An adaptive anti-interference synchronous transmission method for cable tunnel partial discharge monitoring

Abstract

This invention discloses an adaptive anti-interference synchronous transmission method for partial discharge (PD) monitoring in cable tunnels. It involves deploying field nodes, an edge aggregation gateway, and a remote monitoring center. Upon power-up, the nodes establish an initial wireless mesh self-organizing network. The gateway calibrates its local clock via satellite timing, and the nodes achieve dual high-precision synchronization through PTP master-slave synchronization and LoRa mutual synchronization. The nodes sense and classify electromagnetic interference, predict optimal network parameters using a BP neural network, and adaptively adjust frequency bands, antennas, and bandwidth to reconstruct the mesh network. A sparse representation algorithm is used to separate PD signals from interference at the edge, improving the signal-to-noise ratio (SNR) before adding timestamps for uploading. The remote monitoring center identifies the PD type, generates statistical indicators, and feeds them back to the nodes, updating model parameters and forming a closed-loop optimization. This invention achieves microsecond-level synchronization, high SNR transmission, and high-success-rate data aggregation even in environments with no satellite signals and strong electromagnetic interference, supporting accurate location and early warning of PD defects in cable tunnels.

Description

Technical Field

[0001] This invention belongs to the field of high-voltage power equipment condition monitoring and Internet of Things communication technology, specifically relating to an adaptive anti-interference synchronous transmission method for partial discharge monitoring in cable tunnels. Background Technology

[0002] High-voltage cable tunnels are vital channels for urban power transmission, and monitoring partial discharge of cables within them is crucial for preventing insulation failures and ensuring power grid safety. Because cable tunnels are typically located deep underground in narrow, enclosed spaces, they lack coverage from external wireless signals (such as 4G / 5G, GPS / BeiDou), preventing monitoring data from being transmitted via the public network. Furthermore, the tunnels contain various electromagnetic interferences, including power harmonics and stray radiation from equipment, and their varied physical structures (bends, supports, uneven walls, etc.) pose significant challenges to the reliable acquisition and real-time uploading of partial discharge monitoring data.

[0003] In existing technologies, a common approach is to distribute multiple monitoring nodes within a tunnel, using a wireless ad hoc network (such as ZigBee or Wi-Fi Mesh) to aggregate partial discharge (PD) data to a gateway at the tunnel entrance, and then transmit it to a remote monitoring center via fiber optic cable. However, these approaches typically have the following drawbacks: First, the nodes use fixed wireless communication frequency bands (e.g., only 2.4GHz) and a single antenna type (e.g., only omnidirectional), making it impossible to dynamically adjust according to time-varying electromagnetic interference within the tunnel, which can easily lead to signal submersion or transmission interruption. Second, since there are no satellite signals within the tunnel, clock synchronization between nodes relies on periodic message exchanges, but existing synchronization methods lack compensation mechanisms for electromagnetic interference, resulting in degraded synchronization accuracy and failing to meet the microsecond-level time synchronization requirements for distributed PD positioning. Furthermore, the network parameters (frequency band, bandwidth, antenna) and synchronization process are independent of each other, failing to form a joint optimization, and the PD signals collected by the nodes contain a large amount of environmental interference; directly uploading these signals would consume excessive wireless bandwidth, affecting transmission efficiency and signal-to-noise ratio.

[0004] Therefore, there is an urgent need for a method that can achieve high-precision synchronization, anti-interference transmission, and adaptive networking of partial discharge monitoring data in environments with no satellite signals and strong electromagnetic interference, in order to solve the problems of poor transmission reliability, low synchronization accuracy, and insufficient bandwidth utilization in existing technologies. Summary of the Invention

[0005] The technical problem this invention aims to solve is that, in cable tunnels where there are no satellite signals and strong electromagnetic interference, existing technologies cannot achieve adaptive adjustment of network parameters, microsecond-level high-precision synchronization between nodes, and edge separation of partial discharge signals, resulting in poor transmission reliability, low synchronization accuracy, and insufficient bandwidth utilization.

[0006] To address the aforementioned technical problems, this invention provides an adaptive anti-interference synchronous transmission method for partial discharge monitoring in cable tunnels, comprising:

[0007] S1: Distribute multiple field adaptive networking nodes on cable supports or cable bodies within the cable tunnel; deploy edge aggregation gateways at the entrances and exits of the cable tunnel; deploy the remote monitoring center in the remote monitoring room; wherein the wireless communication coverage of adjacent field adaptive networking nodes overlaps with each other;

[0008] S2: After each field adaptive networking node is powered on, it automatically loads the factory default parameters and establishes an initial wireless Mesh self-organizing network based on the default parameters;

[0009] S3: The user generates a configuration command at the remote monitoring center and sends the configuration command to the edge aggregation gateway through fiber optic cable. The edge aggregation gateway broadcasts the configuration command to all field adaptive networking nodes through the initial wireless mesh self-organizing network. Each field adaptive networking node completes the configuration according to the command.

[0010] S4: The edge aggregation gateway receives satellite timing signals to calibrate its local clock; each field adaptive networking node calibrates its local clock through PTP master-slave synchronization based on the local clock of the edge aggregation gateway, and at the same time, each field adaptive networking node performs mutual synchronization calibration of its local clock.

[0011] S5: Each field adaptive networking node collects electromagnetic interference data and partial discharge mixed signals from the cable tunnel; based on the electromagnetic interference data and partial discharge mixed signals, the optimal networking parameters are predicted through the networking parameter mapping model; each field adaptive networking node adjusts the networking parameters according to the optimal networking parameters and re-establishes the wireless Mesh self-organizing network.

[0012] S6: Each field adaptive networking node uses a dual compensation model to dynamically synchronize and compensate the local clock for high-precision clock based on the interference perception results, thus obtaining a high-precision local clock.

[0013] S7: Each field adaptive networking node uses a sparse representation signal decomposition algorithm based on an overcomplete dictionary to remove interference data in the partial discharge mixed signal according to the signal separation algorithm threshold. Then, it uses a bandpass filter to remove residual noise outside the frequency band to obtain high signal-to-noise ratio pure partial discharge data. Based on a high-precision local clock, it generates a high signal-to-noise ratio pure partial discharge data packet with a high-precision timestamp and transmits it to the edge aggregation gateway through a wireless Mesh self-organizing network.

[0014] S8: The edge aggregation gateway transmits high signal-to-noise ratio (SNR) clean partial discharge (PD) data packets with high-precision timestamps to the remote monitoring center via optical fiber. The remote monitoring center generates PD type identification results based on the high SNR clean PD data packets with high-precision timestamps and generates system operation statistics.

[0015] S9: The remote monitoring center transmits the system operation statistics to the edge aggregation gateway via fiber optic cable, and the edge aggregation gateway transmits them to each field adaptive networking node via a wireless mesh self-organizing network; each field adaptive networking node adaptively updates the parameters of the networking parameter mapping model and the dual compensation model according to the system operation statistics.

[0016] The present invention has at least the following beneficial effects

[0017] This invention significantly improves transmission reliability by collecting electromagnetic interference characteristics and dynamically predicting optimal network parameters using a BP neural network, achieving adaptive adjustment of frequency bands, antennas, and bandwidth, as well as wireless mesh network reconstruction. Through a dual synchronization network combining PTP master-slave synchronization and LoRa inter-node synchronization, and utilizing FPGA hardware-accelerated Kalman filtering and crystal drift compensation models, high-precision clock synchronization at the microsecond level (≤±0.5μs) is achieved even without satellite signals. Simultaneously, a sparse representation signal separation algorithm based on an overcomplete dictionary removes interference components from the partial discharge mixed signal in real time at the node end. Combined with bandpass filtering optimization, this significantly improves the signal-to-noise ratio of the partial discharge signal and reduces wireless bandwidth usage. Finally, system operation statistics generated by the remote monitoring center are fed back to each node, achieving closed-loop adaptive optimization of the network parameter mapping model and the dual compensation model. This results in high signal-to-noise ratio, microsecond-level synchronization, high success rate, and low-latency transmission of partial discharge monitoring data even without satellite signals, providing reliable technical support for the accurate location and early warning of partial discharge defects in cable tunnels. Attached Figure Description

[0018] Figure 1 This is a topology diagram of the overall three-layer architecture of the system of this invention;

[0019] Figure 2 This is a block diagram of the core hardware integration of the present invention;

[0020] Figure 3 This is an overall flowchart of the adaptive anti-interference synchronous transmission method of the present invention;

[0021] Figure 4 This is a block diagram illustrating the core innovative linkage principle of this invention. Detailed Implementation

[0022] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] Please see Figures 1-4 This invention provides an adaptive anti-interference synchronous transmission method for partial discharge monitoring in cable tunnels, comprising:

[0024] S1: Distribute multiple field adaptive networking nodes on cable supports or cable bodies within the cable tunnel; deploy edge aggregation gateways at the entrances and exits of the cable tunnel; deploy the remote monitoring center in the remote monitoring room; wherein the wireless communication coverage of adjacent field adaptive networking nodes overlaps with each other;

[0025] In this embodiment, based on the cable tunnel's route (including straight sections and corner sections), multiple field adaptive networking nodes are sequentially installed on cable supports or the cable itself. The following principles are followed during installation: In straight sections of the tunnel, nodes are arranged at equal intervals, with the spacing ensuring sufficient overlap in the wireless communication coverage of adjacent nodes (typically, the overlap area is no less than 20% of the node's coverage radius), thereby eliminating signal blind spots and supporting multi-hop data forwarding. In corner sections of the tunnel, due to severe wall obstruction, additional nodes should be installed on both sides of the corner, ensuring direct communication between the nodes on both sides, or relay communication through an intermediate node. All nodes should be installed at heights that avoid strong electromagnetic interference sources (such as cable joints, grounding wires, etc.), and the antennas should be oriented towards the tunnel axis or along the cable's direction. Edge aggregation gateways are deployed at tunnel entrances / exits (usually one per entrance / exit), locations with the capability to receive BeiDou / GPS satellite signals. The gateways are connected to a remote monitoring center via pre-buried optical fibers. The remote monitoring center is deployed in a manned monitoring room or power dispatch center, receiving data from each entrance / exit gateway via a dedicated network.

[0026] Field adaptive networking nodes: Each node integrates the following functional modules:

[0027] Power module: Consists of an inductive power-harvesting coil, a lithium battery, and a power management chip. The inductive power-harvesting coil is clamped to a cable and uses the cable's power frequency current to generate an induced electromotive force, which is then rectified and regulated to charge the lithium battery. The power management chip is responsible for switching between power harvesting and battery power supply, and outputting a stable voltage and current for use by other modules.

[0028] Partial Discharge Acquisition and Interference Sensing Unit: This unit comprises a partial discharge acquisition submodule (high-frequency pulse current sensor + ultra-high frequency sensor) and an interference sensing submodule (miniature spectrum analyzer + electromagnetic interference intensity detection chip). The former is used to capture cable partial discharge signals, while the latter is used to scan for electromagnetic interference in the 0~6GHz frequency band within the tunnel in real time, outputting the spectrum, time-domain waveform, interference frequency band, and intensity value.

[0029] The adaptive network transmission unit consists of a dual-band RF module (1.4GHz / 2.4GHz), an antenna switching module (electromagnetic switch), directional / omnidirectional antennas, and a bandwidth scheduling module. The RF module can switch frequency bands according to commands. The antenna switching module selects between directional and omnidirectional antennas based on the tunnel section (straight line / corner) and interference conditions. In cable tunnel corners, due to severe wall obstruction and changes in axis direction, the narrow beam of the directional antenna cannot cover multiple directions. Therefore, the antenna switching module will force a switch to an omnidirectional antenna to ensure reliable communication with adjacent segment nodes and eliminate signal blind spots. The bandwidth scheduling module supports 10 / 20 / 50 / 100Mbps adaptive bandwidth.

[0030] Satellite-free high-precision clock synchronization unit: includes a high-stability temperature-controlled crystal oscillator (OCXO), an IEEE 1588 PTPv2 synchronization chip, a 433MHz LoRa inter-synchronization communication module, and an FPGA compensation unit. The OCXO provides a local high-stability clock; the PTPv2 chip processes synchronization messages from the gateway; the LoRa module is used for direct clock information exchange between nodes; and the FPGA executes Kalman filtering and drift compensation algorithms.

[0031] Edge signal processing unit: Using FPGA, it runs a sparse representation signal separation algorithm to separate the interference components in the mixed signal input from the partial discharge acquisition submodule and output a clean partial discharge signal.

[0032] Microcontroller Unit (MCU): Employs STM32 series chips to run interference classification (random forest), network parameter mapping (BP neural network), and synchronization compensation strategies, and controls other modules via buses (SPI, I2C, UART).

[0033] All modules are integrated on a single PCB, with an explosion-proof alloy housing and an IP67 protection rating. They are secured to cable supports using cable ties or bolts. The field adaptive networking nodes are centered around a microcontroller unit (MCU), connected to the adaptive networking transmission unit and the satellite-free high-precision clock synchronization unit via an SPI bus, connected to the edge signal processing unit (FPGA) via a parallel bus, connected to the partial discharge acquisition and interference sensing unit via an ADC / SPI, and connected to the inter-node synchronization communication module via a UART. Meanwhile, the power module reports its status to the MCU via I2C. All modules work together to complete data acquisition, synchronization, and transmission.

[0034] The edge aggregation gateway integrates the following functional modules:

[0035] Power module: adopts industrial-grade switching power supply (AC220V input, DC12V output) and is equipped with a backup battery.

[0036] Multi-channel network receiving module: Multiple wireless radio frequency channels simultaneously receive Mesh data packets uploaded by each node;

[0037] Satellite-less master clock reference unit: Beidou / GPS dual-mode timing module, ultra-high stability OCXO (drift rate ≤0.001ppm / ℃), PTPv2 master clock chip, used to generate the reference time of the entire network;

[0038] Data aggregation and processing module: High-performance embedded processor (such as ARM Cortex-A series) is responsible for data verification, deduplication, and format conversion.

[0039] Fiber optic transmission module: Single-mode optical module, 1Gbps speed, connected to a remote monitoring center.

[0040] Industrial Ethernet module: Provides an RJ45 interface for local debugging or backup communication.

[0041] The edge aggregation gateway is centered on a data aggregation and processing module. It connects to a multi-channel network receiving module via PCIe / USB to achieve wireless data transmission and reception, connects to a PTPv2 master clock chip via MII / SPI interface to generate synchronization messages, connects to an optical fiber transmission module via SGMII to communicate with the remote end, and monitors the power status via I2C bus. All modules work together to complete data aggregation and forwarding.

[0042] The remote monitoring center integrates the following functional modules:

[0043] Industrial server (dual CPU, large-capacity memory and RAID disk array); monitoring terminal (monitor, keyboard and mouse) and large-screen display system. Pre-installed partial discharge monitoring platform software, including deep learning recognition model, TDOA positioning algorithm, and statistical indicator calculation module.

[0044] S2: After each field adaptive networking node is powered on, it automatically loads the factory default parameters and establishes an initial wireless Mesh self-organizing network based on the default parameters;

[0045] In this embodiment, after each field adaptive networking node is powered on, the microcontroller unit (MCU) first reads the factory default parameters from its internal Flash memory, including: communication frequency band (e.g., 1.4GHz), antenna type (directional antenna), transmission bandwidth (20Mbps), Mesh network ID (preset unique identifier), wireless channel number, and routing protocol parameters (e.g., Hello message interval, route failure time, etc.). The MCU writes these parameters into the dual-band RF module, antenna switching module, and bandwidth scheduling module of the adaptive networking transmission unit via the SPI bus, completing the RF initialization.

[0046] Subsequently, the adaptive networking transmission unit initiates the wireless mesh self-organizing network protocol stack (such as the HWMP routing protocol based on 802.11s). Nodes broadcast Beacon frames on designated channels, carrying the network ID and their own MAC address; simultaneously, they listen for Beacon frames from neighboring nodes. When a node receives a Beacon from a neighboring node, it establishes a peer link through a four-way handshake and exchanges routing information. Each node maintains a neighbor table and calculates the optimal multi-hop path to the edge aggregation gateway through periodic exchange of routing messages (PREQ / PREP).

[0047] The edge aggregation gateway is also configured with the same network ID and acts as the portal node of the Mesh network (connected to the fiber optic backbone). Once the field nodes discover the gateway's MAC address and calculate the routing path through the above process, the initial wireless Mesh self-organizing network is established. Afterward, any node can communicate with the gateway via multi-hop relays to transmit configuration commands or data packets. The entire process requires no manual intervention and is completed automatically upon power-up.

[0048] S3: The user generates a configuration command at the remote monitoring center and sends the configuration command to the edge aggregation gateway through fiber optic cable. The edge aggregation gateway broadcasts the configuration command to all field adaptive networking nodes through the initial wireless mesh self-organizing network. Each field adaptive networking node completes the configuration according to the command.

[0049] Preferably, the configuration instructions include: partial discharge acquisition sampling rate, interference sensing scanning frequency band, initial synchronization parameters, and signal separation algorithm threshold; wherein, the initial synchronization parameters include PTP master-slave synchronization period, inter-node mutual synchronization period, and deviation compensation threshold.

[0050] Preferably, the field adaptive networking node is powered by a power module consisting of an inductive power collection coil, a lithium battery, and a power management chip; the inductive power collection coil is sleeved on a cable to collect electrical energy, and the power management chip is used to convert and store the electrical energy collected by the inductive power collection coil into the lithium battery, and to convert and supply the electrical energy output by the lithium battery.

[0051] Preferably, the field adaptive networking node is configured with an adaptive networking transmission unit and an inter-node synchronization communication module. The adaptive networking transmission unit includes: a dual-band radio frequency module, an antenna switching module, multiple types of antennas, and a bandwidth scheduling module. The dual-band radio frequency module is used for switching between 1.4GHz and 2.4GHz dual bands. The multiple types of antennas include directional antennas and omnidirectional antennas. The antenna switching module is used for automatic switching between directional antennas and omnidirectional antennas. The bandwidth scheduling module is used for adaptive allocation of multiple bandwidths. The inter-node synchronization communication module is used to broadcast the local clock information of the field adaptive networking node according to the inter-node synchronization period via 433MHz LoRa narrowband communication.

[0052] In this embodiment, on the operation interface of the remote monitoring center, maintenance personnel set the following configuration parameters according to the actual environment and monitoring needs of the cable tunnel: partial discharge sampling rate (e.g., 200MS / s), interference sensing scanning frequency band (e.g., 0~6GHz), initial synchronization parameters (including PTP master-slave synchronization period of 1s, inter-node mutual synchronization period of 2s, deviation compensation threshold ±0.1μs), and signal separation algorithm threshold (e.g., residual energy threshold ε=0.01). The industrial server of the remote monitoring center encapsulates these parameters into configuration instructions in JSON format and transmits them to the edge aggregation gateway via optical fiber.

[0053] After receiving the instruction, the fiber optic transmission module of the edge aggregation gateway parses it via the industrial Ethernet module and then forwards it to the multi-channel networking receiving module. The multi-channel networking receiving module, based on the currently established initial wireless mesh self-organizing network, encapsulates the configuration instruction into a broadcast data packet (destination address is the broadcast MAC address) and transmits it at maximum power in the 1.4GHz band with a bandwidth of 20Mbps.

[0054] After receiving the broadcast data packet, the adaptive networking transmission unit of each field adaptive networking node transmits the data packet content to the microcontroller unit (MCU) via the SPI bus. The MCU parses the configuration instructions and processes them according to the fields: it writes the "partial discharge acquisition sampling rate" to the control register of the partial discharge acquisition submodule; it writes the "interference sensing scanning frequency band" to the start and end frequency registers of the micro spectrum analyzer; it writes the PTP master-slave synchronization period, inter-node mutual synchronization period, and deviation compensation threshold from the "initial synchronization parameters" to the corresponding registers of the satellite-free high-precision clock synchronization unit; and it writes the "signal separation algorithm threshold" to the threshold register of the edge signal processing unit (FPGA). After all parameters are written, each module returns an acknowledgment flag. After the MCU confirms that there are no errors, it replies with a configuration success response to the edge aggregation gateway via the adaptive networking transmission unit. The gateway then summarizes and reports the information to the remote monitoring center, completing the configuration process.

[0055] S4: The edge aggregation gateway receives satellite timing signals to calibrate its local clock; each field adaptive networking node calibrates its local clock through PTP master-slave synchronization based on the local clock of the edge aggregation gateway, and at the same time, each field adaptive networking node performs mutual synchronization calibration of its local clock.

[0056] Preferably, step S4 includes: the edge aggregation gateway receiving satellite timing signals, calculating the UTC reference time, calibrating the local clock according to the UTC reference time as the master reference clock; obtaining the master reference time from the master reference clock, generating PTP synchronization messages according to the PTP master-slave synchronization cycle, and broadcasting the PTP synchronization messages to each field adaptive networking node through the wireless Mesh self-organizing network; each field adaptive networking node actively broadcasting local clock information through the inter-node mutual synchronization communication module according to the inter-node mutual synchronization cycle;

[0057] Each field adaptive networking node receives the PTP synchronization message, parses the master reference time in the PTP synchronization message, compares the master reference time with the local clock time, calculates the master-slave time deviation, and calibrates the local clock according to the master-slave time deviation to complete the local clock calibration of each field adaptive networking node.

[0058] If a field adaptive networking node does not receive a PTP synchronization message from the edge aggregation gateway for three consecutive PTP master-slave synchronization cycles, the field adaptive networking node extracts a timestamp from the local clock information of the adjacent field adaptive networking nodes, compares the extracted timestamp with the local clock time of the field adaptive networking node, calculates the mutual slave time deviation, and adjusts the local clock according to the mutual slave time deviation to complete the local clock calibration of each field adaptive networking node.

[0059] After the edge aggregation gateway is activated, its BeiDou / GPS dual-mode timing module receives signals from multiple satellites and calculates a high-precision UTC reference time (accuracy better than ±50ns). The ultra-high stability OCXO master clock source in the satellite-free master clock reference unit calibrates its own frequency and phase according to this UTC time. After calibration, the frequency accuracy of the master clock source is better than ±0.001ppm. The PTPv2 master clock chip obtains the time reference from the master clock source, generates PTP synchronization messages (Sync messages) and follow messages (Follow_Up messages) according to the configured PTP master-slave synchronization period (e.g., 1 second), and broadcasts them on the wireless mesh ad hoc network through the multi-channel networking receiver module.

[0060] Each adaptive networking node's adaptive networking transmission unit listens for and captures PTP synchronization messages on the wireless channel. Since the mesh network may have multiple hops, the gateway's PTP messages are forwarded to the entire network via relay nodes. Each node's adaptive networking transmission unit transmits the received PTP messages to the microcontroller unit (MCU) via the SPI bus. The MCU parses the master clock timestamp in the message and forwards it to the IEEE1588 PTPv2 synchronization chip in the satellite-free high-precision clock synchronization unit. This chip records the precise time when the local high-stability temperature-controlled crystal oscillator (OCXO) receives the message from the clock source, calculates the message transmission path delay using a peer-to-peer delay mechanism, and then calculates the master-slave time deviation. The PTPv2 synchronization chip returns this deviation value to the MCU via SPI. The MCU adjusts the phase or frequency of the local OCXO slave clock source according to the deviation value to synchronize the local clock with the gateway master clock, achieving an accuracy better than ±0.5μs.

[0061] Simultaneously, according to the inter-node synchronization cycle (e.g., 2 seconds), each node's MCU controls the inter-node synchronization communication module (433MHz LoRa) to actively broadcast local clock information. Upon receiving this information, the LoRa module of adjacent nodes sends a timestamp to the MCU and calculates the synchronization deviation. If no gateway message is received for three consecutive PTP cycles or the master-slave deviation exceeds the deviation compensation threshold (e.g., ±0.1μs), the node automatically switches to synchronization mode: the MCU sends a request to adjacent normal nodes via LoRa, receives the response timestamp, calculates the weighted average deviation, and adjusts the local clock. When the node receives the gateway PTP message again, the MCU merges the master-slave synchronization result with the synchronization result according to weights (e.g., master-slave 0.7, synchronization 0.3) as the final calibration value, ensuring the clock remains stable within ±0.5μs over a long period. The MCU records the time deviation generated by each synchronization in its internal storage area for local log recording and status monitoring.

[0062] S5: Each field adaptive networking node collects electromagnetic interference data and partial discharge mixed signals from the cable tunnel; based on the electromagnetic interference data and partial discharge mixed signals, the optimal networking parameters are predicted through the networking parameter mapping model; each field adaptive networking node adjusts the networking parameters according to the optimal networking parameters and re-establishes the wireless Mesh self-organizing network.

[0063] Preferably, step S5 includes:

[0064] Each on-site adaptive networking node scans the electromagnetic interference signal in the cable tunnel in real time using a miniature spectrum analyzer based on the interference sensing scanning frequency band, and obtains the spectral characteristics, time domain characteristics and frequency band of the electromagnetic interference signal; the interference intensity value of the electromagnetic interference signal is collected by the electromagnetic interference intensity detection chip, and the interference intensity value of the electromagnetic interference signal is the average power value of the electromagnetic interference signal.

[0065] Each field adaptive networking node uses a random forest machine learning classification algorithm to identify the interference category of electromagnetic interference signals based on the spectral and temporal characteristics of the electromagnetic interference signals. The interference categories of electromagnetic interference signals include harmonic interference, narrowband wireless interference, pulse interference, or spurious radiation interference.

[0066] Each on-site adaptive networking node identifies the interference intensity level of the electromagnetic interference signal according to the interference intensity value and a preset threshold, where: ≤-80dBm is level 1, -80~-70dBm is level 2, -70~-60dBm is level 3, -60~-50dBm is level 4, and ≥-50dBm is level 5.

[0067] Each field adaptive networking node acquires the partial discharge mixed signal of the cable through a high-frequency pulse current sensor and an ultra-high frequency sensor according to the partial discharge acquisition sampling rate, and extracts the amplitude of the partial discharge mixed signal;

[0068] Each field adaptive networking node uses a BP neural network networking parameter mapping model to predict the optimal networking parameters based on the interference type, interference intensity level, frequency band of the interference signal, and amplitude of the partial discharge mixed signal. The optimal networking parameters include communication frequency band, antenna type, and transmission bandwidth.

[0069] Each on-site adaptive networking node adjusts the communication frequency band, antenna type, and transmission bandwidth of the adaptive networking transmission unit according to the optimal networking parameters, and re-establishes the wireless Mesh self-organizing network.

[0070] In this embodiment, the microcontroller unit (MCU) of each field adaptive networking node reads the issued interference sensing scanning frequency band (e.g., covering the 0-6 GHz range of common interferences in tunnels) from its local configuration and starts a miniature spectrum analyzer for continuous scanning via the SPI bus. The miniature spectrum analyzer stays at each frequency point for a preset time, collects the power spectrum of the electromagnetic interference signal, and records the frequency band where the interference occurs and the time-domain waveform characteristics (such as pulse width and repetition period). Simultaneously, the electromagnetic interference intensity detection chip synchronously measures the average power value of electromagnetic interference in the environment. The above-mentioned spectral characteristics, time-domain characteristics, frequency band of interference, and interference intensity value are packaged into data frames and transmitted to the MCU via SPI.

[0071] After receiving the data, the MCU first runs a random forest classification algorithm: taking spectral features (such as peak frequency and bandwidth) and time-domain features (such as pulse shape and duty cycle) as input, it outputs the interference category (harmonic interference, narrowband wireless interference, pulse interference, or spurious radiation interference). Then, it classifies the interference intensity value into levels 1 to 5 according to a preset power threshold. The above interference category, intensity level, and frequency band of the interference together constitute the "interference perception result," which the MCU temporarily stores in memory and distributes to subsequent processing units.

[0072] Simultaneously, the MCU activates the partial discharge acquisition submodule based on the partial discharge acquisition sampling rate (e.g., 200 MS / s). A high-frequency pulse current sensor and an ultra-high-frequency sensor synchronously acquire the mixed partial discharge signal on the cable, which includes partial discharge pulses and environmental interference. The acquired waveform data is simultaneously sent to the MCU and the edge signal processing unit (FPGA) via a parallel bus. The MCU extracts the peak amplitude of the partial discharge signal from the waveform (e.g., by detecting the maximum value through a sliding window).

[0073] The MCU takes four sets of parameters as input features: interference category, interference intensity level, interference frequency band, and partial discharge signal amplitude. These are fed into a pre-programmed BP neural network networking parameter mapping model. This model has been trained offline using a large amount of tunnel test data, and its hidden layer node number and weight coefficients are stored in the MCU's Flash memory. After forward computation, the model outputs optimal networking parameters, including: communication frequency band (1.4GHz or 2.4GHz), antenna type (directional or omnidirectional), and transmission bandwidth (one of 10 / 20 / 50 / 100 Mbps). For example, when narrowband interference of level 3 or higher is detected in the 2.4GHz band and the partial discharge amplitude is small (not exceeding the limit), the model may output a switch to 1.4GHz, maintain a directional antenna, and a bandwidth of 20Mbps; when the partial discharge amplitude exceeds the warning threshold, the output bandwidth is increased to 100Mbps and the model switches to an omnidirectional antenna to increase coverage reliability.

[0074] The MCU writes the optimal networking parameters mentioned above into the dual-band RF module, antenna switching module, and bandwidth scheduling module of the adaptive networking transmission unit via SPI. The dual-band RF module immediately switches the center frequency and filter bandwidth according to the instructions; the antenna switching module connects the RF path to the specified type of antenna (directional or omnidirectional) by controlling the electromagnetic switch; the bandwidth scheduling module adjusts the modulation and coding scheme (MCS) of the baseband processor and the physical layer data rate to achieve the target bandwidth (e.g., 50Mbps).

[0075] After parameter adjustments are completed, the MCU triggers the adaptive networking transmission unit to reconstruct the wireless mesh network. Nodes send neighbor probe requests, and surrounding nodes respond with their current operating parameters and link quality. Each node re-establishes peer connections based on the new frequency band, antenna orientation, and bandwidth, and updates its routing table to ensure a seamless, interference-resistant optimized mesh topology across the entire network. Once reconstruction is complete, nodes resume data transmission with the edge aggregation gateway. The total latency of the entire adjustment process, from interference detection to mesh reconstruction completion, is controlled within milliseconds to adapt to rapid changes in the interference environment within the tunnel.

[0076] S6: Each field adaptive networking node uses a dual compensation model to dynamically synchronize and compensate the local clock for high-precision clock based on the interference perception results, thus obtaining a high-precision local clock.

[0077] Preferably, step S6 includes:

[0078] Each field adaptive networking node obtains the master-slave time deviation after each master-slave synchronization or the mutual slave time deviation after mutual synchronization.

[0079] If the master-slave time deviation or mutual slave time deviation exceeds the deviation compensation threshold, the dual compensation model is triggered to perform dynamic compensation.

[0080] The dual-compensation model performs dynamic compensation including:

[0081] Pre-set Kalman filter parameters are applied based on the type of electromagnetic interference signal, the intensity level of the electromagnetic interference signal, and the frequency band in which the electromagnetic interference signal is located.

[0082] The time deviation for electromagnetic interference compensation is obtained by performing Kalman filtering on the master-slave time deviation or mutual-slave time deviation based on the loaded Kalman filter parameters. ;

[0083] The drift model, based on synchronization period and drift rate estimation, predicts the time deviation caused by the local clock crystal of the field adaptive networking node. ;

[0084]

[0085] in, This represents the drift rate estimate, with an initial value of 0; Indicates the synchronization period;

[0086] Time deviation based on electromagnetic interference compensation Time deviation caused by the local clock crystal oscillator of the field adaptive networking node Calculate the total compensation deviation ;

[0087] Based on total compensation deviation A high-precision local clock is obtained by compensating the local clock of the adaptive networking nodes in the field. .

[0088] In this embodiment, taking a field adaptive networking node as an example, its microcontroller unit (MCU) reads the time deviation value generated during each PTP master-slave synchronization or mutual synchronization after completion. Assuming the configured deviation compensation threshold is 0.1μs, and the measured master-slave time deviation is +0.25μs (greater than the threshold), the MCU triggers the dual compensation model for dynamic compensation.

[0089] The MCU first transmits the interference sensing results (interference type: narrowband wireless interference, intensity level: 3, frequency band: 2.4GHz) to the FPGA interference and drift compensation hardware unit via the SPI bus. The FPGA has multiple pre-set Kalman filter parameter tables. For the combination of "narrowband wireless interference, intensity level 3, 2.4GHz band", the corresponding observation noise covariance R and process noise covariance Q are selected. Then, the FPGA uses the currently measured time deviation z(k) as input to perform Kalman filter iterations, obtaining the electromagnetic interference compensation time deviation after filtering. Meanwhile, the FPGA obtains the current drift rate estimate from the MCU. Synchronization period Seconds; calculate the time deviation increment caused by crystal oscillator drift. Total compensation deviation FPGA will The signal is sent to the PTPv2 synchronization chip, which reduces the local clock speed by adjusting the phase accumulator of the local OCXO from the clock source. After adjustment, the MCU measures the residual deviation between the local clock and the master clock again. The result is less than 0.1μs, indicating that the compensation is effective. After compensation, the MCU records the original deviation, the residual deviation after compensation, and the interference parameters used in the internal log. If the residual deviation remains large (e.g., greater than 0.15μs) in subsequent cycles, the MCU will mark the drift rate estimate as potentially too low and wait for remote feedback to trigger a drift rate update. The entire dynamic compensation process is completed entirely by FPGA hardware, with a response time of less than 1μs, ensuring that the node clock always maintains a synchronization accuracy within ±0.5μs with the master clock.

[0090] S7: Each field adaptive networking node uses a sparse representation signal decomposition algorithm based on an overcomplete dictionary to remove interference data in the partial discharge mixed signal according to the signal separation algorithm threshold. Then, it uses a bandpass filter to remove residual noise outside the frequency band to obtain high signal-to-noise ratio pure partial discharge data. Based on a high-precision local clock, it generates a high signal-to-noise ratio pure partial discharge data packet with a high-precision timestamp and transmits it to the edge aggregation gateway through a wireless Mesh self-organizing network.

[0091] In this embodiment, the partial discharge acquisition submodule of the field adaptive networking node acquires a 10μs cable signal at a sampling rate of 200MS / s, resulting in a mixed signal vector y with 2000 sampling points. This signal contains a partial discharge pulse (amplitude approximately 20mV, width approximately 50ns) and strong narrowband interference (frequency approximately 1.5GHz, amplitude approximately 5 times that of the partial discharge signal). The node edge signal processing unit (FPGA) internally has a pre-built overcomplete dictionary D, which contains 4000 atoms. It contains 1000 simulated partial discharge pulse atoms with different widths and attenuation coefficients. It contains 3,000 typical interference waveform atoms (including sine waves, pulse trains, comb spectra, etc. of different frequencies).

[0092] FPGA executes Orthogonal Matching Pursuit (OMP) algorithm: Initialize residuals Support set Set the threshold for the signal separation algorithm. (This represents the maximum allowable residual energy). During the iteration process, each time, the atom with the largest absolute value of the inner product with the current residual is selected from the dictionary D, its index is added to the support set, and then the coefficients are updated using the least squares method, and the new residual is calculated. After 28 iterations, the residual energy drops to 0.048, which is less than the threshold ϵ, and the iteration stops. At this point, the support set contains 17 interfering atoms and 11 partial discharge pulse atoms.

[0093] FPGA will be with The corresponding interfering atom coefficients are cleared to zero, only those related to... The corresponding coefficients are then used to reconstruct the pure partial discharge signal. In the reconstructed signal, the interference component was suppressed by approximately 28 dB. Next, the FPGA... Bandpass filtering was performed using a 64th-order FIR bandpass filter with a passband frequency range of 30MHz to 3GHz and a stopband attenuation of 60dB. The filtered result yielded high signal-to-noise ratio (SNR) clean partial discharge data. The signal-to-noise ratio was improved from 5dB to 35dB in the original mixed signal.

[0094] The FPGA reads the current high-precision local clock from the satellite-free high-precision clock synchronization unit. Then The data packet is bound to this timestamp and encapsulated with a node ID and CRC checksum field. The FPGA transmits the data packet directly to the adaptive networking transmission unit via a parallel bus, bypassing the MCU to reduce latency. The adaptive networking transmission unit uses its currently optimized operating frequency band (e.g., 1.4GHz), omnidirectional antenna, and bandwidth (50Mbps) to immediately upload the data packet to the edge aggregation gateway via a wireless mesh self-organizing network. The entire processing latency from signal acquisition to data packet transmission is less than 5ms.

[0095] S8: The edge aggregation gateway transmits high signal-to-noise ratio (SNR) clean partial discharge (PD) data packets with high-precision timestamps to the remote monitoring center via optical fiber. The remote monitoring center generates PD type identification results based on the high SNR clean PD data packets with high-precision timestamps and generates system operation statistics.

[0096] Preferably, the remote monitoring center calls a pre-trained deep learning classification model to identify the defect type of the high signal-to-noise ratio clean partial discharge data in each data packet, wherein the defect type includes: internal discharge, surface discharge, corona discharge, floating discharge, and no discharge;

[0097] In the same time window High signal-to-noise ratio (SNR) clean partial discharge (PD) data with the same defect type are determined to come from the same PD pulse, and the location of the field adaptive networking node corresponding to the same PD pulse and the timestamp of the high SNR clean PD data are located.

[0098] Using the location of the adaptive network nodes in the field as a reference, the estimated location of the partial discharge pulse is obtained by solving the following least squares problem. ;

[0099]

[0100] in, Indicates the first The number of field adaptive networking nodes corresponding to the secondary partial discharge pulse; Indicates the first The first partial discharge pulse corresponds to the first The location of each field adaptive networking node; Indicates the first The first partial discharge pulse corresponds to the first The location of each field adaptive networking node; Indicates the speed of electromagnetic wave propagation; Indicates the first The first partial discharge pulse corresponds to the first Timestamps of high signal-to-noise ratio clean partial discharge data from each field adaptive networking node; Indicates the first The first partial discharge pulse corresponds to the first Timestamps of high signal-to-noise ratio clean partial discharge data from each field adaptive networking node; Indicates the location of the partial discharge pulse; the first Each field adaptive networking node is the node with the smallest timestamp;

[0101] If the partial discharge pulse is of the type of internal discharge or surface discharge, and the amplitude of the partial discharge pulse is greater than the warning threshold, the remote monitoring center will generate a warning message and display it on the monitoring terminal.

[0102] Preferably, the statistical indicators for the generation system include:

[0103] The remote monitoring center uses a fixed time window Calculate the following statistical indicators using units:

[0104] Data transmission success rate :

[0105]

[0106] in, The number of high signal-to-noise ratio clean partial discharge data packets actually received and correctly parsed by the remote monitoring center; The total number of high signal-to-noise ratio clean partial discharge data packets that the remote monitoring center expects to receive;

[0107] Average synchronization accuracy :

[0108]

[0109]

[0110] in, This indicates the local time of the edge aggregation gateway when it receives a high signal-to-noise ratio (SNR) clean partial discharge data packet. This time is automatically added by the edge aggregation gateway when it receives the high SNR clean partial discharge data packet. This indicates the timestamp added by the adaptive networking node when generating high signal-to-noise ratio clean partial discharge data packets; This indicates the transmission latency between the field adaptive networking nodes and the edge aggregation gateway;

[0111] Root mean square error of partial discharge positioning :

[0112]

[0113]

[0114] Where K represents the number of partial discharge pulses; Indicates the first The exact location of the secondary partial discharge pulse was obtained through manual investigation of actual fault points. This represents the positioning error of the k-th partial discharge pulse; Indicates the first Estimated location of the secondary partial discharge pulse.

[0115] In this embodiment, the edge aggregation gateway continuously transmits high signal-to-noise ratio (SNR) clean partial discharge (PD) data packets with high-precision timestamps uploaded by each node to the remote monitoring center via optical fiber. The PD monitoring platform software on the industrial server receives the data packets in real time; each data packet contains a node ID, timestamp, and PD waveform data. A pre-trained deep learning classification model (e.g., a one-dimensional convolutional neural network) is loaded into the server memory. For the waveform data in each data packet, the model outputs the probability of each category; the category with the highest probability is taken as the recognition result, and the confidence level is recorded. A confidence level higher than a preset threshold is considered valid.

[0116] The server maintains a time window. When multiple data packets from different nodes are received within the same window, and all of them are identified as having an "internal discharge" defect type with a confidence level higher than 0.8, these data packets are determined to belong to the same partial discharge pulse. Assume this event involves four nodes with the following coordinates (in meters): Node A (100,0,2), Node B (150,0,2), Node C (180,0,2), and Node D (220,0,2), with corresponding timestamps: , , , (Based on a certain reference time). Using node A with the smallest timestamp as a reference, construct the TDOA equation system. Solve using least squares to obtain the estimated location of the partial discharge source. The partial discharge was identified as "internal discharge" (a high-risk defect) and the waveform amplitude (peak value extracted was 25 pC) exceeded the warning threshold (the warning threshold for 220kV cables is 5 pC). The server displayed a warning window on the monitoring terminal screen, showing "Time: 2025-03-01 15:23:45.123, Location: Approximately 155 meters, Type: Internal Discharge, Amplitude: 25 pC, Immediate inspection recommended." This warning event was also recorded in the database.

[0117] The system generates operational statistics, such as the following metrics within a 24-hour time window: data transmission success rate, root mean square error of partial discharge location, and average improvement in signal-to-noise ratio. All metrics, along with their timestamps, are stored in the database and updated in real time to the monitoring terminal display. These metrics constitute the system's operational statistics.

[0118] S9: The remote monitoring center transmits the system operation statistics to the edge aggregation gateway via fiber optic cable, and the edge aggregation gateway transmits them to each field adaptive networking node via a wireless mesh self-organizing network; each field adaptive networking node adaptively updates the parameters of the networking parameter mapping model and the dual compensation model according to the system operation statistics.

[0119] Preferably, the adaptive updating of the network parameter mapping model and the dual compensation model based on system operation statistics includes:

[0120] Based on data transmission success rate Construct a loss function and update the parameters of the BP neural network network parameter mapping model through backpropagation gradient updates:

[0121]

[0122] Based on average synchronization accuracy and the root mean square error of partial discharge positioning Update the drift rate estimation parameters of the drift model:

[0123]

[0124] in, This represents the drift rate estimation parameters of the drift model before the update; For adaptive step size; and Indicates the weighting parameter; Indicates the preset target precision; Indicates the maximum permissible positioning error; Indicates the feedback cycle.

[0125] In this embodiment, the remote monitoring center packages system operation statistics into a feedback command every hour. This command is transmitted via fiber optic cable to the fiber optic transmission module of the edge aggregation gateway, parsed by the industrial Ethernet module, and forwarded to the multi-channel networking receiving module. The gateway broadcasts the command to all field adaptive networking nodes via wireless mesh self-organizing network. After receiving the feedback command, the adaptive networking transmission unit of each node transmits it to the microcontroller unit (MCU) via SPI bus. The MCU parses the various statistical indicators and performs model parameter updates accordingly: updating the networking parameter mapping model (BP neural network); the BP neural network maintained internally by the MCU has its original weight matrix and bias vector trained offline to obtain initial values. Based on the data transmission success rate, the MCU calculates the loss function and performs an online update using gradient descent: learning rate η=0.01, calculates the output layer error and backpropagates it, and updates all connection weights and biases.

[0126] Update the crystal drift rate in the dual-compensation model: The MCU reads the current estimated crystal drift rate and updates the drift rate estimation parameters of the drift model. The MCU writes the updated drift rate to the FPGA interference and drift compensation hardware unit via SPI. After the update is complete, the MCU stores the new model parameters in local non-volatile memory and records the update time. Waiting for the next 24-hour feedback cycle, it receives statistical indicators from the remote monitoring center again and repeats the above update process. Through this closed-loop feedback, the network parameter mapping model and the dual-compensation model can adapt to changes in the tunnel environment and crystal aging over a long period, ensuring the system maintains high-performance operation.

[0127] In summary, this invention significantly improves transmission reliability by collecting electromagnetic interference characteristics and dynamically predicting optimal network parameters using a BP neural network, achieving adaptive adjustment of frequency bands, antennas, and bandwidth, as well as wireless mesh network reconstruction. Through a dual synchronization network combining PTP master-slave synchronization and LoRa inter-node synchronization, and utilizing FPGA hardware-accelerated Kalman filtering and crystal drift compensation models, high-precision clock synchronization at the microsecond level (≤±0.5μs) is achieved even without satellite signals. Simultaneously, a sparse representation signal separation algorithm based on an overcomplete dictionary removes interference components from the partial discharge mixed signal in real time at the node end. Combined with bandpass filtering optimization, this significantly improves the signal-to-noise ratio of the partial discharge signal and reduces wireless bandwidth usage. Finally, system operation statistics generated by the remote monitoring center are fed back to each node, achieving closed-loop adaptive optimization of the network parameter mapping model and the dual compensation model. This results in high signal-to-noise ratio, microsecond-level synchronization, high success rate, and low-latency transmission of partial discharge monitoring data even without satellite signals, providing reliable technical support for the accurate location and early warning of partial discharge defects in cable tunnels.

[0128] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An adaptive anti-interference synchronous transmission method for partial discharge monitoring in cable tunnels, characterized in that, include: S1: Distribute multiple field adaptive networking nodes on cable supports or cable bodies within the cable tunnel; deploy edge aggregation gateways at the entrances and exits of the cable tunnel; The remote monitoring center is deployed in a remote monitoring room; the wireless communication coverage of adjacent field adaptive networking nodes overlaps with each other. S2: After each field adaptive networking node is powered on, it automatically loads the factory default parameters and establishes an initial wireless Mesh self-organizing network based on the default parameters; S3: The user generates a configuration command at the remote monitoring center and sends the configuration command to the edge aggregation gateway through fiber optic cable. The edge aggregation gateway broadcasts the configuration command to all field adaptive networking nodes through the initial wireless mesh self-organizing network. Each field adaptive networking node completes the configuration according to the command. S4: The edge aggregation gateway receives satellite timing signals to calibrate its local clock; each field adaptive networking node calibrates its local clock through PTP master-slave synchronization based on the local clock of the edge aggregation gateway, and at the same time, each field adaptive networking node performs mutual synchronization calibration of its local clock. S5: Each field adaptive networking node collects electromagnetic interference data and partial discharge mixed signals from the cable tunnel; based on the electromagnetic interference data and partial discharge mixed signals, the optimal networking parameters are predicted through the networking parameter mapping model; each field adaptive networking node adjusts the networking parameters according to the optimal networking parameters and re-establishes the wireless Mesh self-organizing network. S6: Each field adaptive networking node uses a dual compensation model to dynamically synchronize and compensate the local clock for high-precision clock based on the interference perception results, thus obtaining a high-precision local clock. S7: Each field adaptive networking node uses a sparse representation signal decomposition algorithm based on an overcomplete dictionary to remove interference data in the partial discharge mixed signal according to the signal separation algorithm threshold. Then, it uses a bandpass filter to remove residual noise outside the frequency band to obtain high signal-to-noise ratio pure partial discharge data. Based on a high-precision local clock, it generates a high signal-to-noise ratio pure partial discharge data packet with a high-precision timestamp and transmits it to the edge aggregation gateway through a wireless Mesh self-organizing network. S8: The edge aggregation gateway transmits high signal-to-noise ratio (SNR) clean partial discharge (PD) data packets with high-precision timestamps to the remote monitoring center via optical fiber. The remote monitoring center generates PD type identification results based on the high SNR clean PD data packets with high-precision timestamps and generates system operation statistics. S9: The remote monitoring center transmits the system operation statistics to the edge aggregation gateway via fiber optic cable, and the edge aggregation gateway transmits them to each field adaptive networking node via a wireless mesh self-organizing network; each field adaptive networking node adaptively updates the parameters of the networking parameter mapping model and the dual compensation model according to the system operation statistics.

2. The adaptive anti-interference synchronous transmission method for partial discharge monitoring in cable tunnels according to claim 1, characterized in that, The configuration instructions include: partial discharge acquisition sampling rate, interference sensing scanning frequency band, initial synchronization parameters, and signal separation algorithm threshold; wherein, the initial synchronization parameters include PTP master-slave synchronization period, inter-node mutual synchronization period, and deviation compensation threshold.

3. The adaptive anti-interference synchronous transmission method for partial discharge monitoring in cable tunnels according to claim 1, characterized in that, The field adaptive networking node is powered by a power module consisting of an inductive power collection coil, a lithium battery, and a power management chip. The inductive power collection coil is sleeved on a cable to collect electrical energy. The power management chip is used to convert the electrical energy collected by the inductive power collection coil and store it in the lithium battery, and to convert the electrical energy output by the lithium battery and use it for power supply.

4. The adaptive anti-interference synchronous transmission method for partial discharge monitoring in cable tunnels according to claim 2, characterized in that, The field adaptive networking node is configured with an adaptive networking transmission unit and an inter-node synchronization communication module. The adaptive networking transmission unit includes: a dual-band radio frequency module, an antenna switching module, multiple types of antennas, and a bandwidth scheduling module. The dual-band radio frequency module is used for switching between 1.4GHz and 2.4GHz dual bands. The multiple types of antennas include directional antennas and omnidirectional antennas. The antenna switching module is used for automatic switching between directional antennas and omnidirectional antennas. The bandwidth scheduling module is used for adaptive allocation of multiple bandwidths. The inter-node synchronization communication module is used to broadcast the local clock information of the field adaptive networking node according to the inter-node synchronization period via 433MHz LoRa narrowband communication.

5. The adaptive anti-interference synchronous transmission method for partial discharge monitoring in cable tunnels according to claim 4, characterized in that, Step S4 includes: the edge aggregation gateway receives satellite timing signals, calculates the UTC reference time, calibrates the local clock according to the UTC reference time, and uses it as the master reference clock; it obtains the master reference time from the master reference clock, generates PTP synchronization messages according to the PTP master-slave synchronization cycle, and broadcasts the PTP synchronization messages to each field adaptive networking node through the wireless Mesh self-organizing network; each field adaptive networking node actively broadcasts local clock information through the inter-node mutual synchronization communication module according to the inter-node mutual synchronization cycle. Each field adaptive networking node receives the PTP synchronization message, parses the master reference time in the PTP synchronization message, compares the master reference time with the local clock time, calculates the master-slave time deviation, and calibrates the local clock according to the master-slave time deviation to complete the local clock calibration of each field adaptive networking node. If a field adaptive networking node does not receive a PTP synchronization message from the edge aggregation gateway for three consecutive PTP master-slave synchronization cycles, the field adaptive networking node extracts a timestamp from the local clock information of the adjacent field adaptive networking nodes, compares the extracted timestamp with the local clock time of the field adaptive networking node, calculates the mutual slave time deviation, and adjusts the local clock according to the mutual slave time deviation to complete the local clock calibration of each field adaptive networking node.

6. The adaptive anti-interference synchronous transmission method for partial discharge monitoring in cable tunnels according to claim 5, characterized in that, Step S5 includes: Each on-site adaptive networking node scans the electromagnetic interference signal in the cable tunnel in real time using a miniature spectrum analyzer based on the interference sensing scanning frequency band, and obtains the spectral characteristics, time domain characteristics and frequency band of the electromagnetic interference signal; the interference intensity value of the electromagnetic interference signal is collected by the electromagnetic interference intensity detection chip, and the interference intensity value of the electromagnetic interference signal is the average power value of the electromagnetic interference signal. Each field adaptive networking node uses a random forest machine learning classification algorithm to identify the interference category of electromagnetic interference signals based on the spectral and temporal characteristics of the electromagnetic interference signals. The interference categories of electromagnetic interference signals include harmonic interference, narrowband wireless interference, pulse interference, or spurious radiation interference. Each on-site adaptive networking node identifies the interference intensity level of the electromagnetic interference signal according to the interference intensity value and a preset threshold, where: ≤-80dBm is level 1, -80~-70dBm is level 2, -70~-60dBm is level 3, -60~-50dBm is level 4, and ≥-50dBm is level 5. Each field adaptive networking node acquires the partial discharge mixed signal of the cable through a high-frequency pulse current sensor and an ultra-high frequency sensor according to the partial discharge acquisition sampling rate, and extracts the amplitude of the partial discharge mixed signal; Each field adaptive networking node uses a BP neural network networking parameter mapping model to predict the optimal networking parameters based on the interference type, interference intensity level, frequency band of the interference signal, and amplitude of the partial discharge mixed signal. The optimal networking parameters include communication frequency band, antenna type, and transmission bandwidth. Each on-site adaptive networking node adjusts the communication frequency band, antenna type, and transmission bandwidth of the adaptive networking transmission unit according to the optimal networking parameters, and re-establishes the wireless Mesh self-organizing network.

7. The adaptive anti-interference synchronous transmission method for partial discharge monitoring in cable tunnels according to claim 6, characterized in that, Step S6 includes: Each field adaptive networking node obtains the master-slave time deviation after each master-slave synchronization or the mutual slave time deviation after mutual synchronization. If the master-slave time deviation or mutual slave time deviation exceeds the deviation compensation threshold, the dual compensation model is triggered to perform dynamic compensation. The dual-compensation model performs dynamic compensation including: Pre-set Kalman filter parameters are applied based on the type of electromagnetic interference signal, the intensity level of the electromagnetic interference signal, and the frequency band in which the electromagnetic interference signal is located. The time deviation for electromagnetic interference compensation is obtained by performing Kalman filtering on the master-slave time deviation or mutual-slave time deviation based on the loaded Kalman filter parameters. ; The drift model, based on synchronization period and drift rate estimation, predicts the time deviation caused by the local clock crystal of the field adaptive networking node. ; in, This represents the drift rate estimate, with an initial value of 0; Indicates the synchronization period; Time deviation based on electromagnetic interference compensation Time deviation caused by the local clock crystal oscillator of the field adaptive networking node Calculate the total compensation deviation ; Based on total compensation deviation A high-precision local clock is obtained by compensating the local clock of the field adaptive networking nodes. .

8. The adaptive anti-interference synchronous transmission method for partial discharge monitoring in cable tunnels according to claim 7, characterized in that, The remote monitoring center calls a pre-trained deep learning classification model to identify the defect type of the high signal-to-noise ratio clean partial discharge data in each data packet. The defect types include: internal discharge, surface discharge, corona discharge, floating discharge, and no discharge. In the same time window High signal-to-noise ratio (SNR) clean partial discharge (PD) data with the same defect type are determined to come from the same PD pulse, and the location of the field adaptive networking node corresponding to the same PD pulse and the timestamp of the high SNR clean PD data are located. Using the location of the adaptive network nodes in the field as a reference, the estimated location of the partial discharge pulse is obtained by solving the following least squares problem. ; in, Indicates the first The number of field adaptive networking nodes corresponding to the secondary partial discharge pulse; Indicates the first The first partial discharge pulse corresponds to the first The location of each field adaptive networking node; Indicates the first The first partial discharge pulse corresponds to the first The location of each field adaptive networking node; Indicates the speed of electromagnetic wave propagation; Indicates the first The first partial discharge pulse corresponds to the first Timestamps of high signal-to-noise ratio clean partial discharge data from each field adaptive networking node; Indicates the first The first partial discharge pulse corresponds to the first Timestamps of high signal-to-noise ratio clean partial discharge data from each field adaptive networking node; Indicates the location of the partial discharge pulse; the first Each field adaptive networking node is the node with the smallest timestamp; If the partial discharge pulse is of the type of internal discharge or surface discharge, and the amplitude of the partial discharge pulse is greater than the warning threshold, the remote monitoring center will generate a warning message and display it on the monitoring terminal.

9. The adaptive anti-interference synchronous transmission method for partial discharge monitoring in cable tunnels according to claim 8, characterized in that, The statistical indicators for the operation of the generation system include: The remote monitoring center uses a fixed time window Calculate the following statistical indicators using units: Data transmission success rate : in, The number of high signal-to-noise ratio clean partial discharge data packets actually received and correctly parsed by the remote monitoring center; The total number of high signal-to-noise ratio clean partial discharge data packets that the remote monitoring center expects to receive; Average synchronization accuracy : in, This indicates the local time of the edge aggregation gateway when it receives a high signal-to-noise ratio (SNR) clean partial discharge data packet. This time is automatically added by the edge aggregation gateway when it receives the high SNR clean partial discharge data packet. This indicates the timestamp added by the adaptive networking node when generating high signal-to-noise ratio clean partial discharge data packets; This indicates the transmission latency between the field adaptive networking nodes and the edge aggregation gateway; Root mean square error of partial discharge positioning : Where K represents the number of partial discharge pulses; Indicates the first The exact location of the secondary partial discharge pulse was obtained through manual investigation of actual fault points. This represents the positioning error of the k-th partial discharge pulse; Indicates the first Estimated location of the secondary partial discharge pulse.

10. The adaptive anti-interference synchronous transmission method for partial discharge monitoring in cable tunnels according to claim 9, characterized in that, The parameters of the adaptive update of the network parameter mapping model and the dual compensation model based on system operation statistics include: Based on data transmission success rate Construct a loss function and update the parameters of the BP neural network network parameter mapping model through backpropagation gradient updates: Based on average synchronization accuracy and the root mean square error of partial discharge positioning Update the drift rate estimation parameters of the drift model: in, This represents the drift rate estimation parameters of the drift model before the update; For adaptive step size; and Indicates the weighting parameter; Indicates the preset target precision; Indicates the maximum permissible positioning error; Indicates the feedback cycle.

Images