A joint debugging device and method with a 4G communication interface

Abstract

The present application relates to power distribution automation terminal technical field, specifically, it relates to a kind of commissioning device and method with 4G communication interface, comprising: 4G communication unit;Data processing unit;Multi-interface unit;Self-adapting protocol conversion unit is used to automatically identify and convert power distribution terminal communication protocol, based on artificial intelligence technology collection message feature and generate analysis rule base, with multi-interface unit real-time interaction, and support the dynamic update of protocol library.The present application constructs multiple-terminal parallel control architecture, realizes the automation distribution of debugging task and the centralized processing of test result, effectively solves the problem of low efficiency of batch scheduling under traditional single debugging mode, improves the execution efficiency of power distribution terminal batch debugging;The present application dynamically learns message feature and generates analysis rule base by the artificial intelligence algorithm of self-adapting protocol conversion unit, without manual maintenance protocol library, can automatically identify and convert the communication protocol of multiple manufacturers terminal.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power distribution automation terminal, in particular to a joint debugging device with a 4G communication interface and a method. BACKGROUND

[0002] As the core equipment of safe operation of distribution network, the debugging quality of power distribution automation terminal directly affects the system collaborative control efficiency. With the large-scale deployment of distribution network terminal, the traditional debugging mode exposes significant short board: relying on manual test one by one, lacking multi-terminal parallel scheduling ability; laboratory simulation test and field 4G / 5G network, main station actual working condition are out of touch, and the collaborative performance of "main station-terminal-network" cannot be verified; protocol analysis relies on static library, which is difficult to adapt to the communication differences of terminals of multiple manufacturers, resulting in low batch debugging efficiency and poor system adaptability, which becomes the key restriction of efficiency improvement of distribution network operation and maintenance.

[0003] The existing power distribution terminal debugging technology exposes multi-dimensional defects such as control architecture, protocol adaptation and communication reliability when facing large-scale deployment demand: the traditional debugging scheme is mostly designed for single terminal, and the multi-terminal parallel control mechanism is not established, resulting in low efficiency of task allocation and result aggregation in batch debugging; the protocol analysis module relies on static rule library, lacks self-learning ability for new terminal communication protocol, and is difficult to dynamically adapt to the communication differences of devices of multiple manufacturers; the communication link does not design recovery mechanism for actual working conditions such as field 4G network instantaneous interruption and signal fluctuation, and message loss or transmission interruption is easy to cause debugging process interruption. The above problems form a systematic bottleneck in efficiency, compatibility and reliability of batch debugging of power distribution terminal. In view of this, we propose a joint debugging device with a 4G communication interface and a method. SUMMARY

[0004] The purpose of the present application is to provide a joint debugging device with a 4G communication interface and a method to solve the problems of low efficiency of task allocation and result aggregation in batch debugging and easy interruption of debugging process caused by message loss or transmission interruption in the background art.

[0005] To solve the above technical problems, one of the purposes of the present application is to provide a joint debugging device with a 4G communication interface, which comprises:

[0006] 4G communication unit, the 4G communication unit is used for realizing remote communication with the power distribution master station, adapting the network frequency band based on the dual-mode communication module, ensuring the stability of data transmission through signal enhancement technology, and transmitting encrypted debugging data with the data processing unit in both directions;

[0007] A data processing unit is configured to perform edge computing and local storage on power distribution terminal data, run a device state prediction model based on a hybrid architecture of a processor and an FPGA, interact with a 4G communication unit through a bus, and cache debugging data of the last N days;

[0008] A multi-interface unit is configured to connect different types of power distribution terminals, adapt to various communication protocols based on protocol conversion technology, ensure transmission stability through signal isolation technology, and identify and adapt active or passive switch input;

[0009] An adaptive protocol conversion unit is configured to automatically identify and convert the communication protocol of a power distribution terminal, collect message features based on artificial intelligence technology and generate an analysis rule library, interact with the multi-interface unit in real time, and support dynamic updating of the protocol library;

[0010] A dynamic load balancing unit is configured to optimize 4G network transmission efficiency, predict traffic trends based on an LSTM neural network, allocate transmission channels through real-time load dynamic weight algorithms, interact with the 4G communication unit in real time using an API interface to load data, and trigger a breakpoint resume process when communication is interrupted through an interrupt detection circuit.

[0011] As a further improvement of the technical solution, the dual-mode communication module includes a frequency band identification submodule and a dynamic switching submodule, wherein:

[0012] The frequency band identification submodule is configured to collect the frequency, bandwidth, and signal strength parameters of the 4G signal in real time, and identify the current network frequency band type based on a feature matching algorithm;

[0013] The dynamic switching submodule is in communication connection with the frequency band identification submodule, and when it detects that the current frequency band signal strength is lower than a preset threshold or the packet loss rate exceeds a preset threshold, it starts scanning adjacent frequency bands and selects the frequency band with the best signal quality through a multi-objective optimization algorithm.

[0014] As a further improvement of the technical solution, the signal enhancement technology specifically includes:

[0015] A 4x4 MIMO intelligent antenna array is used to improve spatial diversity gain, and an automatic gain control circuit is used to dynamically adjust the signal amplification multiple to ensure the stability of the signal amplitude input to the data processing unit;

[0016] The 4x4 MIMO intelligent antenna array uses beamforming technology to adaptively adjust the phase and amplitude of each antenna array through a digital signal processing algorithm to enhance the signal reception strength in the target direction;

[0017] The automatic gain control circuit automatically switches the gain mode according to the real-time received signal power, improves the front-end amplification gain in a weak signal environment, and reduces the gain in a strong signal environment to avoid signal saturation, thereby forming a signal strength optimization mechanism in a full dynamic range.

[0018] As a further improvement of the technical solution, the processor with a mixed architecture of FPGA includes a control scheduling module and an edge computing module, wherein:

[0019] The control scheduling module is used for edge computing task scheduling, parameter updating of the device state prediction model, and data interaction with the 4G communication unit.

[0020] The edge computing module is based on FPGA hardware and is used for pre-processing, feature extraction, and inference calculation acceleration of the device state prediction model of the power distribution terminal data.

[0021] As a further improvement of the technical solution, the process of the data processing unit for edge computing and local storage of the power distribution terminal data includes the following steps:

[0022] S230.1, performing format conversion, outlier filtering, and normalization processing on the received power distribution terminal data through the edge computing module;

[0023] S230.2, extracting time-domain features and frequency-domain features of voltage, current, and power factor from the pre-processed data based on the edge computing module, inputting them into the device state prediction model for inference, and outputting the device state prediction result;

[0024] S230.3, caching the pre-processed data, feature data, and prediction result to the local storage device through the control scheduling module, wherein the local storage device supports timestamp partition storage, and the caching duration is N days;

[0025] S230.4, when the preset synchronization condition is met, the control scheduling module synchronizes the cached data to the power distribution master station through the 4G communication unit.

[0026] The preset synchronization condition specifically includes the following embodiments:

[0027] Data volume trigger: the cached data capacity of the local storage device reaches a preset threshold (such as 80% of the storage capacity according to the storage resource and business demand configuration);

[0028] Time trigger: the time interval from the last synchronization of data to the power distribution master station exceeds a preset duration (such as 24 hours, which can be flexibly adjusted through a system configuration interface);

[0029] Instruction trigger: the control scheduling module receives a data request instruction issued by the power distribution master station through the 4G communication unit.

[0030] As a further improvement of the technical solution, the multi-interface unit comprises a protocol adaptation module, a signal isolation module and a switch quantity identification module, wherein:

[0031] The protocol adaptation module is based on protocol conversion technology to adapt the communication protocols of different types of power distribution terminals;

[0032] The signal isolation module ensures the stability of data transmission between the power distribution terminal and the data processing unit through signal isolation technology;

[0033] The switch quantity identification module is used to identify and adapt the active or passive switch quantity input of the power distribution terminal.

[0034] As a further improvement of the technical solution, the process of identifying active or passive switch quantity input by the switch quantity identification module includes the following steps:

[0035] S330.1, collect the real-time voltage signal of the switch quantity input port of the power distribution terminal through the switch quantity identification module;

[0036] S330.2, convert the collected analog voltage signal to digital quantity and compare it with the voltage discrimination threshold value built-in the switch quantity identification module;

[0037] S330.3, if the digital quantity is greater than or equal to the threshold value, it is determined to be an active switch quantity, and the light coupling isolation active signal conditioning circuit is triggered to output the conditioned signal to the data processing unit;

[0038] S330.4, if the digital quantity is less than the voltage discrimination threshold value, it is determined to be a passive switch quantity, and the pull-up resistor network passive signal conditioning circuit is triggered to output the conditioned digital signal to the data processing unit.

[0039] As a further improvement of the technical solution, the process of automatically identifying and converting the communication protocol of the power distribution terminal by the adaptive protocol conversion unit includes the following steps:

[0040] S400.1, based on the sliding window and multi-scale feature fusion algorithm, the collected power distribution terminal original message is segmented and processed to extract the following features:

[0041] Message length feature: statistics of message length distribution in fixed time window, calculation of mean , variance , entropy value ;

[0042] ;

[0043] ;

[0044] ;

[0045] where, is the byte length of the th message; is the total number of messages collected within the time window; is the probability of a message with length appearing within the window; is the number of different message lengths within the window;

[0046] Field position feature: identify the positions of frame header , function code , address field , data field , and check field in the message and calculate their proportions;

[0047] ;

[0048] where, is the byte proportion of field in the message; is the function field in the message, including:

[0049] frame header : a fixed byte sequence indicating the start of the message (e.g., 0xAA 0x55 for Modbus protocol);

[0050] function code : a byte indicating the operation type of the message (e.g., 0x03 for Modbus to read registers);

[0051] address field : a byte indicating the device address (e.g., slave address for Modbus);

[0052] data field : a byte segment storing specific business data;

[0053] check field : a byte for error detection (e.g., CRC16, LRC check value).

[0054] Timing feature: extract the time interval distribution , frequency feature of adjacent messages;

[0055] ;

[0056] ;

[0057] where, is the time interval between the th and the The receiving time interval of the packet; The receiving time stamp of the first packet ;

[0058] S400.2, input the extracted features into a lightweight feature enhancement network, and optimize the recognition efficiency based on feature grouping processing and correlation screening:

[0059] Feature grouping processing layer:

[0060] The multi-scale features are divided into length feature groups and position feature groups according to types, and each group is independently mapped;

[0061]

[0062] ;

[0063] Wherein, , is the feature vector after grouping mapping; , is a lightweight multilayer perceptron;

[0064] Feature correlation screening layer:

[0065] Calculate the mutual information between features When the mutual information is less than the preset mutual information threshold , filter the low correlation feature pair;

[0066] ;

[0067] ;

[0068] Wherein, is the first , feature in the fused feature vector; is the joint probability distribution of the feature ; is the mutual information of the feature ; is the maximum value of the mutual information of all feature pairs;

[0069] S400.3, based on the recognition result, generate an analysis rule library through Q-learning reinforcement learning:

[0070] ;

[0071] Wherein, is the expected cumulative reward for taking action in state ; is a state of reinforcement learning; is an action of reinforcement learning; is an action performed is a next state transferred after the action; is a state is an action performed is an immediate reward obtained; is a learning rate; is a discount factor; is a next state is all possible actions in the next state is the maximum Q value in

[0072] S400.4, when the identification probability is less than a preset identification probability threshold , it is determined that a new protocol is detected, triggering incremental learning to update the model:

[0073] ;

[0074] wherein, is a learnable parameter of the lightweight feature enhancement network; is a gradient learning rate; is a loss function is the gradient of the parameter ; is a cross-entropy loss function; is a newly collected unknown protocol message feature set; is a protocol type label of the new sample.

[0075] As a further improvement of the technical solution, the process of optimizing the transmission efficiency of the 4G network by the dynamic load balancing unit includes the following steps:

[0076] S500.1, based on the LSTM neural network, input the historical traffic sequence and timestamp features, and predict the future transmission traffic trend:

[0077] ;

[0078] wherein, is the predicted traffic at the future moment ; represents a historical traffic window with a length of ; is the timestamp encoding at the moment ;

[0079] S500.2, through a real-time load dynamic weight algorithm, calculate the channel priority weight by comprehensively considering bandwidth, delay, and packet loss rate, and allocate transmission channels:

[0080] ;

[0081] wherein, is the actual available bandwidth of the channel; is the maximum supported bandwidth of the channel; is the real-time delay of the channel; is the maximum allowed delay; is the real-time packet loss rate of the channel; is the maximum allowed packet loss rate; , , is the weight coefficient, and ;

[0082] S500.3, upload the channel load data (including predicted traffic , real-time weight ) to the 4G communication unit (100) by the RESTful API interface at the frequency F, and the data format is JSON:

[0083] ;

[0084] wherein, is the interaction data body, including the predicted traffic and the channel weight set;

[0085] S500.4, the interruption detection circuit monitors the 4G signal receiving strength and the data transmission timeout time , when any of the following conditions is met, it is determined that the communication is interrupted and the breakpoint resume is triggered:

[0086] Signal strength condition: , wherein is the preset signal strength threshold;

[0087] Timeout condition: , wherein, is the preset timeout threshold;

[0088] Resume the transmission before the interruption according to the data block number and the checksum , and the resume data structure is:

[0089] ;

[0090] wherein, is the data block number; checksum of the th data block; content of the th data block; is the retransmission data structure; Actual service data carried in the structure.

[0091] The second object of the present application is to provide a commissioning method with a 4G communication interface, based on the commissioning device with a 4G communication interface, comprising the following steps:

[0092] S100, establishing a 4G remote communication link: identifying network frequency bands and dynamically switching through the 4G communication unit, optimizing transmission stability combined with signal enhancement technology, and establishing an encrypted communication connection with the power distribution master station;

[0093] S200, power distribution data edge computing and storage: using the data processing unit to preprocess, feature extract and state predict the power distribution terminal data, and caching the data according to the rules and synchronizing to the master station;

[0094] S300, multi-interface protocol adaptation and signal conditioning: adapting the communication protocols of different power distribution terminals through the multi-interface unit, isolating the transmission signal and identifying and processing active and passive switch quantity inputs;

[0095] S400, communication protocol adaptive conversion: based on the adaptive protocol conversion unit, extracting message features, generating an analysis rule library through machine learning, realizing protocol automatic identification and dynamic updating;

[0096] S500, 4G network load balancing and breakpoint resume: using the dynamic load balancing unit to predict traffic trends, allocate transmission channels, monitor signal interruptions and trigger data resume mechanism.

[0097] Compared with the prior art, the present application has the following advantages:

[0098] 1. The present application realizes the automation of debugging task allocation and centralized processing of test results by constructing a multi-terminal parallel control architecture, effectively solves the problem of low batch scheduling efficiency in the traditional single debugging mode, and improves the execution efficiency of power distribution terminal batch debugging;

[0099] 2. In the present application, the artificial intelligence algorithm of the adaptive protocol conversion unit dynamically learns message features and generates an analysis rule library, without manual maintenance of the protocol library, which can automatically identify and convert the communication protocols of terminals from multiple manufacturers, effectively improving the compatibility of the system for different types of power distribution terminals;

[0100] 3. The present application optimizes the design of 4G communication link, integrates instantaneous interruption detection and breakpoint resume mechanism, monitors signal strength in real time and dynamically triggers message retransmission strategy, effectively guarantees the continuity of data transmission during debugging, reduces the risk of debugging interruption caused by network fluctuations, and improves the reliability of on-site debugging. BRIEF DESCRIPTION OF DRAWINGS

[0101] Figure 1Structure diagram of the joint debugging device with 4G communication interface of the present application;

[0102] Figure 2 Flow chart of the joint debugging method with 4G communication interface of the present application;

[0103] The meanings of various labels in the figure are as follows:

[0104] 100, 4G communication unit; 110, dual-mode communication module; 111, frequency band identification submodule; 112, dynamic switching submodule;

[0105] 200, data processing unit; 210, control scheduling module; 220, edge computing module;

[0106] 300, multi-interface unit; 310, protocol adaptation module; 320, signal isolation module; 330, switch quantity identification module;

[0107] 400, adaptive protocol conversion unit;

[0108] 500, dynamic load balancing unit. DETAILED DESCRIPTION

[0109] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0110] Embodiment 1

[0111] As shown in the figure, the present embodiment provides a joint debugging device with 4G communication interface, comprising: Figure 1

[0112] 4G communication unit 100, 4G communication unit 100 is used to realize remote communication with power distribution master station, based on dual-mode communication module 110 to adapt network frequency band, through signal enhancement technology to guarantee data transmission stability, and bidirectional transmission of encrypted debugging data with data processing unit 200;

[0113] In this step, the dual-mode communication module 110 includes a frequency band identification submodule 111 and a dynamic switching submodule 112, wherein:

[0114] The frequency band identification submodule 111 is used to collect the frequency, bandwidth and signal strength parameters of the 4G signal in real time, and identify the current network frequency band type based on a feature matching algorithm;

[0115] ​The dynamic switching sub-module 112 is in communication connection with the frequency band identification sub-module 111, and when it is detected that the current frequency band signal strength is lower than a preset threshold or the packet loss rate exceeds a preset threshold, the adjacent frequency band scanning is started, and the multi-target optimization algorithm is used to select the frequency band with the optimal signal quality.

[0116] In this step, the signal enhancement technology specifically includes:

[0117] The 4x4 MIMO intelligent antenna array is adopted to improve the spatial diversity gain, and the automatic gain control circuit is combined to dynamically adjust the signal amplification multiple, so as to ensure that the signal amplitude input to the data processing unit 200 is stable.

[0118] Specifically, the 4x4 MIMO antenna array adopts equal-interval linear arrangement (such as an interval of 1 / 2 of the wavelength of the current frequency band center frequency), and the signal is enhanced through the following process:

[0119] Channel estimation: periodically send Zadoff-Chu training sequence (common channel estimation sequence in wireless communication, with autocorrelation), and the receiving end constructs a 4x4 channel matrix according to the feedback signal, and analyzes the signal attenuation and phase offset of each antenna pair;

[0120] Beamforming: based on the channel matrix, the phase and amplitude of the antenna array are dynamically adjusted through a digital algorithm, so that the radiation pattern is focused on the strongest signal direction, and the receiving strength of the target area is improved;

[0121] Scene adaptation: automatically switch to "high gain mode" in weak signal scenes such as underground power distribution rooms, balance coverage and power consumption in open areas, and adapt to different environmental communication needs.

[0122] The 4x4 MIMO intelligent antenna array adopts beamforming technology, and the phase and amplitude of each antenna array are adaptively adjusted through a digital signal processing algorithm to enhance the signal receiving strength in the target direction;

[0123] The automatic gain control circuit automatically switches the gain mode according to the real-time received signal power, increases the front-end amplification gain in a weak signal environment, and reduces the gain in a strong signal environment to avoid signal saturation, forming a full-dynamic-range signal strength optimization mechanism.

[0124] Specifically, the AGC circuit has three gain modes (weak signal 31dB, medium signal 10dB, and strong signal-11dB), and the adjustment process is as follows:

[0125] Signal monitoring: the peak detector samples the input amplitude at a period of 1μs, and when the signal changes across the interval (such as from-60dBm to-30dBm), the gain switching is triggered;

[0126] Smooth transition: the gain is adjusted by a ramp function at a slope of 0.5dB / μs to avoid baseband abnormalities caused by signal mutations.

[0127] Mode adaptation: According to the signal mean and variance statistics, automatically determine the intensity interval and switch the gain mode, covering the full dynamic signal range.

[0128] Further, the embodiment also integrates forward error correction (FEC) and automatic repeat (ARQ) protocol to deal with signal instantaneous interruption, specifically including:

[0129] Instantaneous interruption detection: Real-time monitoring of RSSI, when continuously 500ms below -100dBm, it is determined as signal instantaneous interruption;

[0130] Data protection: The debugging data is divided into 1024 bytes, and CRC check code is added; When instantaneous interruption occurs, mark the interruption position and temporarily store the untransmitted data block;

[0131] Breakpoint resume: After signal recovery (such as RSSI≥-95dBm and lasts for 100ms), resume from the interruption position, verify the integrity through the check code, and trigger ARQ retransmission for missing blocks.

[0132] As a further description of this step, in this embodiment, the frequency band recognition submodule 111 realizes real-time classification of 4G network frequency bands using a feature matching algorithm based on pattern recognition principles. By comparing the collected signal parameters with the pre-set frequency band feature library, the automatic recognition of frequency band types is realized. The specific process is as follows:

[0133] First, the frequency band recognition submodule 111 collects radio frequency signal parameters in real time through the ADC circuit of the baseband chip with a period of 100ms (configurable), including center frequency, signal bandwidth, signal strength (RSSI) and fluctuation coefficient (ratio of RSSI standard deviation to average value), and performs sliding window filtering (window size 10) on the collected data to reduce noise interference;

[0134] Subsequently, the pre-processed parameters are integrated into a four-dimensional feature vector, and the built-in pre-set frequency band feature library is called. This library stores the reference feature parameters of the domestic mainstream 4G frequency bands (such as LTE-800, LTE-1800) based on the operator's public specification configuration;

[0135] Then, the feature matching algorithm calculates the similarity between the to-be-identified feature vector and each reference frequency band in the feature library through weighted comparison, focusing on matching the center frequency and bandwidth parameters (assigning higher weights), and combining signal strength and fluctuation coefficient for comprehensive judgment, selecting the most similar frequency band as the recognition result. For example, when the center frequency deviates within ±5MHz and the bandwidth is consistent, it is preferentially determined as the frequency band;

[0136] Then, the system has a dynamic learning mechanism: if the uncollected frequency band is identified for 5 times in succession, the incremental learning process will be triggered, the new frequency band feature parameters are collected and compared with the existing library, if the difference exceeds the preset threshold (such as the center frequency deviation > 10 MHz), the new feature is automatically added to the library, realizing the self-adaptive compatibility of the new network frequency band.

[0137] As a further description of this step, the 4G communication unit 100 in this embodiment and the data processing unit 200 adopt a hybrid mechanism of "asymmetric key negotiation + symmetric encryption transmission", combined with priority scheduling and adaptive strategy to ensure data security and stable transmission: when the system starts, the public key is exchanged through RSA asymmetric encryption (2048-bit key), the 128-bit AES session key is negotiated based on the Diffie-Hellman protocol, and the RSA channel is updated automatically every 100 transmissions (or every 1 hour); the debugging data is encrypted in AES-CBC mode, a 16-byte initialization vector (IV) is randomly generated to attach the encrypted header, and an HMAC-SHA256 check code (based on the AES key) is calculated to attach the tail; large-size data is divided into blocks of 1024 bytes and marked with priority (remote control command high priority, telemetry data ordinary priority); when transmitting, the queue is scheduled according to priority, high-priority data occupies the channel first, adaptive timeout retransmission is used (initial 500ms, dynamically modified according to round-trip time RTT, retransmission timeout doubles and ≤8000ms), when the packet loss rate exceeds 5% for 5 times in a row (or RSSI <-95dBm), it is automatically downgraded from LTE Cat6 to Cat4 mode, if it is still abnormal, the frequency band identification submodule is scanned to switch to a better frequency band.

[0138] The data processing unit 200 is used for edge computing and local storage of power distribution terminal data, runs a device state prediction model based on a processor and FPGA hybrid architecture, interacts with the 4G communication unit 100 through a bus, and caches the debugging data of the last N days;

[0139] In this step, the processor and FPGA hybrid architecture includes a control scheduling module 210 and an edge computing module 220, wherein:

[0140] The control scheduling module 210 is used for edge computing task scheduling, parameter updating of the device state prediction model, and data interaction with the 4G communication unit 100;

[0141] The edge computing module 220 is based on FPGA hardware, and is used for pre-processing, feature extraction, and inference calculation acceleration of the device state prediction model of the power distribution terminal data.

[0142] In this step, the process of the data processing unit 200 for edge computing and local storage of power distribution terminal data includes the following steps:

[0143] S230.1, format conversion, outlier filtering and normalization processing of the received power distribution terminal data by the edge computing module 220;

[0144] In this step, the edge computing module 220 first converts the ASCII code data uploaded by the terminal into floating point binary values (including unit mapping, such as kV→V); filters abnormal values by the 3σ principle; and maps the electrical parameters to the ([0,1]) interval by Min-Max normalization to ensure data consistency.

[0145] S230.2, based on the edge computing module 220, extracting time domain features and frequency domain features of voltage, current and power factor from preprocessed data, inputting to the device state prediction model for inference, and outputting the device state prediction result;

[0146] In this step, time domain features (mean, variance, peak-valley difference of voltage, current and power factor) and frequency domain features (fundamental frequency and harmonic proportion extracted by FFT fast Fourier transform, FPGA end with pipeline architecture to accelerate the transformation process, output 12-dimensional feature vector) are extracted from the preprocessed data; the device state prediction model uses a random forest algorithm (multiple decision trees for parallel inference, FPGA end realizes parallel judgment of tree nodes through hardware logic), and finally outputs the device state prediction result (such as health degree classification).

[0147] S230.3, caching preprocessed data, feature data and prediction results to a local storage device by the control scheduling module 210, the local storage device supports timestamp partition storage, and the cache duration is N days;

[0148] In this step, the cache duration N can be configured to an integer from 1 to 30 days (default 7 days) according to storage resources and business needs, and can be flexibly adjusted through a local interactive interface or a remote JSON instruction; when the remaining storage capacity is lower than a threshold value, a batch debugging task is detected or a power distribution master station instruction is received, the system automatically adjusts the N value (such as halving when capacity warning, temporarily setting to 3 days when task starts), to ensure effective caching and storage of debugging data and balance of storage resources, and related configurations can be read through hardware registers and updated in real time through instruction interfaces.

[0149] S230.4, when the preset synchronization condition is met, the control scheduling module 210 synchronizes the cached data to the power distribution master station through the 4G communication unit 100.

[0150] In this step, the preset synchronization condition includes the following embodiments:

[0151] Data volume trigger: the cache data capacity of the local storage device reaches a preset threshold (such as 80% of the storage capacity according to storage resources and business needs).

[0152] Time trigger: the time interval from the last synchronization data to the power distribution master station exceeds the preset time length (such as 24 hours, which can be flexibly adjusted through the system configuration interface);

[0153] Instruction trigger: the control scheduling module 210 receives the data request instruction issued by the power distribution master station through the 4G communication unit 100.

[0154] As a further description of this step, the data processing unit 200 in this embodiment realizes edge computing and local storage of power distribution terminal data using a hybrid architecture of processor and FPGA, specifically including:

[0155] The ARMCortex-A72 processor carries the control scheduling logic (such as task allocation, model parameter update, and master station interaction), and the XilinxSpartan-7 FPGA accelerates the computationally intensive tasks (such as data preprocessing, feature extraction, and model inference). The two achieve low-latency data interaction through the AXI bus (transmission rate 500 Mbps); wherein:

[0156] The control scheduling module 210 allocates computing resources based on the polling scheduling algorithm (preferentially guaranteeing the device state prediction task), and receives the model parameter update instruction (such as the number of decision trees and learning rate adjustment) of the power distribution master station through the 4G communication unit every hour;

[0157] The edge computing module 220 relies on the parallel operation capability of the FPGA to perform three-level acceleration processing on the power distribution terminal data.

[0158] The multi-interface unit 300 is used to connect different types of power distribution terminals, adapt to multiple communication protocols based on protocol conversion technology, ensure transmission stability through signal isolation technology, and identify and adapt active or passive switching value inputs;

[0159] In this step, the multi-interface unit 300 includes a protocol adaptation module 310, a signal isolation module 320, and a switching value identification module 330, wherein:

[0160] The protocol adaptation module 310 adapts the communication protocols of different types of power distribution terminals based on protocol conversion technology;

[0161] In this embodiment, when the multi-protocol address domains overlap (such as IEC60870-5-104 and Modbus address conflict), the following priority arbitration is performed:

[0162] Protocol priority: protection class protocol (such as IEC61850-9-2) > control class protocol (such as IEC60870-5-104) > monitoring class protocol (such as Modbus);

[0163] Port priority: when the protocol, the target port is 2404 (Modbus) or 2408 (104) message is parsed first;

[0164] Timestamp priority: when the first two are the same, parse the latest message with timestamp (error ≤10ms is considered to be repeated, discarded directly).

[0165] The above rules are configured through a hardware register (address 0x4000_0000), supporting remote upgrade.

[0166] The signal isolation module 320 ensures the stability of data transmission between the power distribution terminal and the data processing unit 200 through signal isolation technology.

[0167] The switch quantity identification module 330 is used to identify and adapt the active or passive switch quantity input of the power distribution terminal.

[0168] In this step, the process of identifying active or passive switch quantity input by the switch quantity identification module 330 includes the following steps:

[0169] S330.1, collect the real-time voltage signal of the switch quantity input port of the power distribution terminal through the switch quantity identification module 330;

[0170] S330.2, convert the collected analog voltage signal to digital quantity, and compare it with the voltage discrimination threshold value built-in the switch quantity identification module 330;

[0171] S330.3, if the digital quantity ≥ threshold value, it is determined as active switch quantity, and the active signal conditioning circuit of optical coupling isolation is triggered, and the conditioned signal is output to the data processing unit 200;

[0172] S330.4, if the digital quantity < voltage discrimination threshold value, it is determined as passive switch quantity, and the passive signal conditioning circuit of pull-up resistor network is triggered, and the conditioned digital signal is output to the data processing unit 200.

[0173] As a further description of this step, the multi-interface unit 300 in this embodiment adopts an integrated architecture of "protocol conversion chip + isolator + signal conditioning circuit", which realizes the multi-type interface adaptation of the power distribution terminal. The specific hardware selection is as follows:

[0174] The protocol adaptation module 310 selects STM32F407 microcontroller (main frequency 168MHz, built-in CAN, RS-485, ModbusRTU hardware protocol stack), which supports more than 10 kinds of industrial communication protocols (such as IEC60870-5-104, DL / T645-2007);

[0175] The signal isolation module 320 adopts ADuM1201 digital isolator (isolation voltage 2500Vrms, data rate 10Mbps) to realize strong and weak electric physical isolation in cooperation with DC-DC isolation power supply.

[0176] The switch quantity identification module 330 is based on TLV2374 operational amplifier (bandwidth 2.2MHz, used for voltage signal conditioning) and CD4051 analog multiplexer (8-channel input selection) to support simultaneous monitoring of 8-way switch quantity signals.

[0177] Further, in view of the noise coupling problem of ADuM1201 high-speed isolation and DC-DC isolation power supply, the following design is made:

[0178] Hierarchical voltage stabilization: a low-noise LDO (such as TPS7A4700, output noise ≤30μVrms) is connected in series at the output end of the DC-DC isolation power supply to suppress switching noise;

[0179] Layout isolation: the power supply layer of the digital isolation circuit is physically separated from the power supply layer of the main circuit, connected through a π-type filter (10μF electrolytic capacitor + 0.1μF ceramic capacitor + 10μH inductor) to reduce crosstalk;

[0180] Timing matching: an RC delay circuit (10kΩ resistor + 100nF capacitor) is connected in series at the enable end of ADuM1201 to ensure that the isolator is started after the power supply is stable (200μs after power-on).

[0181] As a further description of this step, the protocol adaptation module 310 in this embodiment receives the original message of the power distribution terminal (such as ModbusRTU, DL / T645 format), relies on the hardware protocol stack to analyze the data frame header, address field and function code, and extracts the effective data such as register value; then re-encapsulates the message according to the interface protocol (such as IEC60870-5-104) of the data processing unit 200, and guarantees data integrity through CRC16 check (polynomial 0x8005) in the conversion process. The signal isolation module 320 builds a double isolation barrier: the digital signal is electrically isolated through ADuM1201 digital isolator to block ground loop interference; the power supply is isolated through DC-DC isolator (such as B0505LS-1W model) to reduce ripple effect; through the synergistic effect of the two, ground loop interference (common-mode rejection ratio ≥60dB) and surge impact (surge resistance up to 4kV, meeting IEC61000-4-5 standard) can be effectively suppressed, and the stability of data transmission between the power distribution terminal and the data processing unit can be ensured.

[0182] The adaptive protocol conversion unit 400 is used to automatically identify and convert the communication protocol of the power distribution terminal. It collects message features based on artificial intelligence technology and generates a parsing rule base. It interacts with the multi-interface unit 300 in real time and supports dynamic updates of the protocol base.

[0183] In this step, the adaptive protocol conversion unit 400 automatically identifies and converts the power distribution terminal communication protocol, including the following steps:

[0184] S400.1. Based on the sliding window and multi-scale feature fusion algorithm, the collected original messages from the power distribution terminal are segmented and the following features are extracted:

[0185] Message length characteristics: Statistically analyze the distribution of message lengths within a fixed time window and calculate the mean. ,variance Entropy ;

[0186] ;

[0187] ;

[0188] ;

[0189] in, For the first The length of each message in bytes; This represents the total number of messages collected within the time window. For length is The probability of a message appearing within the window; This refers to the number of different message length types within the window;

[0190] Field location features: Identifying frame headers in messages Function codes Address field Data domain , verification field Location, and calculate its proportion;

[0191] ;

[0192] in, For fields The percentage of bytes in the message; The functional fields in the message include:

[0193] Frame header : A fixed sequence of bytes that identifies the beginning of a message (e.g., 0xAA0x55 in the Modbus protocol);

[0194] Function code : Byte indicating the operation type of the message (e.g. 0x03 of Modbus means reading registers);

[0195] Address field : Byte identifying the device address (e.g. slave address of Modbus);

[0196] Data field : Byte segment storing specific business data;

[0197] Check field : Byte for error detection (e.g. CRC16, LRC check value).

[0198] Timing feature: Extract the time interval distribution of adjacent messages , frequency feature ;

[0199] ;

[0200] ;

[0201] wherein, is the receiving time interval between the i-th and the j-th message; is the receiving timestamp of the i-th message; S400.2, input the extracted features into a lightweight feature enhancement network, and optimize the recognition efficiency based on feature grouping processing and correlation screening: Feature grouping processing layer:

[0202] The multi-scale features are divided into length feature group

[0203] and position feature group according to types, and each group is independently mapped;

[0204]

[0205]

[0206] ;

[0207] wherein, , are the feature vectors after grouping mapping; , are the lightweight multilayer perceptrons;

[0208] Feature correlation screening layer:

[0209] Calculate the mutual information between features , and when the mutual information is less than a preset mutual information threshold ​​​At this time, filter the low-correlation feature pairs;

[0210] ;

[0211] ;

[0212] wherein, is the i-th feature in the fused feature vector; , ; is the joint probability distribution of the features ; is the mutual information of the features ; is the maximum mutual information of all feature pairs;

[0213] S400.3, based on the recognition result, generating an analysis rule library through Q-learning reinforcement learning:

[0214] ;

[0215] wherein, is the expected cumulative reward for taking action in state ; is the state of reinforcement learning; is the action of reinforcement learning; is the next state transferred after executing action ; is the immediate reward obtained by executing action in state ; is the learning rate; is the discount factor; is the maximum Q value among all possible actions in the next state ;

[0216] S400.4, when the recognition probability is less than a preset recognition probability threshold , it is determined that a new protocol is detected, triggering incremental learning to update the model:

[0217] ;

[0218] wherein, is the learnable parameter of the lightweight feature enhancement network; is the gradient learning rate; is the gradient of the loss function on the parameter ; is the cross-entropy loss function; For the newly collected unknown protocol message feature set; For the protocol type label of the new sample.

[0219] As a further description of this step, when the message recognition probability is less than the preset threshold value for 5 consecutive times , the incremental learning process is triggered: first, collect ≥50 new protocol messages, extract the feature set and manually label the protocol type , then update the network parameters using the stochastic gradient descent (SGD) algorithm, and the update formula is ; After updating, a new rule library version (such as v1.3) is generated, and the new and old versions run in parallel for 72 hours, and the new rule accuracy is verified ≥0.9 before being officially used. For example, when identifying the IEC60870-5-101 protocol, the initial recognition probability is 0.6, which triggers incremental learning, and after adding the rule "frame header = 68H, function code = 7th byte", the recognition probability is improved.

[0220] The dynamic load balancing unit 500 is used to optimize the 4G network transmission efficiency, predict the traffic trend based on the LSTM neural network, allocate the transmission channel through the real-time load dynamic weight algorithm, and use the API interface to interact with the 4G communication unit 100 in real time. Load data, and trigger the breakpoint resume process when the communication is interrupted by the interrupt detection circuit.

[0221] In this step, the process of the dynamic load balancing unit 500 optimizing the 4G network transmission efficiency includes the following steps:

[0222] S500.1, based on the LSTM neural network, input the historical traffic sequence and timestamp features, and predict the future transmission traffic trend:

[0223] ;

[0224] Where, is the predicted traffic at future time ; represents a historical traffic window with a length of ; is the timestamp encoding at time ;

[0225] S500.2, through the real-time load dynamic weight algorithm, calculate the channel priority weight by integrating bandwidth, delay, and packet loss rate , and allocate the transmission channel:

[0226] ;

[0227] Where, is the actual available bandwidth of the channel; is the maximum supported bandwidth of the channel; is the real-time delay of the channel; is the maximum allowed delay; is the real-time packet loss rate of the target channel; is the maximum allowed packet loss rate; , , is a weight coefficient, and ;

[0228] In the embodiment, the weight coefficients in the embodiment satisfy α+β+γ=1, and the value ranges are determined according to the service feature constraints: (bandwidth), (delay), (reliability).

[0229] Delay-sensitive services (such as differential protection, requiring end-to-end delay ≤20 ms): ∈[0.3,0.5], ≥0.4.

[0230] Bandwidth-sensitive services (such as video monitoring, requiring transmission bandwidth ≥1 Mbps): ≥0.6, ∈[0.2,0.3].

[0231] Reliability-sensitive services (such as remote signaling sending, requiring packet loss rate ≤0.1%): ≥0.5, + ≤0.5.

[0232] The service type is identified through the protocol type and the target port number (for example, port 5000 corresponds to a video stream, and port 2404 corresponds to a Modbus protocol), and the weight mode is automatically switched. Network QoS indicators are collected every 5 minutes, and if the delay jitter exceeds 10%, the value of α is gradually reduced, and the adjustment step is ≤0.1.

[0233] S500.3, the channel load data (containing predicted traffic , real-time weight ) are uploaded to the 4G communication unit 100 at a frequency F through a RESTful API interface, and the data format is JSON:

[0234] ;

[0235] Among them, is the interactive data body, containing the predicted traffic and the channel weight set;

[0236] S500.4, the 4G signal receiving strength is monitored by the interruption detection circuit​ and data transmission timeout time When any of the following conditions is met, it is determined that the communication is interrupted and the breakpoint resume is triggered:

[0237] Signal strength condition: Wherein is a preset signal strength threshold;

[0238] Timeout condition: Wherein, is a preset timeout threshold;

[0239] According to the data block number and the checksum The transmission before the interruption is resumed, and the resume data structure is:

[0240] ;

[0241] Wherein, is the data block number; the checksum of the th data block; the content of the th data block; is the retransmission data structure; the actual service data carried in the structure.

[0242] As a further description of this step, the dynamic load balancing unit 500 in this embodiment adopts an STM32F407VGT6 processor (ARM Cortex-M4 core, 168MHz frequency) to deploy a single-layer LSTM network (128 memory units), and realizes traffic trend prediction through the following process:

[0243] First, the historical traffic is collected in 1 minute cycles and stored in a ring buffer (capacity 100), and a sliding window with a length of W=5 (which can be configured to 3-10 through register 0x001) is taken as input;

[0244] Then, the timestamp is converted into a 24x60-dimensional hot vector (e.g. 15:45 corresponds to the 945th dimension with a value of 1), which is input into the model together with the traffic sequence;

[0245] Finally, the model uses the latest 1 hour data (12 samples are generated according to 5 minute window slicing) to perform incremental training every hour, uses the Adam optimizer (learning rate 0.001) to update the parameters, and outputs the predicted traffic value for the next 1 minute , which is used for subsequent load allocation strategy.

[0246] As a further illustration of this step, the interrupt detection circuit in this embodiment is composed of an RF signal detector and a hardware timer: real-time monitoring of RSSI, when lower than (default -95dBm, configurable -100 to -90dBm) or transmission timeout , trigger breakpoint resume mechanism. When the instantaneous interruption occurs, the dual-port RAM buffer stores the unsent data block (maximum 1024KB) and records the last successful transmission block number . After the signal recovery ) and lasts for 200ms, resume from +1, each data block is attached with CRC32 checksum, and the error block triggers ARQ retransmission (maximum 5 times), the resume data structure contains block number, checksum and Base64 encoded content.

[0247] It should be noted that the dynamic load balancing unit 500 in this embodiment and the 4G communication unit 100, the data processing unit 200, and the multi-interface unit 300 form a cooperative working system: the multi-interface unit 300 connects different types of power distribution terminals through the protocol adaptation module 310, and interacts data with the data processing unit 200 through the signal isolation module 320; after the data processing unit 200 completes edge computing based on the ARM and FPGA hybrid architecture, the data is transmitted to the 4G communication unit 100 through the AXI bus; the dynamic load balancing unit 500 sends the traffic trend and real-time load weight strategy based on LSTM prediction to the 4G communication unit 100 through the RESTful API interface; when the 4G signal instantaneous interruption is monitored, the breakpoint resume mechanism is triggered through the interrupt detection circuit, and the local cache data of the data processing unit 200 is cooperated to recover the transmission, forming a closed loop link of "terminal access-data processing-wireless transmission-load optimization", realizing protocol adaptation, edge computing and efficient transmission of power distribution terminal data on 4G network.

[0248] Embodiment 2

[0249] As shown in Figure 2 , the embodiment also provides a joint debugging method with a 4G communication interface, based on the joint debugging device with a 4G communication interface described in embodiment 1, including the following steps:

[0250] S100, establishing a 4G remote communication link: identifying network frequency bands and dynamically switching through the 4G communication unit 100, optimizing transmission stability combined with signal enhancement technology, and establishing an encrypted communication connection with the power distribution master station;

[0251] S200, power distribution data edge computing and storage: using the data processing unit 200 to preprocess, feature extract and state predict the power distribution terminal data, and caching the data according to the rules and synchronizing to the master station;

[0252] S300, multi-interface protocol adaptation and signal conditioning: adapt the communication protocols of different power distribution terminals through the multi-interface unit 300, isolate the transmission signals and identify and process active and passive switch quantity inputs;

[0253] S400, communication protocol adaptive conversion: based on the adaptive protocol conversion unit 400, extract the message features, generate the analysis rule library through machine learning, realize protocol automatic identification and dynamic update;

[0254] S500, 4G network load balancing and breakpoint resume: use the dynamic load balancing unit 500 to predict traffic trends, allocate transmission channels, monitor signal interruptions and trigger data resume mechanisms.

[0255] Those of ordinary skill in the art can understand that the processes for implementing all or part of the steps of the above embodiments can be completed by hardware, or by programs instructing relevant hardware.

[0256] The above shows and describes the basic principles, main features and advantages of the present application. Those skilled in the art should understand that the present application is not limited to the above embodiments, and the above embodiments and descriptions in the specification are only preferred examples of the present application and are not intended to limit the present application. Without departing from the spirit and scope of the present application, various changes and improvements can be made to the present application, and these changes and improvements all fall within the scope of the claimed present application. The scope of protection of the present application is defined by the appended claims and their equivalents.

Claims

1. A joint debugging device with a 4G communication interface, characterized in that, The utility model relates to a kind of power distribution terminal communication device, including: 4G communication unit (100), the 4G communication unit (100) is used to realize remote communication with power distribution master station, based on dual-mode communication module (110) network frequency band adaptation, through signal enhancement technology guarantee data transmission stability, and with data processing unit (200) two-way transmission encrypted debugging data; Data processing unit (200), the data processing unit (200) is used for edge computing and local storage to power distribution terminal data, based on processor and FPGA hybrid architecture running equipment state prediction model, through bus and 4G communication unit (100) interact data, and buffer recent N days debugging data; Multi-interface unit (300), the multi-interface unit (300) is used to connect different types of power distribution terminal, based on protocol conversion technology adapts to a variety of communication protocols, through signal isolation technology guarantee transmission stability, and identify and adapt active or passive switch quantity input; Adaptive protocol conversion unit (400), the adaptive protocol conversion unit (400) is used to automatically identify and convert power distribution terminal communication protocol, based on artificial intelligence technology message features are collected and generate analysis rule base, with multi-interface unit (300) real-time interaction, and support dynamic update of protocol library; The process that the adaptive protocol conversion unit (400) automatically identifies and converts power distribution terminal communication protocol includes the following steps: S400.1, based on sliding window and multi-scale feature fusion algorithm, the original message of power distribution terminal collected is segmented, and the following features are extracted: Packet length feature: statistics the packet length distribution in a fixed time window, calculate the mean , variance , entropy value ; Field position feature: identify the frame header in the packet , function code , address field , data field , check field Position and calculate its proportion; Timing feature: extract the time interval distribution of adjacent packets , frequency feature ; S400.2, the extracted features are input into lightweight feature enhancement network, based on feature grouping processing + correlation screening optimization identification efficiency: Feature grouping processing layer: The multi-scale features are divided into length feature groups and position feature groups by type, each group being independently mapped; Feature correlation screening layer: Computing mutual information between features When the mutual information is less than a preset mutual information threshold filtering the pair of low-correlation features S400.3, based on identification result, generate analysis rule base through Q-learning reinforcement learning: ； wherein, is a state takes an action with a desired cumulative reward; is a state for reinforcement learning; is an action for reinforcement learning; is an action is a next state transitioned to after the action is performed; is a state is an action is an immediate reward obtained after the action is performed in the state is a learning rate; is a discount factor; is a next state is a maximum Q-value among all possible actions in the next state is a maximum Q-value among all possible actions in the next state S400.4、When the identification probability is less than the preset identification probability threshold determine that a new protocol is detected, trigger incremental learning to update the model: ； wherein, are learnable parameters of the lightweight feature enhancement network; is a gradient learning rate; is a loss function is a gradient of the parameter ; is a cross-entropy loss function; is a newly collected unknown protocol message feature set; is a protocol type label of the new sample; Dynamic load balancing unit (500), the dynamic load balancing unit (500) is used to optimize 4G network transmission efficiency, based on LSTM neural network prediction flow trend, through real-time load dynamic weight algorithm distribution transmission channel, using API interface and 4G communication unit (100) real-time interaction load data, and through interrupt detection circuit triggers breakpoint resume process when communication instantaneous outage.

2. The device according to claim 1, wherein, The dual-mode communication module (110) includes frequency band identification submodule (111) and dynamic switching submodule (112), wherein: The frequency band identification submodule (111) is used to real-time acquisition 4G signal frequency, bandwidth and signal strength parameters, based on feature matching algorithm identification current network frequency band type; The dynamic switching submodule (112) is communicatively connected with the frequency band identification submodule (111), when detecting that current frequency band signal strength is lower than preset threshold or packet loss rate exceeds preset threshold, start adjacent frequency scanning, and select the frequency band of optimal signal quality through multi-objective optimization algorithm.

3. The device of claim 1, wherein the device is configured to communicate with a 4G network. The signal enhancement technology specifically includes: 4 × 4 MIMO intelligent antenna array is used to improve spatial diversity gain, combined with automatic gain control circuit dynamic adjustment signal amplification multiple, ensure that the signal amplitude input to data processing unit (200) is stable; The 4*4 MIMO smart antenna array adopts beamforming technology, and the phase and amplitude of each antenna array are adaptively adjusted by a digital signal processing algorithm to enhance the signal receiving strength in the target direction. The automatic gain control circuit automatically switches the gain mode according to the real-time received signal power, improves the front-end amplification gain in a weak signal environment, and reduces the gain in a strong signal environment to avoid signal saturation, thereby forming a signal strength optimization mechanism with full dynamic range.

4. The device of claim 1, wherein the device is configured to communicate with a 4G network. The processor with a mixed architecture of FPGA includes a control scheduling module (210) and an edge computing module (220), wherein: The control scheduling module (210) is used for edge computing task scheduling, parameter updating of a device state prediction model, and data interaction with the 4G communication unit (100); The edge computing module (220) is based on FPGA hardware and is used for pre-processing, feature extraction, and inference calculation acceleration of the power distribution terminal data.

5. The device of claim 4, wherein the device is configured to communicate with the 4G network. 5 The process of the data processing unit (200) for edge computing and local storage of the power distribution terminal data includes the following steps: S230.1, performing format conversion, outlier filtering, and normalization processing on the received power distribution terminal data through the edge computing module (220); S230.2, extracting time-domain features and frequency-domain features of voltage, current, and power factor from the pre-processed data based on the edge computing module (220), inputting them into the device state prediction model for inference, and outputting the device state prediction result; S230.3, caching the pre-processed data, feature data, and prediction result to the local storage device through the control scheduling module (210), wherein the local storage device supports timestamp partition storage, and the cache duration is N days; S230.4, when the preset synchronization condition is met, the control scheduling module (210) synchronizes the cached data to the power distribution master station through the 4G communication unit (100).

6. The device of claim 1, wherein the device is configured to communicate with a 4G network. The multi-interface unit (300) includes a protocol adaptation module (310), a signal isolation module (320), and a switching value identification module (330), wherein: The protocol adaptation module (310) is based on protocol conversion technology and adapts to the communication protocols of different types of power distribution terminals; The signal isolation module (320) ensures the stability of data transmission between the power distribution terminal and the data processing unit (200) through signal isolation technology; The switching value identification module (330) is used to identify and adapt active or passive switching value inputs of the power distribution terminal.

7. The device according to claim 6, wherein, The process of the switching value identification module (330) for identifying active or passive switching value inputs includes the following steps: S330.1, collecting the real-time voltage signal of the switching value input port of the power distribution terminal through the switching value identification module (330); S330.2, converting the collected analog voltage signal into a digital quantity and comparing it with the voltage discrimination threshold value built-in the switching value identification module (330); S330.3, if the digital quantity is greater than or equal to the threshold value, it is determined as an active switching value, and the light-coupled isolated active signal conditioning circuit is triggered to output the conditioned signal to the data processing unit (200). S330.4, if the digital quantity is less than the voltage discrimination threshold, it is determined that there is no passive switching quantity, and the pull-up resistance network type passive signal conditioning circuit is triggered to output the conditioned digital signal to the data processing unit (200).

8. The device of claim 1, wherein the device is configured to communicate with a 4G network. The process of optimizing the transmission efficiency of the 4G network by the dynamic load balancing unit (500) includes the following steps: S500.1, based on the LSTM neural network, input the historical traffic sequence and timestamp features to predict the future transmission traffic trend: ； wherein, is a predicted traffic for a future time instant ; denotes a historical traffic window of length ; is a timestamp encoding of the time instant ; S500.2, Through real-time load dynamic weight algorithm, the bandwidth, delay, packet loss rate are integrated to calculate the channel priority weight , Assign transmission channel: ； wherein, is the actual available bandwidth of the channel; is the maximum supported bandwidth of the channel; is the real-time delay of the channel; is the maximum allowed delay; is the real-time packet loss rate of the channel; is the maximum allowed packet loss rate; , , is a weight coefficient, and ; S500.3, through the RESTful API interface, upload the channel load data to the 4G communication unit (100) at a frequency F, and the data format is JSON: ； wherein, is an interaction data body comprising a set of predicted traffic and channel weights; S500.4, the interruption detection circuit monitors the 4G signal receiving strength and data transmission timeout time When any of the following conditions is met, it is determined that the communication is interrupted and the breakpoint continues transmission is triggered: Signal strength condition: wherein is a preset signal strength threshold value; Time-out condition: wherein, is a pre-set time-out threshold; By data block number With checksum To resume the transmission before the interruption, the resume data structure is: ； wherein, is a data block number; a checksum of the th data block; the content of the th data block; is a retransmission data structure; actual service data carried in the structure.

9. A method for joint debugging of a device with a 4G communication interface, based on the device with a 4G communication interface for joint debugging according to any one of claims 1-8, characterized in that, Including the following steps: S100, establish a 4G remote communication link: identify the network frequency band and dynamically switch through the 4G communication unit (100), optimize the transmission stability combined with signal enhancement technology, and establish an encrypted communication connection with the power distribution master station; S200, power distribution data edge computing and storage: use the data processing unit (200) to preprocess, feature extraction and state prediction of power distribution terminal data, and cache data according to the rules and synchronize to the master station; S300, multi-interface protocol adaptation and signal conditioning: adapt the communication protocol of different power distribution terminals through the multi-interface unit (300), isolate the transmission signal and identify and process active and passive switching quantity input; S400, communication protocol adaptive conversion: based on the adaptive protocol conversion unit (400), extract the message features, generate the analysis rule library through machine learning, realize protocol automatic identification and dynamic update; S500, 4G network load balancing and breakpoint resume: use the dynamic load balancing unit (500) to predict traffic trends, allocate transmission channels, monitor signal interruptions and trigger data resume mechanism.

Images