CN-122027065-A - Multi-node time synchronization method and dual-mode communication module

CN122027065ACN 122027065 ACN122027065 ACN 122027065ACN-122027065-A

Abstract

The invention is applicable to the technical field of power communication, and relates to a multi-node time synchronization method and a dual-mode communication module, wherein the method comprises the following steps of S10, dual-mode link interference prediction and cross pre-synchronization driven by AI; S20, multi-factor federal learning clock drift compensation, S30, electric Hongshe national security trusted time synchronization and safety protection, wherein the dual-mode communication module comprises a high-precision clock module, a PLC communication unit, an RF communication unit, an electric hongshe trusted execution unit and an AI reasoning unit. The invention has simple structure and convenient operation, solves the problems of low time setting precision, weak anti-interference capability and insufficient safety protection of the traditional dual-mode communication module in the low-voltage area, and provides a high-reliability and high-safety time reference for the area service.

Inventors

DENG QINGTANG
CHEN JUNJIAN
SUN QIN
ZHUANG KAIXIANG
YANG KANGPING
Jiang Weihuang

Assignees

南方电网数字电网研究院股份有限公司

Dates

Publication Date: 20260512
Application Date: 20260112

Claims (10)

1. A multi-node time synchronization method, comprising the steps of: S10, AI-driven dual-mode link interference prediction and cross pre-synchronization, wherein a CCO main node collects characteristic parameters of a PLC and an RF link and inputs the characteristic parameters to an LSTM neural network for interference prediction; triggering dual-mode cross presynchronization when the interference probability of any link is predicted to exceed a threshold value, namely sending a reference timestamp through a PLC link and simultaneously sending a check factor of the reference timestamp through an RF link; S20, multi-factor federal learning clock drift compensation, wherein each node locally builds and trains a multi-factor drift model based on temperature, load and noise based on historical clock deviation data to obtain local model parameters; uploading local model parameters to a CCO master node by each node; the CCO master node adopts a weighted federal average algorithm to aggregate local model parameters of all nodes to generate a global drift compensation model; S30, the reliable time setting and safety protection of the electronic mobile terminal E are that the structure of a time setting message is expanded to comprise a master node ID, a nanosecond time stamp, a link identifier, an SM3 hash value, an SM2 signature and a monotonic serial number, after receiving the time setting message, a slave node sequentially verifies the SM3 hash value, the SM2 signature and the monotonic serial number, and in a time setting initialization stage, a challenge-response two-way authentication based on an SM2 algorithm is executed by a CCO master node and the slave node.
2. The method according to claim 1, wherein in S10, the check factor is an SM3 hash value of the reference timestamp, and the characteristic parameters include SNR, packet loss rate, power line load of the PLC link, and SNR, channel occupancy, and noise level of the RF link.
3. The multi-node time synchronization method of the dual-mode communication module according to claim 2, wherein the dimension of the input layer of the LSTM neural network is 6, the hidden layer comprises 2 LSTM layers, 128 neurons are respectively, the output layer is a PLC interference probability and an RF interference probability, and the inference time is less than or equal to 10ms.
4. The multi-node time synchronization method of the dual-mode communication module according to claim 1, wherein in S20, after the global drift compensation model is generated, the global model is calibrated by a high-stability node reference cluster selected from nodes based on Raft distributed consensus algorithm.
5. The multi-node time synchronization method of the dual-mode communication module according to claim 4, wherein the weighting coefficient of the master node to each node model parameter is positively correlated with the node clock stability during federal learning aggregation, and the global model aggregation error is less than or equal to 0.01ms/h.
6. The multi-node time synchronization method of the dual-mode communication module according to claim 1, wherein in S30, when a security event of continuous signature failure or repeated sequence number is detected, a link isolation mechanism is automatically triggered and reported to the network management system.
7. The method of multi-node time synchronization of a dual-mode communication module as set forth in claim 6, wherein, the safe starting flow verification of the electric-control TEE trusted execution unit is passed through the safe starting flow verification of the electric-control operation system, the call delay of the national encryption algorithm is less than or equal to 1ms, and the requirement on real-time synchronization is met.
8. A dual mode communication module employing the method of any of claims 1 to 7, comprising: The high-precision clock module adopts a GPS/Beidou dual-mode time service chip, integrates a temperature compensation circuit and an AI cooperative calibration circuit, and is used for providing a reference clock and sensing multi-factor drift; The PLC communication unit supports an HPLC protocol and has low-delay transmission and interference prediction capability; the RF communication unit supports wireless radio frequency communication, is used as a backup of a PLC link and is used for transmitting a check factor; The electronic secret TEE trusted execution unit is internally provided with a national encryption algorithm SM2/SM3 hardware acceleration module, and a private key is stored in a TEE safe area and is used for digitally signing and checking a time message; and the AI reasoning unit is provided with a lightweight neural network processor and is used for executing LSTM link interference prediction and federal learning model aggregation.
9. The dual-mode communication module of claim 8, wherein the software element of the dual-mode communication module comprises: The AI link interference prediction module is used for predicting the interference probability of a future link based on six-dimensional characteristics of an LSTM neural network, such as an input SNR, a packet loss rate, a power line load, an RF channel occupancy rate, a temperature and a noise level; The dual-mode cross pre-synchronization module is used for triggering a PLC link transmission reference time stamp and an RF link to transmit a cross synchronization mechanism of the check factor when the predicted interference probability is larger than a preset threshold; The multi-factor drift compensation module is used for constructing and training a local drift model based on temperature, load and noise and supporting federal learning parameter aggregation to generate global compensation parameters; and the TEE national encryption time module is used for calling a national encryption algorithm in the TEE trusted execution unit to realize signature, signature verification and bidirectional trusted authentication of the time message.
10. The dual-mode communication module of claim 9, further comprising a power on-demand adaptation unit for performing drive adaptation and compliance detection with a power on-demand operating system, supporting remote inquiry of time status and security events by a power on-demand network manager.

Description

Multi-node time synchronization method and dual-mode communication module Technical Field The invention belongs to the technical field of power communication, and particularly relates to a multi-node time synchronization method and a dual-mode communication module. Background In the novel power system, the low-voltage station area needs to realize services such as 500-node low-time-delay data acquisition (less than or equal to 1 minute), undisturbed topology identification (ranging accuracy less than or equal to 30 meters) and the like, and the services depend on multi-node high-accuracy time synchronization. The prior art has the defects that the link has weak anti-interference capability, the traditional dual-mode time synchronization depends on 'single link leading+passive switching', and dynamic environments such as power line impulse noise, RF channel interference and the like are easy to cause timestamp transmission distortion; the drift compensation precision is insufficient, the traditional scheme only depends on single factor fitting (such as temperature) of historical data, does not combine special multi-factor coupling influence of power scenes such as load, noise and the like, has long-term synchronization deviation exceeding 5ms under extreme environments (-40 ℃ to 70 ℃), has low safety protection level, is easy to be subjected to identity forging, timestamp tampering and replay attack due to the fact that a time synchronization protocol adopts CRC (cyclic redundancy protocol) or simple encryption, such as an NTP scheme disclosed by a fox searching network, and cannot meet the 2.0-level requirements of power and the like, has poor adaptability of a electronic system, is difficult to realize time synchronization protection due to the fact that a part of time synchronization scheme does not deeply integrate an electronic-aided TEE (terminal equipment) trusted environment and a national secret algorithm, and is difficult to utilize the safety capability of the electronic system. The patent with the publication number of CN108449791B provides a self-adaptive time synchronization method based on temperature compensation, which firstly utilizes the correlation between clock drift and temperature to establish a temperature-crystal oscillator frequency model, and nodes can compensate the offset of clocks according to the change condition of temperature under the model, so that the precision of synchronization among the nodes is improved. And secondly, under the condition that the network time delay is a Gaussian model, combining with the correlation theory of probability time synchronization, the node can compensate the current time according to the maximum synchronization error allowed by the network and the accumulated clock deviation, and estimate the corresponding resynchronization interval. The patent only depends on a single factor (temperature) to realize time synchronization, has low safety protection and has the same defects as the prior art. Therefore, how to solve the problems of low time setting precision, weak anti-interference capability and insufficient safety protection of the traditional dual-mode communication module in the low-voltage area, and provide a time reference with high reliability and high safety for the area service is not known. Disclosure of Invention Aiming at the defects of the prior art, the invention aims to provide a multi-node time synchronization method so as to solve the problems of low time setting precision, weak anti-interference capability and insufficient safety protection of the traditional dual-mode communication module in a low-voltage station area, and provide a high-reliability and high-safety time reference for station area service. In order to solve the technical problems, the invention adopts the following technical scheme: in a first aspect, the present invention provides a multi-node time synchronization method, comprising the steps of: S10, AI-driven dual-mode link interference prediction and cross pre-synchronization, wherein a CCO main node collects characteristic parameters of a PLC and an RF link and inputs the characteristic parameters to an LSTM neural network for interference prediction; triggering dual-mode cross presynchronization when the interference probability of any link is predicted to exceed a threshold value, namely sending a reference timestamp through a PLC link and simultaneously sending a check factor of the reference timestamp through an RF link; S20, multi-factor federal learning clock drift compensation, wherein each node locally builds and trains a multi-factor drift model based on temperature, load and noise based on historical clock deviation data to obtain local model parameters; uploading local model parameters to a CCO master node by each node; the CCO master node adopts a weighted federal average algorithm to aggregate local model parameters of all nodes to generate a global drift compensation model; S30, the reliabl