation amount for the hardware delay of the slave unit is output in step S1. To identify the static delay difference of the cellular network in step S3, The dynamic time delay disturbance is estimated in real time by AEKF.

[0133] Furthermore, this method incorporates dedicated correction logic for cellular network-specific scenarios. When cell handover signaling is detected from the 4G / 5G communication module within the cell, the sliding timestamp buffer window is immediately cleared, and the state estimation covariance matrix of the adaptive extended Kalman filter (AEKF) is reset. This allows the AEKF to quickly reconverge under the new link, eliminating synchronization disturbances caused by the instantaneous delay jump during cell handover.

[0134] In this invention, the master clock unit can transmit and interact with the slave unit's 4G communication module via a 4G cellular network through a 4G communication module; it can also transmit and interact with the slave unit's 5G communication module via a 5G cellular network through a 5G communication module.

[0135] When using 5G cellular networks for transmission, it also includes time slot scheduling pre-compensation for TDD standards.

[0136] Specifically, the time slot allocation parameters issued by the 5G cellular network are obtained to determine the number of downlink time slots within a single scheduling cycle. Uplink time slots and special time slot numbers ;

[0137] Setting single-slot reference transmission delay and special time slot conversion delay ;

[0138] Calculate downlink average scheduling delay Uplink average scheduling delay ;

[0139] Calculate the inherent time delay difference of time slot scheduling ;

[0140] Will The global static latency compensation is superimposed on the 5G cellular network to offset the fixed asymmetric latency error caused by the uneven allocation of uplink and downlink time slot resources in 5G TDD.

[0141] In this embodiment, the cellular network-specific correction process comprises three layers: removing abnormal latency samples caused by uplink random access backoff from the multi-period observation sample set; pre-calculating and compensating for the inherent latency difference introduced by time slot scheduling under the 5G TDD standard; and clearing the sliding timestamp buffer window and resetting the state estimation covariance matrix when cell handover is detected. These three layers of correction, from the perspectives of data sample purification, physical layer time slot deviation pre-compensation, and channel abrupt state reset, respectively, specifically eliminate the impact of cellular network-specific error sources on synchronization accuracy.

[0142] S5. The clock offset estimate is corrected using the real-time complete uplink and downlink total delay difference, and the slave unit clock is corrected using clock servo control.

[0143] The specific formula for correcting the clock offset estimate using the real-time complete uplink and downlink total delay difference in step S5 is as follows:

[0144]

[0145] in, Estimate the actual clock offset; The master clock unit periodically sends Sync messages, FollowUp messages, and MAC hardware capture messages with timestamps. To capture the timestamp of the Sync message reception from the unit MAC hardware; For the local transmission timestamp of the DelayReq message sent from the unit; The master clock MAC hardware captures the DelayResp message reception timestamp; This represents the real-time, complete total uplink and downlink latency difference.

[0146] The clock servo control is a feedforward PI clock servo; the real clock offset estimate is input into the PI controller, and the real-time complete uplink and downlink total delay difference is directly injected into the servo output as the feedforward correction amount. The PI controller outputs the clock adjustment amount, and the servo output adjustment amount acts on the TCXO temperature-controlled crystal oscillator of the slave unit to correct the local system clock of the slave unit and realize the microsecond-level synchronization between the slave unit clock and the master clock.

[0147] In this embodiment, after the local TCXO crystal oscillator calibration is completed in step S6, the slave unit has obtained a local clock synchronized with the master clock. Based on this, the following synchronization detection and safety latching operations are performed.

[0148] The unit uses the corrected local synchronization clock as the sampling reference to synchronously trigger the ADC analog-to-digital converters of its two internal voltage acquisition channels, which synchronously sample the outputs of the voltage transformers on the main grid side and the load side, respectively. Since the local synchronization clock has eliminated the deviation introduced by the asymmetric delay of the cellular network, the synchronization error of the sampling time of the electrical quantities on both sides can be stably controlled within 2μs.

[0149] The frequency, amplitude, and phase parameters of the voltage on both sides of the unit are extracted, and the closing conditions are determined according to the power system synchronization grid connection standard criteria. At the same time, the local clock synchronization error is monitored in real time from the unit, and the current synchronization error is compared with a preset safety threshold: if the synchronization error is within the allowable range, a closing permission signal is output; if the synchronization error exceeds the preset safety threshold, a closing blocking signal is directly output, prohibiting the grid connection closing operation, so as to avoid serious faults such as phase detection distortion, excessive closing inrush current, and regional grid disconnection caused by synchronization inaccuracy.

[0150] like Figure 2 As shown, this embodiment also provides a cellular network-based distribution network master-slave clock synchronization system for executing a cellular network-based distribution network master-slave clock synchronization method, including: a PTP master clock unit of the distribution network master station, a 4G / 5G cellular communication network, and at least one synchronization detection slave unit of a remote ring network cabinet.

[0151] The master clock unit includes an industrial-grade 4G / 5G communication module, a MAC layer hardware timestamp capture module, a TCXO temperature-controlled crystal oscillator clock source, an RF temperature acquisition sensor, and local non-volatile memory.

[0152] The master clock unit captures the timestamp of the PTP message through its MAC layer hardware timestamp capture module, and corrects the hardware delay deviation of its own timestamp through the radio frequency temperature acquisition sensor and the temperature-delay lookup table in the local non-volatile memory.

[0153] The unit includes an industrial-grade 4G / 5G communication module, a MAC layer hardware timestamp capture module, a TCXO temperature-controlled crystal oscillator clock source, an RF temperature acquisition sensor, a local non-volatile memory, an AEKF algorithm processing unit, a cellular-specific correction logic processing unit, and a feedforward PI clock servo unit.

[0154] Temperature data acquired from the unit's RF temperature acquisition sensor is used to obtain the hardware latency compensation amount through a temperature-latency lookup table in the local non-volatile memory. The MAC layer hardware timestamp capture module captures the timestamp of the PTP message and inputs it to the AEKF algorithm operation unit and the cellular-specific correction logic processing unit. The outputs of the AEKF algorithm operation unit and the cellular-specific correction logic processing unit are superimposed and input to the feedforward PI clock servo unit. The output adjustment amount of the feedforward PI clock servo unit is applied to the TCXO temperature control crystal oscillator clock source.

[0155] The master clock unit and the slave unit exchange PTP messages through the 4G / 5G cellular communication network.

[0156] This invention employs a 4G / 5G communication and indoor ring network cabinet experimental environment, setting up four control schemes to conduct synchronization accuracy comparison tests. The test load covers three levels: 0% (no load), 40% (medium load), and 80% (high load). Each set of operating conditions is continuously tested for 24 hours, and two indicators, static synchronization error and dynamic synchronization jitter, are collected. The test results are shown in the table below:

[0157]

[0158] Experimental conclusions: The accuracy improvement of a single compensation module is limited; the synchronization accuracy of the layered infrastructure is significantly improved, but the jitter under high load conditions still exceeds 4μs; after superimposing a three-layer cellular dedicated correction module and a feedforward PI servo, the no-load static error is as low as 0.36μs, and the dynamic jitter under 80% high load is only 1.8μs. The synchronization accuracy produces unexpected technical effects that exceed the sum of the individual effects of each module, and has outstanding substantial characteristics.

[0159] Control group 2 only compensated using a hardware temperature LUT, resulting in limited improvement in synchronization accuracy. Control group 3 employed a hierarchical joint compensation architecture combining temperature LUT, least squares, and AEKF (excluding cellular-specific correction), achieving a significant improvement in synchronization accuracy compared to control group 2. However, when the network load reached 80%, the synchronization jitter still exceeded 4μs, failing to stably meet the synchronization detection requirements of the distribution network. In contrast, when tested using the method of this invention, the no-load static error was as low as 0.36μs, and the dynamic jitter at 80% high load was only 1.8μs. The comparison clearly shows that the method of this invention significantly improves synchronization accuracy compared to control group 3 under high load jitter conditions.

[0160] This invention is used in scenarios involving simultaneous detection of ring main units in urban indoor environments:

[0161] The power distribution network master station room is equipped with a PTP master clock unit, and multiple ring main units in the room are equipped with industrial 5G power private network modules for simultaneous detection. The ring main units are located in enclosed spaces such as underground cable tunnels or building mezzanines, where there is no satellite signal coverage and GPS timing cannot be used.

[0162] After power-on, the device automatically performs real-time RF temperature compensation and continuously interacts with standard PTP messages to collect quaternary hardware timestamps. The sliding timestamp buffer window dynamically and adaptively adjusts its size from 30 to 100 groups according to the network load, and identifies the static latency difference of the cellular network in batches using the least squares method. It simultaneously performs uplink abnormal sample removal, 5G URLLC time slot pre-compensation, and cell handover signaling monitoring. The AEKF adaptively adjusts the process noise matrix to track dynamic latency disturbances. The real-time total latency difference is substituted into the reconstructed clock offset formula, and the local TCXO crystal oscillator is corrected through feedforward PI servo correction. The corrected synchronous clock drives the main network side and the load side voltage ADC to sample synchronously, monitor the synchronization error in real time, and block the closing signal when the error exceeds the standard.

[0163] After 24 hours of continuous operation testing, the static synchronization error remained stable between 0.3 and 0.4 μs, the dynamic jitter under high load was always below 2 μs, the phase detection results of the ring main unit were undistorted, and the closing logic was safe and reliable.

[0164] This invention is used in the scenario of simultaneous detection of long-distance ring main units across different areas in county-level rural power grids:

[0165] Using 4G cellular networks as the synchronization message transmission link, the maximum distribution distance between ring network cabinets reaches 5 kilometers, making it impossible to use short-range wireless time synchronization solutions such as LoRa.

[0166] The system continuously monitors 4G module cell handover signaling. When cell handover is detected, it automatically clears the timestamp cache and resets the AEKF filter covariance matrix to eliminate instantaneous handover delay jumps. It automatically identifies and removes abnormal delay samples caused by uplink random backoff. The hierarchical joint compensation architecture, together with the feedforward PI servo, continuously corrects the local clock. The synchronous clock drives the synchronous acquisition of electrical quantities on both sides of the line to complete remote long-distance synchronous grid connection detection.

[0167] The test results show that the synchronization jitter is stably controlled within 1.9μs in cross-regional long-distance scenarios, and there is no accuracy drift during long-term continuous operation.

[0168] The above description, in conjunction with specific embodiments, provides a further detailed explanation of the present invention. It should not be construed that the specific implementation of the present invention is limited to these descriptions. For those skilled in the art, various simple deductions or substitutions can be made without departing from the concept of the present invention, and all such modifications and substitutions should be considered within the scope of protection of the present invention.

Claims

1. A master-slave clock synchronization method for power distribution network based on cellular network, used for time synchronization of master clock unit of power distribution network master station and slave unit of remote ring cabinet through 4G / 5G cellular communication network; characterized in that: include S1. Obtain the operating temperature of the master clock unit and slave unit, and obtain the current hardware latency compensation amount of the master clock unit and slave unit; S2. Construct a sliding timestamp buffer window, collect timestamp information corresponding to multiple sets of master-slave clock message interactions based on the PTP protocol, and construct a multi-period observation sample set; S3. Based on a multi-period observation sample set, an asymmetric delay PTP observation equation set is introduced. The static delay difference of the cellular network is solved by least squares batch identification, and the global static delay compensation is obtained by superimposing it with the basic compensation amount from the unit. S4. Obtain the clock deviation residual of each round of PTP interaction as the observation value, update it using an adaptive extended Kalman filter, obtain the dynamic delay disturbance at the current moment, and superimpose the dynamic delay disturbance with the global static delay compensation to obtain the real-time complete uplink and downlink total delay difference. S5. The clock offset estimate is corrected using the real-time complete uplink and downlink total delay difference, and the slave unit clock is corrected using clock servo control.

2. The method of claim 1, wherein the method further comprises: The specific methods for obtaining the basic hardware latency compensation amount in step S1 include: Acquire 4G / 5G RF transmission path delay of master clock unit and slave unit within the temperature range of -40℃ to 70℃. and RF receiving path delay Construct a temperature-RF delay lookup table; Obtain the real-time operating temperature of the master clock unit and slave unit, look up the table to obtain the RF transmission path delay and RF reception path delay of the master clock unit and slave unit at the current temperature, and calculate the basic hardware delay compensation amount for each unit: Basic hardware delay compensation amount = RF transmission path delay of the corresponding unit - RF reception path delay of the corresponding unit.

3. The master-slave clock synchronization method for a distribution network based on a cellular network according to claim 1, characterized in that: Constructing a multi-period observation sample set involves continuously collecting 30 to 100 complete quaternary timestamps as the multi-period observation sample set. The quaternary timestamps include the timestamps sent periodically by the master clock unit for Sync messages, FollowUp messages, and MAC hardware capture messages. Capture Sync message receive timestamp from unit MAC hardware Local transmission timestamp of the DelayReq message sent from the unit Master clock MAC hardware capture of DelayResp message receive timestamp .

4. The master-slave clock synchronization method for a distribution network based on a cellular network according to claim 3, characterized in that: It also includes removing abnormal samples from the constructed multi-period observation sample set: the uplink round-trip delay of each DelayReq message group is counted, the uplink round-trip delay is compared with the preset uplink delay threshold, and if it exceeds the threshold, it is determined to be an abnormal sample and removed from the multi-period observation sample set.

5. The master-slave clock synchronization method for a distribution network based on a cellular network according to claim 1, characterized in that: In step S4, the state vector of the adaptive extended Kalman filter (AEKF) is: wherein, is a clock offset, is a clock frequency offset, is a static latency difference of the cellular network solved for step S3, is a dynamic latency perturbation; The state equation and the observation equation adapting to the drift characteristics of the power crystal oscillator and the evolution law of the cellular time delay are established, a PTP synchronization iteration period is set as , and a 4×4 order state transition matrix is constructed. The state prediction equation is wherein, is the prior state estimate at time k; is the process noise, , is normally distributed; is the process noise covariance matrix; Covariance prediction is performed based on the state transition matrix. The covariance prediction formula is as follows: wherein, is the prior state estimation covariance matrix; is the state estimation filter covariance matrix after completion of the previous iteration, used to represent the uncertainty of the estimation error of each component of the four-dimensional state vector; is the transpose matrix of the state transition matrix F; The observation equation is wherein is the PTP clock bias residual; observation matrix , is the observation noise, is the observation noise covariance matrix; Calculate the Kalman gain based on the observation matrix and the observation noise covariance matrix: wherein, is the Kalman gain matrix; is the transpose of the observation matrix H; The state vector is corrected using Kalman gain, and the updated state estimation filter covariance matrix is ​​obtained: in, It is a 4×4 identity matrix; The state update is completed by correcting the prior state variables using observed residuals. extracting a cellular network dynamic latency disturbance component from an output state vector extracting a cellular network dynamic latency disturbance component from an output state vector .

6. The method of claim 5, wherein the method further comprises: The process noise covariance matrix Q is based on the mean of round-trip delay jitter. To make adaptive adjustments, specifically: in, is the round-trip delay of the i-th PTP round; N is the number of quaternary timestamp sample groups collected within the sliding timestamp buffer window; The historical average round-trip latency within the sliding timestamp cache window; if Greater than the preset threshold Then the process noise matrix is ​​adjusted to Amplify the process noise matrix; otherwise, adjust the process noise matrix to... This reduces the noise matrix during the process.

7. The method of claim 5, wherein the method further comprises: It also includes cell handover processing: real-time monitoring of cell handover signaling reported from the 4G / 5G communication module of the unit; when a cell handover action is detected, clearing the sliding timestamp buffer window and resetting the state estimation filter covariance matrix.

8. The method of claim 1, wherein the method further comprises: When using 5G cellular networks for transmission, it also includes time slot scheduling pre-compensation for TDD standards, specifically: Obtain the time slot allocation parameters issued by the 5G cellular network to determine the number of downlink time slots within a single scheduling cycle. Uplink time slots and special time slot numbers ; Setting single-slot reference transmission delay and special slot transition delay ; Computing downlink average scheduling latency uplink average scheduling latency ; Computing time slot scheduling inherent latency difference values ; The global static latency compensation amount is superimposed to the 5G cellular network. The global static latency compensation amount is superimposed to the 5G cellular network.

9. The method of claim 1, wherein the method further comprises: The specific formula for correcting the clock offset estimate using the real-time complete uplink and downlink total delay difference in step S5 is as follows: in, Estimate the actual clock offset; The master clock unit periodically sends Sync messages, FollowUp messages, and MAC hardware capture messages with timestamps. To capture the timestamp of the Sync message reception from the unit MAC hardware; For the local transmission timestamp of the DelayReq message sent from the unit; The master clock MAC hardware captures the DelayResp message reception timestamp; This represents the real-time, complete total uplink and downlink latency difference. The clock servo control is a feedforward PI clock servo; the real clock offset estimate is input into the PI controller, and the real-time complete uplink and downlink total delay difference is directly injected into the servo output as the feedforward correction amount. The PI controller outputs the clock adjustment amount, and the servo output adjustment amount acts on the TCXO temperature control crystal oscillator of the slave unit.

10. A cellular network-based master-slave clock synchronization system for power distribution networks, configured to perform a cellular network-based master-slave clock synchronization method according to any one of claims 1 to 9, characterized in that: include: The distribution network master station's PTP master clock unit, 4G / 5G cellular communication network, and at least one remote ring network cabinet's synchronization detection slave unit; The main clock unit includes an industrial-grade 4G / 5G communication module, a MAC layer hardware timestamp capture module, a TCXO temperature-controlled crystal oscillator clock source, an RF temperature acquisition sensor, and local non-volatile memory. The master clock unit captures the timestamp of the PTP message through its MAC layer hardware timestamp capture module, and corrects the hardware delay deviation of its own timestamp through the radio frequency temperature acquisition sensor and the temperature-delay lookup table in the local non-volatile memory. The unit includes an industrial-grade 4G / 5G communication module, a MAC layer hardware timestamp capture module, a TCXO temperature-controlled crystal oscillator clock source, an RF temperature acquisition sensor, local non-volatile memory, an AEKF algorithm processing unit, a cellular-specific correction logic processing unit, and a feedforward PI clock servo unit. Temperature data acquired from the unit's RF temperature acquisition sensor is used to obtain the hardware delay base compensation amount through the temperature-delay lookup table in the local non-volatile memory. The MAC layer hardware timestamp capture module captures the timestamp of the PTP message and inputs it to the AEKF algorithm operation unit and the cellular dedicated correction logic processing unit. The outputs of the AEKF algorithm operation unit and the cellular dedicated correction logic processing unit are superimposed and input to the feedforward PI clock servo unit. The output adjustment amount of the feedforward PI clock servo unit is applied to the TCXO temperature control crystal oscillator clock source. The master clock unit and the slave unit exchange PTP messages through the 4G / 5G cellular communication network.

Images