A service-aware fgOTN heterogeneous link collaborative redundancy protection method

Abstract

This invention relates to the field of network transmission technology, specifically to a service-aware fgOTN heterogeneous link collaborative redundancy protection method, comprising: real-time monitoring of heterogeneous communication links in the distribution network; mapping physical layer parameters to heterogeneous metrics and generating a weighted link quality ranking table; calculating the slope of link metric changes; if preset deterioration conditions are met, calculating the estimated remaining lifetime of the link and the physical alignment time to the target link through time-series extrapolation; executing differentiated handover decisions based on time adequacy: when the estimated remaining lifetime is greater than the physical alignment time, performing lossless handover with frame-level delay alignment; otherwise, performing rapid rescue handover; if the target link is congested after handover, dynamically reducing the bandwidth of low-priority services based on a resource boundary negotiation mechanism. This invention achieves differentiated protection through predictive game theory decisions, realizing the utilization of fgOTN flexibility to ensure the continuity of high-priority services in congested scenarios.

Description

Technical Field

[0001] This invention relates to the field of network transmission technology, specifically to a service-aware fgOTN heterogeneous link collaborative redundancy protection method. Background Technology

[0002] Existing power distribution network or industrial communication protection schemes mainly rely on the traditional 1+1 / 1:N protection mechanism of SDH / OTN or routing switching strategies based on fixed thresholds. However, such technologies generally suffer from the defect of "trigger lag," that is, they are usually only passively triggered to switch over after the detection of optical path signal loss (LOS) or long-term excessive bit error rate. They lack the ability to perceive and predict the gradual deterioration trend of the link, which makes it impossible to avoid data packet loss at the moment of failure. In addition, when dealing with the switching of heterogeneous links such as optical fiber, 5G, and power line carrier, existing technologies lack a comprehensive consideration of the service alignment time and the remaining life of the link. They often adopt a "one-size-fits-all" hard switching mode, which makes it difficult to balance the need to ensure lossless data phase and the need to deal with the high speed of sudden interruption. When backup link resources are scarce, traditional mechanisms lack dynamic negotiation means to link source traffic and pipeline resources, which can easily cause high-priority services to be interrupted again due to switching to congested paths. In heterogeneous communication networks, under scenarios of gradual link deterioration and resource constraints, the existing protection mechanisms lack differentiated switching strategies, making it difficult to balance the timeliness of business data integrity and connection recovery under sudden failures, which is an issue that needs to be addressed. To address this, a service-aware fgOTN heterogeneous link collaborative redundancy protection method is proposed. Summary of the Invention

[0003] This invention aims to provide a service-aware fgOTN heterogeneous link collaborative redundancy protection method. By sensing the link deterioration trend in advance and flexibly selecting the switching mode based on the time window game result, the redundancy protection method can achieve high reliability, low latency and congestion resistance service continuity guarantee in the heterogeneous environment of the distribution network.

[0004] To achieve the above objectives, the present invention provides the following technical solution: A service-aware fgOTN heterogeneous link collaborative redundancy protection method includes: Real-time monitoring of heterogeneous communication links in the power distribution network is performed, and the collected physical layer parameters are mapped to heterogeneous metrics. The metrics are then weighted and synthesized according to the weight of the current service type to generate a link quality ranking table. Based on the link quality ranking table, the slope of the index change of the current link is calculated in real time; if the slope of change meets the preset gradual deterioration condition, the estimated remaining lifetime of the current link under non-sudden interruption is calculated through time series extrapolation, and the physical alignment time required to switch to the optimal target link is calculated. Perform differentiated handover decision based on time adequacy: compare the estimated remaining lifetime with the physical alignment time; if the estimated remaining lifetime is greater than the physical alignment time and there is time adequacy to complete lossless handover, perform lossless handover including frame-level delay alignment; if the estimated remaining lifetime is less than or equal to the physical alignment time, perform rapid rescue handover with bypass alignment logic. After locking onto the optimal target link, if the target link is detected to be in a congested state, during the steady-state maintenance phase after the handover is completed, the bandwidth of low-priority services is dynamically reduced based on the resource boundary negotiation mechanism, and the container channel for high-priority services is adapted.

[0005] Preferably, the real-time monitoring of heterogeneous communication links in the power distribution network specifically includes: A link-layer operation, management, and maintenance protocol is adopted to establish a bidirectional handshake session between the source and destination network elements. Periodic exchanges of detection messages carrying send and receive timestamps are performed to calculate the round-trip delay of the bidirectional link and verify bidirectional connectivity, obtaining connectivity status data. The digital diagnostic monitoring interface of the fiber optic link module is used to read the input optical power value. The spectrum analysis unit of the power line carrier chip is used to measure the power ratio and count the number of transient spikes, obtaining the signal-to-noise ratio and impulse noise. The physical layer measurement report interface of the 5G wireless link module is used to parse the channel state information reference signal, obtaining the reference signal received power.

[0006] Preferably, mapping the collected physical layer parameters to heterogeneous metrics specifically includes: Obtain the overload optical power threshold and the receiving sensitivity threshold, calculate the relative position of the collected optical power value within the range of the sensitivity threshold and the overload optical power threshold, and obtain the optical path health through a linear normalization formula; Using the signal-to-noise ratio value as the base, an initial link quality score is determined. The interference penalty value of the impulse noise count is calculated using a logarithmic function. The interference penalty value is subtracted from the initial link quality score to obtain the transmission reliability. Based on the received power of the reference signal, the modulation and coding strategy index is matched, and combined with the historical spectral efficiency statistics of the base station cell, the available transmission rate of the current air interface physical layer is estimated. The amount of data at the air interface queuing depth is divided by the available transmission rate to calculate the expected queuing delay. The probability that the expected queuing delay meets the service time limit requirement is calculated using the exponential decay model to obtain the delay confidence level.

[0007] Preferably, the step of weighting and synthesizing based on the weights of the current service type to generate a link quality ranking table specifically includes: Parse the Virtual Local Area Network (VLAN) priority tag in the header of the current transmitted service data frame; use the VLAN priority tag as an index key to retrieve the corresponding service weight vector in the preset weight matrix; use connectivity status data to perform bidirectional availability verification on all heterogeneous links, and directly remove links marked as unidirectional failures or interruptions; For the remaining links that have passed the verification, the optical path health, transmission reliability, and latency confidence are set as positive indicators, and the distribution network link latency and bit error rate are set as negative indicators. The positive and negative indicators are weighted algebraically summed using the service weight vector to obtain the comprehensive quality score of each remaining heterogeneous link. The links are then arranged in descending order of the comprehensive quality score to generate a link quality ranking table.

[0008] Preferably, the step of calculating the estimated remaining lifetime of the current link through time-series extrapolation and calculating the physical alignment time required to switch to the optimal target link specifically includes: A multidimensional temporal feature tensor containing transmission reliability slope, delay confidence slope, and historical bit error rate is constructed at the network element edge. The multidimensional temporal feature tensor is input into a preset lightweight long short-term memory network model to perform temporal deduction of the current link state, output the failure probability curve, and calculate the expected remaining lifetime. Using the round-trip delay value in the connectivity state data as a physical reference, calculate the relative delay difference between the optimal target link and the current link; based on the relative delay difference, calculate the physical alignment time required to eliminate the inter-link phase difference and fill the receiver's dejitter buffer queue by caching the target link data.

[0009] Preferably, the execution of the differentiated handover decision based on time adequacy specifically includes: The estimated remaining lifetime is compared with the physical alignment time. If the estimated remaining lifetime is greater than the physical alignment time, it is determined that there is enough time margin to complete the lossless handover. The optimal target link is locked and the frame-level delay alignment procedure is started to maintain the phase synchronization of the data stream during the handover process. If the estimated remaining lifetime is less than or equal to the physical alignment time, lock the optimal target link and force bypass delay alignment logic, and perform a rapid rescue switch in packet loss tolerance mode.

[0010] Preferably, the dynamic reduction of low-priority service bandwidth based on the resource boundary negotiation mechanism specifically includes: The resource elasticity boundary is defined based on the difference between the committed rate and the peak rate of low-priority services, and the maximum transferable bandwidth margin is calculated; the bandwidth gap of high-priority services is matched with the margin to determine the negotiated reduction target value. Send a backpressure signaling message carrying the target value to the low-priority service source end to initiate boundary negotiation, drive the source end to adjust the token bucket parameters to perform traffic shaping to reduce the transmission code rate, and continuously sample the transmission queue depth; When the queue depth is detected to drop to a low threshold and remain stable, and the logical traffic boundary contraction is confirmed to be effective, the lossless bandwidth adjustment protocol is triggered to physically contract the time slot boundary of the flexible optical service unit for low-priority services, and redirect the released time slot resources to high-priority service channels.

[0011] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. By predicting the estimated remaining lifetime of the link through time-series simulation and using this prediction in a game theory approach compared to the physical alignment time, lossless handover with frame-level alignment is prioritized when there is sufficient time margin to ensure zero data loss. In the event of a sudden avalanche failure, a rapid emergency handover is enforced to maintain connectivity. This differentiated strategy precisely resolves the contradiction between data integrity and handover speed in heterogeneous networks, significantly improving transmission reliability.

[0012] 2. By mapping the parameters of different physical media such as optical fiber, power line carrier, and 5G wireless to homogenized optical path health, transmission reliability, and latency confidence, and combining them with service weights to generate a dynamic sorting table, this mechanism effectively avoids misjudgments and frequent oscillations caused by fluctuations in a single physical parameter, ensuring that in complex power distribution network environments, the protection logic can always lock onto the transmission path with the best overall quality.

[0013] 3. By calculating the resource elasticity boundary and establishing a time-series linkage between source-end traffic shaping and pipeline time slot contraction, the OSUflex time slot resources occupied by low-priority services are dynamically released to high-priority services without interrupting physical connections. This achieves seamless bandwidth adaptation in congestion scenarios, greatly improving the survivability of critical services in extreme network environments. Attached Figure Description

[0014] Figure 1 This is a flowchart illustrating the service-aware fgOTN heterogeneous link collaborative redundancy protection method of the present invention. Figure 2 This is a schematic diagram of the heterogeneous metric mapping process in Embodiment 1 of the present invention; Figure 3 This is a schematic diagram of the process of dynamically reducing bandwidth during congestion in Embodiment 1 of the present invention. Detailed Implementation

[0015] 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.

[0016] Please see Figures 1 to 3 This invention provides a service-aware fgOTN heterogeneous link collaborative redundancy protection method, referring to... Figure 1 The diagram below illustrates the process of this invention, and the technical solution is as follows: A service-aware fgOTN heterogeneous link collaborative redundancy protection method includes: Real-time monitoring of heterogeneous communication links in the power distribution network is performed, and the collected physical layer parameters are mapped to heterogeneous metrics. The metrics are then weighted and synthesized according to the weight of the current service type to generate a link quality ranking table. Based on the link quality ranking table, the slope of the index change of the current link is calculated in real time; if the slope of change meets the preset gradual deterioration condition, the estimated remaining lifetime of the current link under non-sudden interruption is calculated through time series extrapolation, and the physical alignment time required to switch to the optimal target link is calculated. Perform differentiated handover decision based on time adequacy: compare the estimated remaining lifetime with the physical alignment time; if the estimated remaining lifetime is greater than the physical alignment time and there is time adequacy to complete lossless handover, perform lossless handover including frame-level delay alignment; if the estimated remaining lifetime is less than or equal to the physical alignment time, perform rapid rescue handover with bypass alignment logic. After locking onto the optimal target link, if the target link is detected to be in a congested state, during the steady-state maintenance phase after the handover is completed, the bandwidth of low-priority services is dynamically reduced based on the resource boundary negotiation mechanism, and the container channel for high-priority services is adapted.

[0017] Example 1 In this embodiment, under the scenario of distribution network automation communication, the system establishes a two-way handshake session based on the link layer operation, management and maintenance protocol between the source and sink network elements of the fgOTN heterogeneous network to realize real-time monitoring of optical fiber, power line carrier and 5G wireless links.

[0018] First, based on the dynamic waiting recovery mechanism of historical trend analysis, the least squares method is used to perform regression analysis on the historical quality score after the physical layer of the link is restored to determine the stability and improvement trend of the link status. When the physical layer of the faulty link is detected to be reconnected, the historical quality score data of the link in the past N detection cycles is immediately traced back; the least squares method is used to perform linear regression fitting on these discrete data points to calculate the trend slope of quality change and the variance of the actual value deviation from the regression line. The slope characterizes the rate of link quality recovery (i.e., the strength of the "improving" trend), while the variance quantifies the jitter of the link signal in the early stages of recovery. The controller dynamically adjusts the duration of the waiting recovery timer based on these two dimensions: if the slope is significantly positive and the variance is below a preset threshold, the link is determined to be "steadily improving," and the WTR duration is automatically shortened to accelerate the switchback; conversely, if the variance is too large or the slope is too flat, the WTR duration is extended to allow time for spatial confirmation. This process continues until the algorithm determines that the link meets the steady convergence condition, at which point the final switchover signal is triggered, switching the service traffic back to the primary link. This solution effectively solves the "ping-pong switching" technical problem caused by physical link parameter oscillations or poor contact under the traditional fixed WTR mechanism. By introducing regression analysis to deeply quantify and predict the link health, it ensures that services can switch back to high-quality links as quickly as possible while significantly reducing the risk of secondary interruptions caused by premature switchback, thus significantly improving the robustness of heterogeneous link collaborative protection.

[0019] Specifically, the source network element sends a detection message containing a 48-bit high-precision transmission timestamp to the destination every 10 milliseconds. The destination network element records the reception timestamp upon receiving the message and sends back a response packet containing its own processing delay. The source processor extracts the four timestamp values ​​from the response packet, performs a difference calculation, deducts the inherent 10-microsecond processing deviation within the destination, calculates the round-trip delay, and outputs connectivity status data in microseconds.

[0020] Furthermore, for the fiber optic communication link in the fgOTN architecture, the system accesses the digital diagnostic monitoring interface inside the optical module through the I2C bus, reads the values ​​of the 16-bit registers with addresses 104 and 105 in real time, and converts them into the absolute value of the input optical power with a range of -40dBm to +8dBm.

[0021] Specifically, the read optical power value is transmitted as an input signal to the optical link attenuation analysis module. This module compares the current power value with the preset negative 25dBm warning threshold and outputs a binary status bit representing the physical layer link health.

[0022] Furthermore, for power line carrier communication links, the spectrum analysis unit inside the carrier chip performs a 1024-point fast Fourier transform on the analog signal with a sampling frequency of 25 MHz to calculate the power spectral density distribution in the operating frequency band from 2 MHz to 12 MHz.

[0023] Specifically, the system counts the number of transient spikes with an amplitude exceeding 500 millivolts within a 20-millisecond power frequency cycle, and inputs the power ratio and transient spike count as feature vectors into a preset 1D-CNN convolutional neural network.

[0024] Furthermore, this 1D-CNN convolutional neural network includes an input layer with an input dimension of 1024. The first convolutional layer is configured with 16 convolutional kernels, with a kernel size of 3 and a stride of 1, for extracting spectral envelope features. The second layer is a 2x2 max pooling layer, used to compress the feature dimension and extract the saliency of impulse noise. The third layer is a fully connected layer containing 64 neurons. After processing with the ReLU activation function, the output layer uses the Softmax function to output the probability components representing the intensity of impulse noise.

[0025] Specifically, for 5G wireless links, the system uses the physical layer measurement report interface of the wireless module to parse the channel state information reference signal in the 5G New Radio frame structure in real time.

[0026] Furthermore, the processor extracts the reference signal received power value from the received measurement report frame. The value is limited to the range of -140dBm to -44dBm, and the measurement resolution is set to 1dB, which serves as the basis for evaluating the spatial attenuation performance of the wireless channel.

[0027] Specifically, the system synchronizes and aligns the round-trip delay obtained from the link layer protocol, the optical power obtained from the digital diagnostic interface, the impulse noise probability output by the convolutional neural network, and the received power parsed from the measurement report to construct a 4×1 heterogeneous link status feature matrix. This status feature matrix serves as the output of the closed-loop data stream and is transmitted in real time to the subsequent path coordination scheduling engine, ensuring that each monitoring dimension is accessed by the redundancy protection algorithm in 32-bit floating-point format, thereby achieving white-box perception and accurate assessment of link status in complex distribution network environments.

[0028] Table 1: Heterogeneous Link Monitoring Parameter Configuration Table

[0029] Service-adaptive parameter settings: Dynamically adjust delay parameters based on current service priority: Relay protection (priority 7): 5 milliseconds, ordinary control services (priority 5): 20 milliseconds, video surveillance (priority 1): 50 milliseconds; By employing in-band management technology to establish a two-way handshake session and combining it with the acquisition of underlying parameters from multiple physical media, this solution can deeply perceive subtle changes in the physical layer of heterogeneous links and the symmetry of two-way connectivity. It overcomes the blind spot of traditional mechanisms that rely solely on the upper-layer logical state, thereby providing accurate and real-time multi-dimensional data support for upper-layer decision-making and effectively improving the sensitivity and accuracy of fault identification.

[0030] Furthermore, the collected raw parameters of the optical path, carrier, and wireless physical layer are input into the metric mapping engine, and transformed into normalized metrics through a specific nonlinear algorithm. Reference Figure 2 This is a flowchart illustrating the mapping of heterogeneous metrics. Specifically, the system acquires the input optical power value fed back in real time from the fiber optic link and sends it to the optical performance evaluation module. This module has a preset lower limit threshold for receiver sensitivity of -28dBm and a preset upper limit threshold for overload optical power of -8dBm. The optical performance evaluation module calculates the difference between the real-time optical power value and the -28dBm sensitivity threshold, and divides this difference by a fixed span of 20dB to obtain the linear position ratio of the optical power within the effective dynamic range. This linear position ratio is multiplied by 100 for normalization and outputs an optical path health score representing the operational margin of the optical path physical layer. This score is limited to an integer range of 0 to 100.

[0031] Furthermore, for power line carrier links, the system extracts the signal-to-noise ratio (SNR) value fed back by the carrier chip as the calculation base for the initial link quality score, directly linearly mapping the SNR within 30 dB to the initial score range of 0 to 100 points; the system calls the impulse noise interference assessment model to perform nonlinear penalty calculation on the number of detected transient spikes. This model uses a logarithmic function with base 10 to perform logarithmic operation on the sum of the number of spikes plus 1.

[0032] Furthermore, the obtained logarithmic value is multiplied by a constant factor of 15 to obtain the interference penalty value generated by the impulse noise, where the constant factor of 15 is a weight parameter pre-set based on the interference intensity sample of the frequency converter in the distribution network. This interference penalty value is subtracted from the initial link quality score, and the final difference is used as the transmission reliability characterizing the data transmission stability of the carrier link under strong interference environment.

[0033] Furthermore, for 5G wireless links, the system inputs the parsed reference signal received power into the link capacity estimation model, which contains a modulation and coding strategy index table with 29 orders. Specifically, the system matches the corresponding modulation order in the index table based on the current reference signal received power, and combines the historical average spectral efficiency statistics of the base station cell over the past 3600 seconds to multiply the current subcarrier bandwidth by the modulation order coefficient to estimate the available transmission rate of the current air interface physical layer.

[0034] Furthermore, the system monitors the air interface queuing depth of the wireless module's transmit buffer in real time, obtains the number of bytes of data currently waiting to be transmitted, divides it by the estimated available transmission rate, and calculates the expected queuing delay in milliseconds.

[0035] Specifically, the expected queuing delay is input into the delay confidence assessment model, which employs an exponential decay algorithm based on the natural logarithm base. The delay confidence assessment model determines the score by calculating the negative power of the natural logarithm base, where the exponent term of the power is the ratio of the expected queuing delay to the maximum tolerable delay of 20 milliseconds preset by the distribution network automation business. As the expected queuing delay gets closer to 20 milliseconds, the value output by the exponential decay model gets closer to 0, and finally outputs a floating-point number between 0 and 1 as the delay confidence score characterizing the deterministic transmission capability of the wireless link.

[0036] Furthermore, the entire mapping process forms a closed-loop data stream. The inputs are optical power, signal-to-noise ratio, spike count, reference signal received power, and queuing depth. After processing by three algorithms—linear normalization, logarithmic penalty, and exponential decay—the outputs are standardized optical path health, transmission reliability, and time delay confidence. These three indicators serve as heterogeneous link metric feature vectors and are transmitted in real time to the subsequent cooperative path decision module to complete the redundancy handover determination of the fgOTN link within a 10-millisecond handover time limit.

[0037] By mapping the differentiated original parameters of heterogeneous physical media such as optical fiber, power line carrier, and 5G wireless to homogeneous standard metrics, the data dimension barriers between heterogeneous networks are resolved through interference penalties and latency probabilities, a unified evaluation benchmark is established, and misjudgments caused by fluctuations in a single physical parameter are avoided, thereby enhancing the objectivity and compatibility of link quality assessment in multimodal network environments.

[0038] A dynamic mapping mechanism between service QoS labels (such as DSCP or 802.1p) and fgOTN OSU (Optical Service Unit) containers: Specifically, the edge access device first performs deep parsing on incoming service packets to extract their QoS priority fields. For high-priority production control services, the system directly maps them to dedicated OSU containers with fixed bandwidth and low latency, establishing rigid pipeline transmission. For low-priority office or browsing services, they are mapped to shared bandwidth statistical multiplexing OSU containers. The system monitors the cache queue depth of each container in real time. When a burst of high-priority services is detected, the system uses fgOTN's lossless adjustment mechanism to dynamically increase the number of time slots in the corresponding OSU containers within milliseconds, borrowing the idle bandwidth of low-priority services. This mapping is not statically configured but is adjusted in real time according to the characteristics of the service flow. It achieves the best balance between bandwidth utilization and critical service SLA assurance. By introducing fine-grained slice awareness below the physical layer, it improves the resource scheduling flexibility of heterogeneous networks when carrying multimodal mixed services.

[0039] Furthermore, the system pre-constructs a service priority mapping matrix based on a two-dimensional data structure in the non-volatile memory of the fgOTN network element control plane. This matrix serves as the core logical benchmark for subsequent execution path optimization and weighted synthesis.

[0040] Specifically, the mapping matrix uses priority tags 0 to 7 as defined by the 802.1Q standard in the Ethernet frame structure as row indices, and horizontally defines column indices corresponding to five physical layer metric parameters: optical path health, transmission reliability, delay confidence, distribution network link delay, and bit error rate. For the 7th level relay protection service indicated by the row index, the system sets the weighting coefficient for delay confidence to 0.35 and the weighting coefficient for distribution network link delay to 0.4 in the corresponding column space, while setting the remaining weighting coefficients for transmission reliability and bit error rate to 0.15 each. The weighting factor is set to 0.1, which makes the sum of the weighting factors representing latency perception performance reach 0.75, thus forming a latency-sensitive configuration row that meets the high real-time requirements of relay protection. For the Level 1 video surveillance service indicated by the row index, the system sets the weighting factor of transmission reliability to 0.45 and the weighting factor of bit error rate to 0.3 in the corresponding column space, while reducing the weighting factors of latency confidence and distribution network link latency to 0.15 and 0.1 respectively, ensuring that the link quality assessment result shifts to the data transmission integrity dimension and forms a throughput-sensitive configuration row.

[0041] Furthermore, the media access control sublayer interface of the network element captures Ethernet data frames entering the fgOTN system in real time, and obtains the virtual LAN priority label of the current service by parsing the priority code point field of bits 13 to 15 in the data frame header. Specifically, the service parsing engine converts the captured 3-bit raw data into a decimal index key value between 0 and 7, and inputs it as the classification feature of the current service flow into the weight query module. The decimal index key value is used to address in the preset weight matrix to extract the service weight vector containing five 32-bit floating-point weights.

[0042] The construction process of the weight matrix includes: establishing a two-dimensional data structure mapping table in the network element control plane; defining the virtual local area network priority label as the row index; and defining the weighting coefficients corresponding to the optical path health, transmission reliability, delay confidence, distribution network link delay, and bit error rate as the column index; for the high-priority relay protection service indicated by the row index, in the corresponding column index, setting the sum of the weighting coefficients of delay confidence and distribution network link delay to be greater than the sum of the weighting coefficients of transmission reliability and bit error rate, forming a delay-sensitive configuration row; for the low-priority video surveillance service indicated by the row index, in the corresponding column index, setting the sum of the weighting coefficients of transmission reliability and bit error rate to be greater than the sum of the weighting coefficients of delay confidence and distribution network link delay, forming a throughput-sensitive configuration row; and storing the delay-sensitive configuration row and the throughput-sensitive configuration row in a fixed manner to obtain the weight matrix.

[0043] The system maps all indicators to a unified scale of 0-100 points: optical path health, transmission reliability, and delay confidence were originally 0-100 points or were converted to 0-100 points through exponential decay. Distribution network link delay score = 100 - min(actual delay / 20ms × 100, 100) Bit error rate score = 100 - min(actual bit error rate / ×100,100) Constraints of the weight vector: w1+w2+w3+w4+w5=1, all wi≥0; corresponding to the weighting coefficients for optical path health, transmission reliability, delay confidence, distribution network link delay and bit error rate, respectively; The weighting of business priorities is configured as follows: Relay protection (priority 7): [0.1, 0.15, 0.35, 0.2, 0.2] (time delay sensitive); Video surveillance (priority 1): [0.15, 0.3, 0.15, 0.1, 0.3] (reliability sensitive); Overall quality score formula: Overall quality score = Σ(positive indicator score × weight) - Σ(negative indicator score × weight); The result range is 0 to 100 points.

[0044] The effectiveness of the weight matrix was verified through simulation experiments. Taking relay protection services as an example, the performance comparison of using the weight vector [0.1, 0.15, 0.35, 0.2, 0.2] compared with the equal weight [0.2, 0.2, 0.2, 0.2] is shown in Table 2 below: Table 2: Performance Comparison with Equal Weighting

[0045] Sensitivity analysis of the weights shows that when the time delay confidence weight (0.35) is increased by 10%, the path selection accuracy does not change by more than 2%, indicating that the system has appropriate robustness to this parameter.

[0046] Furthermore, the system calls the real-time connectivity status data generated by the aforementioned link layer operation management and maintenance protocol to perform atomic-level closed-loop availability verification on all current fgOTN heterogeneous physical paths based on a two-way handshake mechanism; Specifically, the verification module reads the interaction results between the source and destination network elements within the current 10-millisecond detection period, and determines the link status by comparing the integrity of the bidirectional timestamps carried in the detection messages. If the system detects that the response message in any direction is lost or the round-trip delay exceeds the warning threshold of 20 milliseconds, the system will forcibly clear the availability flag of the heterogeneous link and directly remove the link marked as a one-way failure or interruption from the current unsorted resource pool.

[0047] Furthermore, the path selection operator performs a weighted algebraic sum operation on the positive and negative indicators of the remaining valid heterogeneous links that have passed the verification, based on the business weight vector retrieved from the weight matrix, thereby quantifying the comprehensive quality score of each link under the current business requirements. Specifically, the calculation unit sets the optical path health, transmission reliability, and latency confidence, which represent better performance, as positive indicators with larger values, and the distribution network link latency and bit error rate, which represent better performance with smaller values, as negative indicators. During processing, the calculation unit multiplies the values ​​of the three positive indicators by their corresponding weighting coefficients in the service weight vector and sums them up. At the same time, it multiplies the values ​​of the two negative indicators by their corresponding weighting coefficients and subtracts them from the summation result. Finally, it outputs a floating-point score that characterizes the comprehensive carrying capacity of the link.

[0048] Furthermore, the scheduling module calls the built-in comparison sorting algorithm to perform real-time optimization and sorting of all remaining heterogeneous links based on the calculated floating-point comprehensive quality score; Specifically, the sorting engine uses fast sorting logic to process the port identifiers of all candidate links in descending order, and selects the link with the highest score as the primary transmission path for the current service. This generates a link quality sorting table that is updated in real time and includes link type, physical port number and quality level. This table is then sent to the fgOTN slice scheduling module to drive the service flow to complete path reselection in microseconds.

[0049] By parsing the header tags of service frames to identify service level protocols and dynamically matching weight vectors, this solution achieves precise adaptation between service requirements and link characteristics. It can intelligently select the transmission path with the best overall quality for service flows of different priorities, thereby maximizing the transmission performance of critical services under limited network resources and reflecting the intelligent scheduling concept centered on services.

[0050] Furthermore, the original sequences of transmission reliability, delay confidence, and historical bit error rate over the past 100 sampling periods are extracted from the computing engine at the edge of the fgOTN network element. The rate of change of transmission reliability over time (i.e., the reliability slope) and the variability derivative of delay confidence on the time axis (i.e., the delay confidence slope) are calculated using first-order difference operations. These parameters are then stacked with the historical bit error rate data along the time dimension to construct a multidimensional temporal feature tensor of dimension 100×3. This multidimensional temporal feature tensor is input into a pre-defined lightweight long short-term memory network model. This model is configured with one LSTM layer containing 64 hidden neurons, whose internal weights are pre-trained using the Adam optimizer. A dropout layer with a ratio of 0.2 is then connected to enhance the prediction generalization performance. The end is connected to a fully connected layer containing one output neuron and using the Sigmoid activation function to nonlinearly map the temporal feature vector to continuous probability values ​​between 0 and 1.

[0051] Specifically, the lightweight long short-term memory network model performs recursive prediction processing with a step size of 1 on the input temporal feature tensor, selectively retains historical link decay features using the forget gate inside the LSTM unit, calculates the gain of the current slope change feature using the input gate, and finally generates a set of failure probability sequences that extend with the prediction step size at the output. Each value in this sequence represents the instantaneous probability of the physical link changing from a normal carrying state to an interrupted failure state, thus forming a failure probability curve that reflects the link quality degradation trend.

[0052] The temporal tensor is input into a pre-built lightweight long short-term memory network model. This model is trained offline using over 3000 link fault evolution data samples accumulated during the system's operation. Each sample contains the temporal features of the 100 sampling periods before the fault occurred (corresponding to 1 second of historical monitoring data) and its corresponding fault occurrence time label. The model uses the Adam optimizer with an initial learning rate set to 0.001 and a batch size of [missing information]. The size is 32, and training stops when the effective loss no longer decreases significantly. The LSTM performs recursive processing with a step size of 1. At each step, it receives a sampling point from the temporal tensor, retains historical features through a forget gate, processes the current features through an input gate, and generates a probability value between 0 and 1 at the output of each time step. The LSTM iteratively processes all 100 sampling points of the temporal tensor and finally outputs a failure probability sequence containing 100 probability values. Based on the statistical analysis of historical failure data, the system sets 0.85 as the critical failure judgment threshold. In the generated failure probability sequence, the system sequentially traverses the probability values ​​of each sampling point and retrieves the index position where the first probability value reaches or exceeds 0.85. Let this index be N, then the expected remaining survival time = N × 10 ms.

[0053] Furthermore, the system sets a critical failure threshold of 0.85, sequentially searches the generated failure probability curve for the first time index point where the probability value reaches or exceeds 0.85, and performs a subtraction operation between the future time corresponding to the time index point and the current system time to calculate the expected remaining survival time in milliseconds.

[0054] Specifically, the scheduling engine reads the round-trip delay values ​​between the current working link and the optimal target link in real time from the storage space of the link layer operation management and maintenance protocol module. It then uses a one-way delay estimation method to divide the obtained round-trip delay values ​​by 2, thereby extracting the one-way transmission time reference of the two heterogeneous links at the physical layer signal transmission layer.

[0055] Furthermore, the calculation module performs a difference operation between the unidirectional transmission time base of the current working link and the unidirectional transmission time base of the optimal target link. By taking the absolute value of the calculation result, the relative delay difference between the two heterogeneous links is obtained. This value characterizes the physical quantity of phase offset on the time axis when the cross-link data packet arrives at the destination network element.

[0056] Specifically, the calculation process of physical alignment time includes compensation for relative delay difference and filling of the receiver's de-jitter buffer queue. The system obtains the bandwidth rate of the current service based on the constant bit rate characteristics of the fgOTN service, calculates the time increment required to fill the receiver's de-jitter buffer queue with a preset length of 128 bytes, and sums the time increment with the aforementioned relative delay difference to obtain the total physical alignment time, which includes packet buffering, phase difference elimination, and queue pre-filling.

[0057] Furthermore, the calculated estimated remaining lifetime and physical alignment time are input into the handover decision-maker for logical comparison. If the estimated remaining lifetime is greater than 1.5 times the physical alignment time, it is determined that the physical conditions for preemptive redundancy handover are met. This determination result, along with the physical alignment time parameter, is output to the fgOTN cross-connection control matrix, driving the target link to start parallel data buffering at the specified phase offset. This achieves a closed loop of accurate physical alignment and lossless service flow handover for heterogeneous links before actual failure.

[0058] By using a lightweight LSTM network to extrapolate multidimensional temporal feature tensors, this step transforms the reactive response into proactive prediction. It accurately quantifies the remaining survival time window before link failure and the physical alignment time required to eliminate phase difference, providing a forward-looking spatiotemporal quantitative basis for subsequent differentiated handover decisions. This effectively solves the handover data damage problem caused by perception lag in traditional protection mechanisms.

[0059] The network element switching control unit obtains the estimated remaining lifetime in milliseconds from the output of the lightweight long short-term memory network model in real time, and simultaneously receives the physical alignment time output by the physical alignment calculation module. These two 32-bit floating-point numbers are then input into a difference comparator based on the hardware clock cycle.

[0060] Specifically, the difference comparator obtains the estimated remaining lifetime and physical alignment time delay margin by performing a subtraction operation, and performs a logical judgment on the sign bit of the delay margin according to a preset 0 value benchmark, thereby determining the execution strategy of the current heterogeneous link protection action.

[0061] The calculation process for the physical alignment time is as follows: The scheduling engine obtains the round-trip delay between the current link and the target link from the link layer protocol module. For fiber optic links, the round-trip delay is usually symmetrical, and dividing by 2 gives the one-way delay. For power line and 5G links, the system adopts dynamic asymmetric compensation: monitoring the round-trip delay sequence in the past 20 sampling periods, and calculating the average value of the uplink and downlink delays as the benchmark. The relative delay difference is calculated by subtracting the one-way delay of the current link from the one-way delay of the target link and then taking the absolute value; this value is always non-negative and represents the time offset of data packets arriving at the destination from the two links. The receiving end maintains a reordering buffer of standard length 128 bytes. The system calculates the filling time using the formula: Buffer Filling Time = (Buffer Size × 8) / Service Bandwidth Rate. The physical alignment time is then the sum of the relative latency difference and the buffer filling time. For example, if the relative latency difference is 5... If the service rate is 100Mbps and the buffer is 128 bytes, then the buffer fill time = 128 × 8 / 100M ≈ 10.24 The total physical alignment time was approximately 15.24 seconds. .

[0062] Furthermore, when the determination result is that the estimated remaining lifetime is greater than the physical alignment time and the delay margin is greater than 1.5 milliseconds, the system determines that there is currently enough time margin to complete the lossless handover and generates a lossless handover enable signal. Specifically, in this lossless handover mode, the control plane immediately locks the optimal target link determined by the link quality sorting table and drives the receiving logic unit of the target link to enter the frame-level delay alignment state. The phase offset of the two service streams is identified by parsing the sequence number field in the fgOTN overhead frame. Bit-level shift compensation is performed on the data stream of the target link using a 1024-bit elastic sliding window, so that the clock phase of the data packets of the target link at the destination cross-connection point is strictly synchronized with the current working link. To ensure the complete execution of the lossless handover operation, the system sets a safety margin threshold of 1.5 milliseconds. The complete decision logic is as follows: if the time margin (remaining time - alignment time) is ≥ 1.5ms, perform lossless handover; if 0 ≤ time margin < 1.5ms, enter "cautious mode" and re-determine after a 50ms delay; if the time margin < 0, immediately perform fast handover.

[0063] Furthermore, when the determination result is that the expected remaining lifetime is less than or equal to the physical alignment time, or the delay margin value is negative, the system immediately generates a rapid handover enable signal. Specifically, in this rapid rescue handover mode, the system forcibly rewrites the control register of the cross-connect matrix, directly changes the priority flag bit of the target link from the standby state to the active state, and forcibly bypasses all physical alignment logic related to frame sequence number verification and receiver dejitter buffer queue, directly completing the physical layer connection redirection of the physical link within the current clock cycle. At this time, the system allows the loss of unaligned data packets caused by the sudden change in delay at the moment of handover, so as to ensure that the distribution network control command restores the physical path within a sub-millisecond time.

[0064] Furthermore, the entire decision-making process forms a closed-loop control flow. The input consists of two dynamic time parameters representing the time window game. After logical distribution by the hardware comparator, the output consists of different control primitives directed to the fgOTN physical cross-connection matrix, realizing differentiated and precise control of heterogeneous link switching behavior under different fault severity levels.

[0065] By rigorously comparing the expected survival time with the physical alignment time, this solution can intelligently select the optimal switching strategy based on the urgency of the fault evolution. When there is sufficient time, it can achieve zero-loss migration of services through frame-level alignment, while in case of sudden emergencies, it can force rapid rescue, thus achieving the best balance between data integrity and business continuity, and effectively solving the technical problem of a one-size-fits-all approach to heterogeneous network protection strategies.

[0066] Based on the technical foundation of differentiated handover decisions with sufficient time leeway, this decision logic reflects a profound understanding of the failure evolution patterns in heterogeneous networks: 1. Physical basis of time window prediction: By performing time-series analysis on the evolution of link faults using an LSTM model, the system can predict the occurrence time of the fault in advance, leaving a time window for lossless switching. This is a fundamental improvement over the traditional passive alarm mechanism. 2. Hardware implications of physical alignment time consumption: "Phase alignment" during the handover process involves compensating for the delay difference between heterogeneous links. This requires precise time slot scheduling and buffer management at the hardware layer (fgOTN cross-connect matrix), rather than just logical judgment in software. 3. Improvements to the system architecture through differentiated handover strategy: When time is sufficient (remaining time > alignment time and there is enough time margin to complete lossless handover), the system prioritizes frame-level alignment. This requires maintaining a 1024-byte elastic buffer at the receiving end and performing bit-level precise control on the read pointer. When time is tight (remaining time < alignment time), the system forces bypass of the alignment logic and directly modifies the control register of the cross-connect matrix to achieve microsecond-level physical link redirection. The execution of this judgment logic actually guides the cooperation of multiple control planes (buffer management, token bucket adjustment, time slot mapping) of the fgOTN network element hardware, which is an improvement to the overall design of the communication system.

[0067] A differential delay dynamic compensation buffer mechanism based on timestamp alignment is introduced to solve the problem of asynchronous transmission delay caused by the difference in physical media between primary and backup links (such as optical fiber and 5G / microwave). In heterogeneous protection scenarios, the fgOTN primary link typically has low latency, while heterogeneous backup links (such as public network tunnels or wireless links) have higher latency and greater jitter. The system adds a high-precision timestamp sequence number to each service data packet at the sending end and sets up an elastic reordering buffer at the receiving end. The controller calculates the time difference between the arrival of packets from the primary and backup links in real time and dynamically adjusts the depth of the buffer accordingly. When a switchover is triggered, the receiving end uses the residual data in the buffer to align the packet sequences of the old and new links, ensuring that the data flow of the upper-layer logical links remains continuous and in the correct order at the moment of physical channel switching. This effectively solves the problem of data out-of-order or packet loss at the moment of switchover caused by the huge difference in latency between the primary and backup paths in heterogeneous network environments. By adding time-domain alignment processing to the switchover logic, a truly "lossless switchover" is achieved, greatly improving the smoothness of the user's service experience during network fault switching.

[0068] Furthermore, the system defines the elastic carrying capacity of resources in real time by obtaining the traffic contract parameters of low-priority services in the fgOTN resource management plane.

[0069] Specifically, the resource manager reads the committed information rate and peak information rate values ​​corresponding to low-priority video services from the network element configuration database. The committed information rate is preset to 100 megabits per second and the peak information rate is set to 500 megabits per second. The system calculates the difference between the peak information rate and the committed information rate to obtain a bandwidth range of 400 megabits per second and defines this range as the current resource elasticity boundary. The 400 megabits per second data rate that can be temporarily stripped within this boundary is the maximum transferable bandwidth margin.

[0070] Furthermore, the control plane performs supply and demand matching calculations based on the real-time bandwidth gap on the target link for high-priority services such as relay protection and the aforementioned maximum transferable bandwidth margin.

[0071] Specifically, when it is determined that there is a 50 Mbps transmission bandwidth gap after a high-priority service performs a redundancy switch, the system inputs the gap value as a demand vector to the supply and demand coordination engine. Through a comparison algorithm, it is confirmed that the 400 Mbps concession margin is greater than the 50 Mbps demand value, and thus 50 Mbps is determined as the bandwidth reduction target value for this boundary negotiation.

[0072] Furthermore, the network element sends a 64-byte backpressure signaling message containing the aforementioned 50 Mbps reduction target value to the source network element of the low-priority service through the in-band management channel.

[0073] Specifically, after receiving the backpressure signaling, the source network element uses its internal traffic shaping module to reconfigure the execution parameters of the dual token bucket algorithm, dynamically reducing the token addition rate, which represents the peak rate, from 500 megabits per second to 450 megabits per second, and simultaneously reducing the preset storage depth of the peak burst size in the token bucket. This forces the video data stream flowing through the network element to be constrained within the shrunken logical boundary before entering the fgOTN slice.

[0074] Furthermore, the system confirms the effectiveness of the traffic shaping algorithm by periodically sampling the storage depth of the low-priority service sending queue.

[0075] Specifically, the traffic monitoring unit reads the real-time occupied bytes of the sending queue every 10 milliseconds and divides the value by the total capacity of the queue to obtain the depth percentage. When the queue depth is continuously reduced from the original 80% to below 10% for 5 consecutive sampling cycles and the absolute deviation of the fluctuation remains within 2%, it is determined that the physical contraction of the logical traffic boundary has officially taken effect.

[0076] Furthermore, after confirming that the logical flow control is effective, the system triggers a lossless bandwidth adjustment protocol based on the flexible optical service unit architecture to perform time slot resource redirection mapping at the physical layer.

[0077] Reference Figure 3 This is a schematic diagram of the process of dynamically reducing bandwidth during congestion. Specifically, the cross-connect matrix controller retrieves the corresponding 40 flexible time slot resources in the fgOTN frame structure based on the reduction target of 50 megabits per second. By modifying the pointer offset in the flexible time slot overhead, these time slots are deassociated from the virtual container mapping path of low-priority services, and the mapping routing table of high-priority services is updated synchronously. The 40 released time slot addresses are physically redirected and mapped to the bearer channel of high-priority services.

[0078] Furthermore, the entire resource reduction and redirection process achieves a complete closed loop from logical contract boundaries to physical time slot scheduling. The input is the service commitment and peak rate parameters. After supply and demand matching calculation, backpressure signaling interaction and token bucket parameter shaping, the final output is the reconstruction result of the fgOTN physical time slot table, ensuring deterministic bandwidth guarantee for high-priority services during heterogeneous link protection switching.

[0079] A resource boundary negotiation and traffic engineering control mechanism based on supply and demand matching was introduced. When backup links are congested, the low-priority service resources can be smoothly released to adapt to the high-priority service needs through the linkage of source-end rate limiting and pipeline time slot contraction. This achieves seamless bandwidth relay without interrupting physical connections and improves the resource utilization of critical services in network congestion scenarios.

[0080] Table 3: Performance improvement of this application compared to existing protection schemes (1+1 protection based on fixed threshold triggering)

[0081] The above data is derived from a simulation model of a power distribution network automation application scenario. Specifically, it is verified using a historical dataset containing 1,000 link fault samples. The fault scenarios include three typical faults: gradual decrease in optical power, slow increase in bit error rate, and accumulation of impulse interference. The simulated network topology is a heterogeneous redundancy configuration of 3 optical fibers + 2 power line carriers + 1 5G backup link. Key performance indicators: Improved early warning time: Through LSTM timing prediction, the system can detect faults 70% earlier, allowing sufficient time for lossless switching; Improved lossless switching success rate: The differentiated switching strategy (lossless switching when time is available, fast switching when time is limited) improves the success rate from 70% to 95% compared to traditional fixed threshold protection; Significantly reduced packet loss rate: In scenarios with target link congestion, low-priority bandwidth is dynamically released through resource boundary negotiation, reducing the packet loss rate of high-priority services from 2.5% to 0.08%, a reduction of 96.8%.

[0082] This embodiment breaks the rigid choice between lossless switching and rapid recovery in traditional protection schemes by constructing a differentiated decision mechanism based on the game of survival time and alignment time. It achieves zero-packet lossless protection for gradual faults and millisecond-level rapid rescue for sudden faults. At the same time, combined with resource boundary negotiation in congestion scenarios, it ensures the transmission continuity and service quality reliability of high-priority services throughout the entire life cycle in heterogeneous network environments.

[0083] Example 2 To address the complex issue of sub-microsecond clock synchronization jitter failure caused by the accumulation of physical characteristic differences across multiple paths in the fgOTN multi-hop cascaded network, the system employs a deep spatiotemporal feature analysis model based on the Transformer architecture and a dynamic phase alignment mechanism for physical layer time slots.

[0084] Furthermore, the system introduces a Transformer timing prediction model based on a multi-head self-attention mechanism at the fgOTN destination network element node to extract features and estimate in real time the micro-level latency fluctuation trend accumulated across heterogeneous links.

[0085] Specifically, the model's input receives the inter-frame arrival interval sequence of the past 500 sampling periods obtained by the underlying physical measurement module. Each set of data is encapsulated in the format of a 32-bit floating-point matrix. The model architecture consists of two stacked encoder modules, each configured with four parallel attention heads and a 64-dimensional feature embedding vector space. A feedforward neural network with 256 hidden layer neurons nested inside it uses a layer normalization algorithm to normalize the input signal. By performing the autocorrelation dot product operation of each sampling point in the input feature matrix relative to a 200-millisecond global sliding time window, the temporal correlation weight between different sampling points is calculated, thereby identifying the nonlinear evolution law of link fluctuations. In the output layer, a linear regression mapping layer is used to generate a prediction deviation value representing the expected cumulative jitter of the link in the next 10 milliseconds.

[0086] Furthermore, the system uses the link jitter prediction deviation value output by the Transformer prediction model as a closed-loop control command and sends it in real time to the high-precision time slot offset compensation unit of the fgOTN physical layer to perform hardware-level calibration operations.

[0087] Specifically, after receiving the predicted deviation value, the compensation unit first converts it into the physical offset of the corresponding flexible optical service unit frame structure read pointer through linear mapping logic. Then, it maps the offset into a set of 8-bit binary control codes and writes them into the configuration register of the elastic buffer of the destination network element. The hardware logic circuit forces the rewriting of the starting probe address of the read pointer in the 2048-byte first-in-first-out buffer queue, thereby performing reverse compensation at the bit level for the phase lead or lag of the target link data stream relative to the local reference clock. This ensures that the physical phase of each service frame across heterogeneous physical media achieves microsecond-level coincidence at the destination node, realizing real-time elimination of nondeterministic link disturbances at the physical layer.

[0088] In summary, this embodiment achieves proactive prediction and sub-millisecond hardware compensation for nondeterministic jitter in heterogeneous links by embedding a high-precision Transformer timing prediction network and a physical layer time slot offset dynamic fine-tuning mechanism in the heterogeneous redundancy protection architecture. It constructs a full-stack high-reliability communication guarantee closed loop of fgOTN, from real-time extraction of underlying physical micro-features to deterministic execution at the hardware layer.

[0089] 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. A service-aware fgOTN heterogeneous link collaborative redundancy protection method, characterized in that, include: Real-time monitoring of heterogeneous communication links in the distribution network is performed, and the collected physical layer parameters are mapped to heterogeneous metrics. These metrics are then weighted and synthesized according to the weight of the current service type to generate a link quality ranking table. Specifically, this includes: Parse the Virtual Local Area Network (VLAN) priority tag in the header of the current transmitted service data frame; use the VLAN priority tag as an index key to retrieve the corresponding service weight vector in the preset weight matrix; use connectivity status data to perform bidirectional availability verification on all heterogeneous links, and directly remove links marked as unidirectional failures or interruptions; For the remaining links that have passed the verification, optical path health, transmission reliability, and latency confidence are set as positive indicators, while distribution network link latency and bit error rate are set as negative indicators. The positive and negative indicators are weighted algebraically summed using the service weight vector to obtain the comprehensive quality score of each remaining heterogeneous link. The links are then arranged in descending order of the comprehensive quality score to generate a link quality ranking table. Based on the link quality ranking table, the slope of the index change of the current link is calculated in real time. If the slope of change meets the preset gradual deterioration condition, the estimated remaining lifetime of the current link under non-sudden interruption conditions is calculated through time-series deduction, and the physical alignment time required to switch to the optimal target link is calculated. The calculation process of physical alignment time includes compensation for relative delay difference and filling process of receiver de-jitter buffer queue. The system obtains the bandwidth rate of the current service according to the constant bit rate characteristics of fgOTN service, calculates the time increment required to fill receiver de-jitter buffer queue with a preset length of 128 bytes, and sums the time increment with the aforementioned relative delay difference to obtain the total physical alignment time including packet buffering, phase difference elimination and queue pre-filling. Perform differentiated handover decision based on time adequacy: compare the estimated remaining lifetime with the physical alignment time; if the estimated remaining lifetime is greater than the physical alignment time and there is sufficient time leeway to complete a lossless handover, perform a lossless handover including frame-level delay alignment; if the estimated remaining lifetime is less than or equal to the physical alignment time, perform a rapid salvage handover with bypass alignment logic; in this rapid salvage handover mode, the system forcibly rewrites the control register of the cross-connect matrix. The register directly changes the priority flag of the target link from the standby state to the active state and forces the bypass of all physical alignment logic related to frame sequence number verification and receiver jitter buffer queue. It directly completes the physical layer connection redirection of the physical link within the current clock cycle. At this time, the system allows the loss of unaligned data packets caused by sudden changes in time delay during the switching moment, so as to ensure that the power distribution network control command restores the physical path in sub-millisecond time. After locking onto the optimal target link, if the target link is detected to be in a congested state, during the steady-state maintenance phase after the handover is completed, the bandwidth of low-priority services is dynamically reduced based on the resource boundary negotiation mechanism, and the container channel for high-priority services is adapted.

2. The service-aware fgOTN heterogeneous link collaborative redundancy protection method according to claim 1, characterized in that, The real-time monitoring of heterogeneous communication links in the power distribution network specifically includes: A link-layer operation, management, and maintenance protocol is adopted to establish a bidirectional handshake session between the source and destination network elements. Periodic exchanges of detection messages carrying send and receive timestamps are performed to calculate the round-trip delay of the bidirectional link and verify bidirectional connectivity, obtaining connectivity status data. The digital diagnostic monitoring interface of the fiber optic link module is used to read the input optical power value. The spectrum analysis unit of the power line carrier chip is used to measure the power ratio and count the number of transient spikes, obtaining the signal-to-noise ratio and impulse noise. The physical layer measurement report interface of the 5G wireless link module is used to parse the channel state information reference signal, obtaining the reference signal received power.

3. The service-aware fgOTN heterogeneous link collaborative redundancy protection method according to claim 1, characterized in that, The process of mapping the collected physical layer parameters to heterogeneous metrics specifically includes: Obtain the overload optical power threshold and the receiving sensitivity threshold, calculate the relative position of the collected optical power value within the range of the sensitivity threshold and the overload optical power threshold, and obtain the optical path health through a linear normalization formula; Using the signal-to-noise ratio (SNR) value as a base, an initial link quality score is determined. An interference penalty value for the impulse noise count is calculated using a logarithmic function. The transmission reliability is obtained by subtracting the interference penalty value from the initial link quality score. Based on the reference signal received power, the modulation and coding strategy index is matched, and the historical spectral efficiency statistics of the base station cell are combined to estimate the available transmission rate of the current air interface physical layer. The data volume of the air interface queuing depth is divided by the available transmission rate to calculate the expected queuing delay. The probability that the expected queuing delay meets the service time limit requirement is calculated using the exponential decay model to obtain the delay confidence level.

4. The service-aware fgOTN heterogeneous link collaborative redundancy protection method according to claim 1, characterized in that, The step of calculating the estimated remaining lifetime of the current link through time-series extrapolation and calculating the physical alignment time required to switch to the optimal target link specifically includes: A multidimensional temporal feature tensor containing transmission reliability slope, delay confidence slope, and historical bit error rate is constructed at the network element edge. The multidimensional temporal feature tensor is input into a preset lightweight long short-term memory network model to perform temporal deduction of the current link state, output the failure probability curve, and calculate the expected remaining lifetime. Using the round-trip delay value in the connectivity state data as a physical reference, calculate the relative delay difference between the optimal target link and the current link; based on the relative delay difference, calculate the physical alignment time required to eliminate the inter-link phase difference and fill the receiver's dejitter buffer queue by caching the target link data.

5. The service-aware fgOTN heterogeneous link collaborative redundancy protection method according to claim 1, characterized in that, The execution of the differentiated handover decision based on time adequacy specifically includes: The estimated remaining lifetime is compared with the physical alignment time. If the estimated remaining lifetime is greater than the physical alignment time, it is determined that there is enough time margin to complete the lossless handover. The optimal target link is locked and the frame-level delay alignment procedure is started to maintain the phase synchronization of the data stream during the handover process. If the estimated remaining lifetime is less than or equal to the physical alignment time, lock the optimal target link and force bypass delay alignment logic, and perform a rapid rescue switch in packet loss tolerance mode.

6. The service-aware fgOTN heterogeneous link collaborative redundancy protection method according to claim 1, characterized in that, The mechanism for dynamically reducing bandwidth for low-priority services based on resource boundary negotiation specifically includes: The resource elasticity boundary is defined based on the difference between the committed rate and the peak rate of low-priority services, and the maximum transferable bandwidth margin is calculated. The bandwidth gap of high-priority services is matched with the margin to determine the negotiation reduction target value. A backpressure signaling carrying the target value is sent to the source end of the low-priority service to initiate boundary negotiation, driving the source end to adjust the token bucket parameters to perform traffic shaping to reduce the transmission code rate, and continuously sampling the transmission queue depth. When the queue depth is detected to drop to a low threshold and remain stable, after confirming that the logical traffic boundary contraction has taken effect, the lossless bandwidth adjustment protocol is triggered to physically contract the time slot boundary of the flexible optical service unit of low-priority services, and the released time slot resources are redirected and mapped to the high-priority service channel.

Images