A method and system for evaluating an optical network spectrum fragmentation grooming strategy

Abstract

This invention provides a method and system for evaluating optical network spectrum fragmentation management strategies. The method includes: constructing a predictive digital twin that couples spectrum fragmentation states with physical layer impairment evolution; simulating multiple optical network spectrum fragmentation management strategies in parallel using deterministic service event sequences within the predictive digital twin, and recording the state evolution trajectory of each strategy and a failure log containing a preset failure mode classification; evaluating the multiple optical network spectrum fragmentation management strategies based on the state evolution trajectory and failure log, and outputting the evaluation results. This invention, by constructing a predictive digital twin, provides a physically realistic testing ground for strategy evaluation, and by evaluating multiple optical network spectrum fragmentation management strategies using state evolution trajectories and failure logs, improves the decision-making accuracy of optical network spectrum fragmentation management strategies.

Description

Technical Field

[0001] This invention relates to the field of optical communication network technology, specifically to an evaluation method and system for optical network spectrum fragmentation management strategies. Background Technology

[0002] In Elastic Optical Networks (EON), while the flexible allocation of spectrum resources improves utilization, it also brings about spectrum fragmentation problems: the random establishment and teardown of services will generate a large number of discontinuous, small gaps in the spectrum space that are difficult to be utilized by subsequent high-capacity services, which will seriously degrade the network congestion rate.

[0003] Current industry research primarily focuses on designing more efficient spectrum defragmentation algorithms. However, a long-neglected core issue is how to scientifically, accurately, and with low risk evaluate the long-term effectiveness of different defragmentation strategies. Existing evaluation methods have the following fundamental limitations: First, traditional offline simulation evaluations are distorted. Simulations based on fixed service models and idealized assumptions cannot reproduce the complex, dynamic, and non-steady-state service traffic and impairment environments of real networks. Their evaluation results often differ significantly from actual network performance, leading to poor performance after the simulation-optimized strategy is implemented. Second, direct testing on live networks is highly risky. Any defragmentation operation involves service interruption and rerouting. Conducting strategy trials on live networks directly impacts service quality. Furthermore, existing evaluation methods focus on designing better single-strategy algorithms, failing to conduct comparative analyses of how different strategies might be implemented, resulting in extremely high trial-and-error costs.

[0004] Therefore, there is an urgent need for an evaluation method and system for optical network spectrum fragmentation strategies that can fairly compare multiple strategies and systematically characterize their performance differences under different network states in a near-real dynamic environment, so as to achieve accurate decision-making on optical network spectrum fragmentation strategies. Summary of the Invention

[0005] In view of this, it is necessary to provide a method and system for evaluating optical network spectrum fragmentation strategies, in order to solve the technical problem that existing technologies cannot fairly compare multiple strategies and systematically characterize their performance differences under different network states in a near-real dynamic environment, resulting in inaccurate decision-making on optical network spectrum fragmentation strategies.

[0006] To address the aforementioned technical problems, in a first aspect, the present invention provides a method for evaluating optical network spectrum fragmentation management strategies, comprising: Construct a predictive digital twin that couples spectral fragmentation states with physical layer damage evolution; In the predictive digital twin, multiple optical network spectrum fragmentation strategies are simulated in parallel using deterministic service event sequences, and the state evolution trajectory of each optical network spectrum fragmentation strategy and failure logs including preset failure mode classifications are recorded. The multiple optical network spectrum fragmentation strategies are evaluated based on the state evolution trajectory and the failure log, and the evaluation results are output. The evaluation results are used to characterize the performance differences of each optical network spectrum fragmentation strategy under different network states.

[0007] In one possible implementation, constructing a predictive digital twin that couples spectral fragment states with physical layer damage evolution includes: The physical optical network is parametrically abstracted to establish a static network model that includes parameters of nodes, links, spectrum resources, and basic link impairments. A fragmentation damage prediction model is constructed, which takes the predicted service load and current spectrum fragmentation characteristics as input and the physical layer damage increment caused by fragmentation as output. The network static model is coupled with the fragmentation damage prediction model to generate the predictive digital twin, which is used to output dynamic damage parameters for physical layer performance calculation of each link in real time.

[0008] In one possible implementation, the current spectrum fragmentation characteristics include a fragmentation index and the normalized bandwidth of the largest consecutive free slot.

[0009] In one possible implementation, the predictive digital twin employs a deterministic service event sequence to simulate multiple optical network spectrum fragmentation strategies in parallel, and records the state evolution trajectory of each optical network spectrum fragmentation strategy and a failure log containing a preset failure mode classification, including: For each of the optical network spectrum fragmentation strategies, a digital twin copy with the same initial state is cloned, and the digital twin copy is driven to run synchronously by a deterministic service event sequence with a fixed seed. When the optical network spectrum fragmentation strategy is unable to process the incoming service requests, the failure modes are determined sequentially according to a preset order. Record the state evolution sequence and failure log of each optical network spectrum fragmentation strategy. The failure log includes the failure occurrence time, service request information and failure mode.

[0010] In one possible implementation, the failure modes include no available path, delay violation, cost exceeding limits, signal quality failure, and policy-driven rejection; then, the step of determining the failure modes sequentially in a preset order includes: Determine if there is a physical path in the current optical network that meets the bandwidth requirements; If there is no physical path that meets the bandwidth requirements, the failure mode is determined to be no available path. If a physical path exists that meets the bandwidth requirements, then determine whether the transmission latency of the physical path that meets the bandwidth requirements is greater than the latency limit threshold specified in the service level agreement corresponding to the current service request. If the transmission delay is greater than the delay limit threshold specified in the service level agreement corresponding to the current service request, the failure mode is a delay violation. If the transmission delay is less than or equal to the delay limit threshold specified in the service level agreement corresponding to the current service request, then determine whether the minimum cleanup and migration cost corresponding to the current service request is greater than the cost threshold. If the minimum cleanup and migration cost corresponding to the current business request is greater than the cost threshold, the failure mode is cost overrun. If the minimum cleanup and migration cost corresponding to the current business request is less than or equal to the cost threshold, then obtain the predicted physical layer damage increment determined based on the predictive digital twin, and determine the current optical signal-to-noise ratio based on the predicted physical layer damage increment, and determine whether the current optical signal-to-noise ratio is less than the required signal-to-noise ratio value. If the current optical signal-to-noise ratio is less than the required signal-to-noise ratio value, the failure mode is that the signal quality is substandard. If the current optical signal-to-noise ratio is greater than or equal to the required signal-to-noise ratio value, the failure mode is policy-driven active rejection.

[0011] In one possible implementation, the evaluation of the multiple optical network spectrum fragmentation strategies based on the state evolution trajectory and the failure log, and the output of the evaluation results, include: Based on multiple predefined evaluation scenarios and corresponding dynamic weighting functions, the weighted performance score of each optical network spectrum fragmentation strategy under different evaluation scenarios is calculated, and the weighted performance score is used as the evaluation result.

[0012] In one possible implementation, the step of calculating the weighted performance score of each optical network spectral fragmentation strategy under different evaluation scenarios based on multiple predefined evaluation scenarios and corresponding dynamic weight functions includes: Multiple evaluation scenarios are defined, and each evaluation scenario corresponds to a dynamic weight function that changes dynamically with time and network state. For each of the optical network spectrum fragmentation strategies and each of the evaluation scenarios, the instantaneous cost vector is determined based on the state evolution trajectory, and the weighted instantaneous cost is determined based on the dynamic weighting function. The weighted instantaneous costs are accumulated over time throughout the entire evaluation period to obtain the weighted performance score of each optical network spectrum fragmentation strategy in each evaluation scenario.

[0013] In one possible implementation, the method further includes: An evolutionary dataset is constructed based on failure logs, network status snapshots, and business contexts from historical evaluation periods. The parameters in the dynamic weight function are dynamically calibrated based on the evolutionary dataset and the online learning mechanism. Unsupervised clustering analysis is performed on the evolutionary dataset to identify new typical failure scenarios, and the test case library is updated based on the new typical failure scenarios.

[0014] In one possible implementation, the method further includes: The frequency of occurrence of various failure modes is statistically analyzed based on the failure logs of each optical network spectrum fragmentation strategy, and a strategy failure mode spectrum is constructed based on the frequency of occurrence. For each of the aforementioned evaluation scenarios, a ranking list of scenario strategies is generated based on the weighted performance score of each optical network spectrum fragmentation management strategy. For each evaluation scenario, the optimal recommended strategy for that evaluation scenario is determined based on the scenario strategy ranking list, and the failure mode distribution of the optimal strategy is extracted from the strategy failure mode spectrum. Calculate the relative difference in performance scores between the optimal recommendation strategy and the preset benchmark strategy in each evaluation scenario; A structured strategy selection report is generated based on the optimal recommendation strategy, the failure mode distribution of the optimal strategy, and the relative difference in performance scores.

[0015] Secondly, the present invention also provides an evaluation system for optical network spectrum fragmentation management strategies, comprising: A digital twin building block is used to construct predictive digital twins that couple spectral fragment states with physical layer damage evolution; The strategy parallel simulation module is used to simulate multiple optical network spectrum fragmentation strategies in parallel using deterministic service event sequences in the predictive digital twin, and to record the state evolution trajectory of each optical network spectrum fragmentation strategy and failure logs including preset failure mode classifications. The strategy evaluation module is used to evaluate the multiple optical network spectrum fragmentation strategies based on the state evolution trajectory and the failure log, and output the evaluation results. The evaluation results are used to characterize the performance differences of each optical network spectrum fragmentation strategy under different network states.

[0016] The beneficial effects of this invention are: The optical network spectrum fragmentation management strategy evaluation method provided by this invention, by constructing a predictive digital twin that couples the spectrum fragmentation state with the physical layer damage evolution, can simulate the dynamic behavior of real networks with high fidelity in a safe and lossless virtual environment, solving the core problems of distortion in traditional simulation evaluation and high risk in live network testing, and providing a test field that is close to physical reality for strategy evaluation.

[0017] Secondly, by employing deterministic service event sequences in parallel simulation of multiple optical network spectrum fragmentation strategies in a predictive digital twin, this invention ensures that all optical network spectrum fragmentation strategies to be evaluated face a fair test under identical and dynamically changing service loads and environments, eliminating the interference of random factors on strategy comparison and achieving scientific and repeatable comparison of strategy performance.

[0018] Furthermore, this invention evaluates multiple optical network spectrum fragmentation strategies based on state evolution trajectories and failure logs, and outputs evaluation results to characterize the performance differences of each strategy under different network states. This provides a comprehensive and quantitative basis for strategy selection, improves the decision-making accuracy of optical network spectrum fragmentation strategies, and thus significantly enhances the scientific and refined level of optical network resource management. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 A schematic flowchart of an embodiment of the optical network spectrum fragmentation management strategy evaluation method provided by the present invention; Figure 2 For the present invention Figure 1 A schematic diagram of an embodiment of step S101; Figure 3 For the present invention Figure 1 A schematic flowchart of an embodiment of step S102; Figure 4 A schematic flowchart of an embodiment of the present invention, which sequentially determines failure modes in a preset order; Figure 5 A schematic diagram of an embodiment of the calculation of weighted performance score provided by the present invention; Figure 6 A schematic diagram illustrating an embodiment of the present invention for constructing and evolving an evolutionary dataset; Figure 7 A schematic diagram of an embodiment of the generation strategy selection report provided by the present invention; Figure 8 This is a schematic diagram of an embodiment of the optical network spectrum fragmentation strategy evaluation system provided by the present invention. Detailed Implementation

[0021] 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 a part of the embodiments of the present invention, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0022] It should be understood that the illustrative drawings are not drawn to scale. The flowcharts used in this invention illustrate operations implemented according to some embodiments of the invention. It should be understood that the operations in the flowcharts may be implemented out of order, and steps without logical contextual relationships may be reversed or performed simultaneously. Furthermore, those skilled in the art, guided by the content of this invention, may add one or more other operations to the flowcharts, or remove one or more operations from the flowcharts. Some block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities may be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor systems and / or microcontroller systems.

[0023] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a mutually exclusive, independent, or alternative embodiment. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0024] This invention provides a method and system for evaluating optical network spectrum fragmentation management strategies, which will be described below.

[0025] Figure 1 This is a schematic flowchart of an embodiment of the optical network spectrum fragmentation strategy evaluation method provided by the present invention, as shown below. Figure 1 As shown, the evaluation methods for optical network spectrum fragmentation management strategies include: S101. Construct a predictive digital twin that couples the spectral fragment state with the physical layer damage evolution.

[0026] This invention, through the construction of a predictive digital twin capable of predicting the state of spectrum fragmentation and the evolution of physical layer impairments, shifts the perspective of strategy evaluation from reviewing past static slices to extrapolating future dynamic environments. By constructing a predictive digital twin that can anticipate changes in service load and the evolution of physical layer impairments, the evaluation system can examine the long-term adaptability, robustness, and physical layer feasibility of each strategy in the face of unknown service requests and dynamic signal quality constraints on a time evolution dimension that approximates the real network. This allows for the selection of spectrum management strategies that are not only optimal at the present moment but also maintain excellent performance in complex future changes.

[0027] S102. In the predictive digital twin, multiple optical network spectrum fragmentation strategies are simulated in parallel using deterministic service event sequences, and the state evolution trajectory of each optical network spectrum fragmentation strategy and failure logs containing preset failure mode classifications are recorded. S103. Evaluate multiple optical network spectrum fragmentation strategies based on state evolution trajectories and failure logs, and output the evaluation results. The evaluation results are used to characterize the performance differences of each optical network spectrum fragmentation strategy under different network states.

[0028] It should be understood that the optical network spectrum fragmentation strategy evaluation method in the embodiments of the present invention can be implemented in any device based on the optical network spectrum fragmentation strategy evaluation method, such as a flexible optical network spectrum resource control device. Specifically, the optical network spectrum fragmentation strategy evaluation method is stored in the aforementioned device as a pre-programmed program. When the device starts, the program is invoked, and the optical network spectrum fragmentation strategy evaluation method is implemented.

[0029] Compared with existing technologies, the optical network spectrum fragmentation management strategy evaluation method provided in this invention constructs a predictive digital twin that couples spectrum fragmentation state with physical layer damage evolution. This enables high-fidelity simulation of the dynamic behavior of real networks in a safe and lossless virtual environment, solving the core problems of distortion in traditional simulation evaluation and high risks in live network testing. It provides a test field that closely resembles physical reality for strategy evaluation.

[0030] Secondly, by employing deterministic service event sequences in parallel simulation of multiple optical network spectrum fragmentation strategies in a predictive digital twin, this embodiment of the invention ensures that all optical network spectrum fragmentation strategies to be evaluated face a fair test under identical and dynamically changing service loads and environments, eliminating the interference of random factors on strategy comparison and achieving scientific and repeatable comparison of strategy performance.

[0031] Furthermore, this embodiment of the invention evaluates multiple optical network spectrum fragmentation strategies based on state evolution trajectories and failure logs, and outputs evaluation results to characterize the performance differences of each strategy under different network states. This provides a comprehensive and quantitative basis for strategy selection, improves the decision-making accuracy of optical network spectrum fragmentation strategies, and thus significantly enhances the scientific and refined level of optical network resource management.

[0032] In some embodiments of the present invention, such as Figure 2 As shown, step S101 includes: S201. Parametrically abstract the physical optical network and establish a static network model that includes parameters of nodes, links, spectrum resources, and basic link impairments.

[0033] Specifically, the physical optical network is abstracted as a graph structure G = (V, E, Λ). Here, V is the set of nodes, E is the set of fiber links, and Λ is the set of global spectrum resources. A basic impairment parameter vector Θ_e_base is defined for each link e∈E.

[0034] S202. Construct a fragmentation damage prediction model. The fragmentation damage prediction model takes the predicted service load and the current spectrum fragmentation characteristics as input and the physical layer damage increment caused by fragmentation as output.

[0035] Specifically, a fragmentation damage prediction function M_FD is defined. This function takes the predicted traffic load L(t) for the future time window [t, t+Δt] and the spectral fragmentation feature Frag(t) at the current time t as input. The components of Frag(t) include, but are not limited to: the fragmentation index and the normalized bandwidth of the largest consecutive idle frequency slot. The output of the fragmentation damage prediction function is the predicted damage increment vector ΔΘ_e_frag(t) = M_FD(L(t), Frag(t)).

[0036] S203. Couple the network static model with the fragmented damage prediction model to generate a predictive digital twin. The predictive digital twin is used to output the dynamic damage parameters of each link for physical layer performance calculation in real time.

[0037] For simulation time t, the dynamic impairment parameter used by link e in the predicted digital twin is: Θ_e_sim(t) =Θ_e_base +ΔΘ_e_frag(t).

[0038] The dynamic impairment parameter refers to a quantitative indicator in the predictive digital twin that characterizes the transmission quality of the physical layer of an optical fiber link and dynamically changes with time and the state of spectral fragmentation. It is defined as the sum of the basic impairment parameter of the link and the impairment increment output by the fragmentation impairment prediction model. The basic impairment parameter reflects the inherent physical characteristics of the link (such as nonlinear coefficients, noise figure, etc.), while the impairment increment is a dynamic compensation amount calculated based on the predicted service load and the current spectral fragmentation characteristics, so that the parameter can reflect the time-varying impact of the degree of spectral fragmentation on the physical layer performance in real time.

[0039] This invention, through coupling a static network model with a fragmentation impairment prediction model, constructs a predictive digital twin capable of outputting dynamic impairment parameters in real time. This digital twin possesses the ability to perceive the impact of changes in spectral fragmentation states on physical layer transmission quality, thus overcoming the limitation of fixed physical layer parameters in traditional simulations. This allows signal quality determination in subsequent policy evaluation to be based on dynamic impairment parameters that are linked to the current fragmentation state in real time, significantly improving the simulation environment's approximation of real network physical layer behavior. Ultimately, the policy evaluation results output by this twin can more accurately reflect the physical layer constraints faced by each strategy in actual deployment, providing a more reliable basis for policy selection.

[0040] In some embodiments of the present invention, such as Figure 3 As shown, step S102 includes: S301. For each optical network spectrum fragmentation strategy, clone digital twin copies with identical initial states, and drive the synchronous operation of the digital twin copies with a deterministic service event sequence with a fixed seed.

[0041] Specifically, when N optical network spectrum fragmentation strategies are included, for each of the N optical network spectrum fragmentation strategies {π_1, π_2, ..., π_N}, a digital twin copy {G_1, G_2, ..., G_N} with identical initial state is cloned. All digital twin copies are driven by the same deterministic service event sequence Φ, which is generated by a pseudo-random process with a fixed seed. Each event event_k = (t_k, type_k, r_k) contains absolute time, type (arrival / departure from departure), and service request r_k.

[0042] S302. When the optical network spectrum fragmentation strategy cannot handle the incoming service requests, the failure modes shall be determined in a preset order. S303. Record the state evolution sequence and failure log of each optical network spectrum fragmentation strategy. The failure log includes the failure occurrence time, service request information and failure mode.

[0043] Specifically, the state evolution sequence refers to a series of continuously changing system state snapshots recorded over time by each optical network spectrum fragmentation strategy and its corresponding digital twin copy during the simulation process, driven by a deterministic service event sequence. In particular, this sequence, with time as the axis, fully depicts the continuous changes in network resource occupancy distribution, spectrum fragmentation morphology, link dynamic damage parameters, and internal state variables caused by the decision-making behavior of each strategy in response to the arrival or departure of each service request. This provides traceable and replayable complete data support for subsequent performance evaluation and failure mode analysis.

[0044] In a specific embodiment of the present invention, failure modes include no available path, delay violation, cost exceeding limits, signal quality failure, and policy-driven active rejection; then, as follows Figure 4 As shown, step S302, which involves determining failure modes in a preset order, includes: S401. Determine if there is a physical path in the current optical network that meets the bandwidth requirements; S402. If there is no physical path that meets the bandwidth requirements, the failure mode is determined to be no available path. S403. If there is a physical path that meets the bandwidth requirements, determine whether the transmission delay of the physical path that meets the bandwidth requirements is greater than the upper limit threshold of the delay specified in the service level agreement corresponding to the current service request. S404. If the transmission delay is greater than the delay limit threshold specified in the service level agreement corresponding to the current service request, the failure mode is a delay violation. S405. If the transmission delay is less than or equal to the delay limit threshold specified in the service level agreement corresponding to the current service request, then determine whether the minimum reorganization and migration cost corresponding to the current service request is greater than the cost threshold.

[0045] It should be noted that the minimum reorganization and migration cost can be calculated using a graph-theoretic heuristic algorithm. Specifically: First, enumerate the top K shortest physical paths between the source and destination nodes. For each path, evaluate the minimum number of service migrations or migration slot distances required to reorganize the current idle spectrum segment into continuous bandwidth. Take the minimum value among all paths as the minimum reorganization and migration cost for that service request. If the network scale is too large, resulting in excessively high computational complexity, a preset computation time threshold can be set, and a greedy algorithm can be used to obtain an approximate solution within the threshold.

[0046] S406. If the minimum cleanup and migration cost corresponding to the current business request is greater than the cost threshold, the failure mode is cost overrun. S407. If the minimum reorganization and migration cost corresponding to the current business request is less than or equal to the cost threshold, then obtain the predicted physical layer damage increment determined based on the predictive digital twin, and determine the current optical signal-to-noise ratio based on the predicted physical layer damage increment, and determine whether the current optical signal-to-noise ratio is less than the required signal-to-noise ratio value.

[0047] It should be noted that the predicted physical layer damage increment is the dynamic damage parameter: Θ_e_sim(t) = Θ_e_base + ΔΘ_e_frag(t).

[0048] S408. If the current optical signal-to-noise ratio is less than the required signal-to-noise ratio value, the failure mode is that the signal quality is substandard. S409. If the current optical signal-to-noise ratio is greater than or equal to the required signal-to-noise ratio value, the failure mode is policy-driven active rejection.

[0049] Specifically, the order and logic for determining failure modes are shown in the table below:

[0050] The principle behind the failure mode determination in this invention, which follows the sequence of no available path - latency violation - cost exceedance - substandard signal quality - policy-driven rejection, is based on a progressive decision-making chain in business request processing, from resource availability to service quality constraints to policy business logic. Specifically, it first checks the physical foundation of the underlying network for connectable resources, then progressively verifies whether the resource meets transmission performance requirements, and finally addresses the economic or rule constraints of the policy itself. The aim is to accurately pinpoint the root cause of service failure when it occurs, avoiding the masking of deeper problems by superficial failures (e.g., simply attributing failures caused by signal quality degradation to insufficient resources), thus providing clear diagnostic information for subsequent policy optimization. In other words, by constructing a hierarchical and traceable failure mode determination logic, this invention enables the evaluation method to progressively peel away the causes of failure from the physical layer, performance layer, to the policy layer. This not only quantifies the capability boundaries of each policy in different dimensions but also reveals the intrinsic mechanism of its failure, providing precise optimization directions for targeted policy improvement and significantly enhancing the diagnostic depth and guiding value of the evaluation results.

[0051] In some embodiments of the present invention, step S103 specifically includes: Based on multiple predefined evaluation scenarios and corresponding dynamic weighting functions, the weighted performance score of each optical network spectrum fragmentation strategy under different evaluation scenarios is calculated, and the weighted performance score is used as the evaluation result.

[0052] It is important to note that the multiple evaluation scenarios are mutually exclusive. This means that the diverse and potentially conflicting operational objectives of network operators (such as prioritizing reliability or cost-effectiveness) are formalized into quantifiable evaluation criteria. Specifically, each evaluation scenario represents a particular operational intent, while the dynamic weighting function adjusts the penalty intensity for different failure costs (such as congestion, migration costs, and latency violations) in real time based on time and network conditions. This allows for compatibility and trade-offs between multidimensional and even mutually exclusive objectives within a single evaluation framework.

[0053] In other words, by introducing an intent-aware multi-scenario evaluation mechanism, the policy performance score is no longer an abstract value detached from the actual operational context, but is deeply bound to the real concerns of network managers (such as ensuring high-level services and controlling operation and maintenance costs). This provides operators with clearly targeted policy rankings for different service positioning, realizing a leap from general performance comparison to intent-driven intelligent selection, and significantly improving the practicality and decision-making guidance value of the evaluation results.

[0054] In specific embodiments of the present invention, such as Figure 5 As shown, based on multiple predefined evaluation scenarios and corresponding dynamic weighting functions, the weighted performance scores of each optical network spectrum fragmentation strategy under different evaluation scenarios are calculated, including: S501. Define multiple evaluation scenarios, each with a corresponding dynamic weight function that changes dynamically with time and network status.

[0055] In a specific embodiment of the present invention, when the evaluation scenario is a reliability-priority scenario, the dynamic weight function W_rel(t, S) is: W_rel(t,S)= [w_no_path=10.0, w_delay=5.0, w_cost=1.0, w_quality=8.0,w_policy=0.5]; When the evaluation scenario is a cost-priority scenario, the dynamic weighting function W_eco(t, S) is: W_eco(t,S)=[w_no_path=2.0,w_delay=1.0, w_cost=10.0, w_quality=3.0, w_policy=0.1]; Where w_no_pat, w_delay, w_cost, w_quality, and w_policy are respectively the cost weights for no available path, delay violation, cost exceeding limits, signal quality failure, and policy-driven rejection.

[0056] It should be noted that S represents the real-time state of the optical network, such as the network load rate, and the dynamic weight function can be conditionally assigned values ​​based on S. In other words, the dynamic weight function is the basic weight vector, which defines the relative importance of different failure types (blocking / cost / migration), forming a static baseline for the evaluation scenario. Based on this static baseline, the weights can be dynamically fine-tuned according to the instantaneous state of the network (such as the current load rate) to reflect the differentiated impact of the same failure type under different network conditions.

[0057] S502. For each optical network spectrum fragmentation strategy and each evaluation scenario, determine the instantaneous cost vector based on the state evolution trajectory, and determine the weighted instantaneous cost based on the dynamic weighting function.

[0058] It should be noted that: instant cost vector = [No available path (0 / 1), delay violation (0 / 1), cost exceeded (0 / 1 or exceeded value), signal quality not up to standard (0 / 1), policy actively rejected (0 / 1)].

[0059] Specifically, the weighted instantaneous cost c_i^k(t) is:

[0060] in, = [w_1(t), w_2(t), ..., w_5(t)] is the dynamic weight vector. = [c_1(t), c_2(t), ..., c_5(t)] is the instantaneous cost vector, both of which are five-dimensional row vectors. The dot product is the sum of the products of each component.

[0061] S503. Accumulate the weighted real-time costs on the time axis throughout the entire evaluation period to obtain the weighted performance score of each optical network spectrum fragmentation strategy in each evaluation scenario.

[0062] Specifically, the weighted efficiency score is:

[0063] Where m is the sequence number of the business event (from 1 to M). γ Time discount factor (0 < γ ≤1), t m Let m be the time when the m-th event occurs. For strategy i in evaluation scenario k at time t m The weighted immediate cost generated at each moment. The lower the weighted performance score, the better the strategy performance.

[0064] Given that existing evaluation methods are usually one-time static evaluations, their evaluation criteria and test cases do not change once determined, and they cannot learn and improve from historical evaluation data; at the same time, the evaluation system lacks the ability to perceive unforeseen failure scenarios, causing its judgment to gradually lag behind changes in the real network environment over time, making it difficult to cope with the evolution of business models or the emergence of new failure modes.

[0065] To solve the above-mentioned technical problems, in some embodiments of the present invention, such as Figure 6 As shown, the evaluation method for optical network spectrum fragmentation management strategies also includes: S601. Construct an evolutionary dataset based on failure logs, network status snapshots, and business contexts from historical evaluation periods; S602. Dynamically calibrate the parameters in the dynamic weight function based on the evolutionary dataset and online learning mechanism.

[0066] Specifically, a weight calibration model is trained based on an evolutionary dataset. This model learns the implicit correlations between different failure modes and final business metrics (such as customer satisfaction loss and economic costs) under different network contexts (e.g., workload level, fragmentation severity, business tier distribution). The output of the weight calibration model is periodically used to fine-tune the weight vector function W_k(t, S) to make the evaluation criteria more reflective of the actual impact of failures.

[0067] S603. Perform unsupervised clustering analysis on the evolutionary dataset to identify new typical failure scenarios, and update the test case library based on the new typical failure scenarios.

[0068] Specifically, unsupervised clustering analysis (such as using density-based clustering algorithms) is performed on failure events in the evolutionary dataset to automatically identify typical failure scenario patterns that occur frequently or lead to serious consequences and are not covered in the initial test sequence Φ. The characteristics of these new patterns (such as specific business arrival combinations and fragment distribution patterns) are transformed into new adversarial test templates and dynamically added to the business event sequence generator to enrich the test case library for subsequent evaluation, thereby achieving the goal of proactively discovering unknown defects in the strategy.

[0069] This invention, through the construction of an evolutionary dataset and the introduction of an online learning mechanism, enables the evaluation system to possess self-optimization capabilities. On one hand, by dynamically calibrating the parameters in the weighting function, the evaluation criteria can reflect the real-time impact of different failure modes on actual business operations. On the other hand, by automatically mining new typical failure scenarios and updating the test case library through unsupervised clustering analysis, the evaluation system can proactively discover and test unknown defects in the strategy. This mechanism achieves continuous evolution of the evaluation system's discriminative power and predictive ability, fundamentally solving the aging problem of static evaluation systems.

[0070] Furthermore, traditional assessment methods typically only output a single overall score or ranking, failing to reveal the underlying mechanisms of strategy failure and making it difficult to connect assessment results with specific operational intentions. In other words, when faced with the question of which strategy is more suitable for their network, network operators lack decision support tools that can intuitively present the boundaries of strategy capabilities, the distribution of failure modes, and a quantitative comparison with benchmark strategies, resulting in limited practical guiding value of the assessment results.

[0071] To solve this technical problem, in some embodiments of the present invention, such as Figure 7 As shown, the evaluation method for optical network spectrum fragmentation management strategies also includes: S701. Calculate the frequency of occurrence of various failure modes based on the failure logs of each optical network spectrum fragmentation strategy, and construct a strategy failure mode spectrum based on the frequency of occurrence.

[0072] Specifically, the strategy failure mode spectrum refers to the distribution ratio of failure modes.

[0073] S702. For each evaluation scenario, a ranking list of scenario strategies is generated based on the weighted performance score of each optical network spectrum fragmentation strategy. S703. For each evaluation scenario, determine the optimal recommended strategy under the evaluation scenario based on the scenario strategy ranking list, and extract the failure mode distribution of the optimal strategy from the strategy failure mode spectrum. S704. Calculate the relative difference in performance scores between the optimal recommended strategy and the preset benchmark strategy in each evaluation scenario.

[0074] The preset benchmark strategy is a pre-selected optical network spectrum fragmentation strategy, such as the First-Fit algorithm or a passive allocation strategy that does not perform active fragmentation, which is used as a reference baseline for performance comparison.

[0075] Specifically, for the evaluation scenario, the benchmark weighted performance score corresponding to its prediction benchmark strategy is obtained, and the relative difference in performance scores ΔScore_k is:

[0076] In the formula, The baseline weighted performance score; This represents the weighted performance score corresponding to the optimal recommendation strategy.

[0077] S705. Generate a structured strategy selection report based on the optimal recommendation strategy, the failure mode distribution of the optimal strategy, and the relative difference in performance scores.

[0078] This invention constructs a strategy failure mode spectrum and generates scenario strategy rankings. The failure mode spectrum reveals the weaknesses of each strategy (e.g., 80% of a strategy's failures stem from cost overruns), while the scenario strategy rankings directly answer the applicability of a strategy under different operational intentions. By calculating the relative difference in performance scores with benchmark strategies, it provides a quantifiable improvement margin. The final generated strategy selection report integrates multi-dimensional information such as optimal recommendations, failure distribution, and performance comparisons, enabling network operators to make accurate and quantifiable strategy selection decisions based on their own operational goals, significantly improving the practicality and decision-making guidance value of the evaluation results.

[0079] In addition, the strategy selection report may include the impact of new typical failure scenarios on strategy ranking, or indicate the calibration changes of the evaluation weights compared to the previous evaluation, reflecting the evolution of the evaluation method.

[0080] To facilitate understanding of the optical network spectrum fragmentation management strategy evaluation method proposed in this embodiment of the invention, a specific metropolitan core optical network scenario will be used as an example for illustration.

[0081] 1. Scenario Setting for the Implementation Example Physical network: A bidirectional ring topology with 6 nodes, link numbers E={e1, e2, ..., e6}. It uses the C-band with a 12.5 GHz spectrum grid and a total of 320 slots (Λ = {1, 2, ..., 320}). The fundamental nonlinear coefficient γ_base_e = 1.3 / W / km for all links.

[0082] Business Model: Service request arrival follows a Poisson process, with an average arrival rate λ = 5 requests / second. The bandwidth requirement bw is uniformly distributed across {50GHz (4 slots), 100GHz (8 slots), 200GHz (16 slots)}, with source-destination node pairs selected uniformly and randomly. Service duration follows an exponential distribution with a mean of 300 seconds. 20% of these are high-priority services, with SLA requirements of: SLA_delay ≤ 5ms and OSNR_Required ≥ 18 dB.

[0083] Strategies to be evaluated: π_A (Preset Baseline Strategy): First-Fit algorithm, no active cleanup.

[0084] π_B: A threshold-based heuristic algorithm. Local defragmentation is triggered when the fragmentation index (FI) of any link > 0.4, with the goal of maximizing a single continuous gap.

[0085] π_C: An intelligent policy based on deep reinforcement learning (DRL), whose policy network is a multilayer perceptron (MLP).

[0086] 2. Implementation process of the embodiments of the present invention S1: Constructing a predictive digital twin of coupled spectral fragmentation states and physical layer damage S1.1 Network Parametric Modeling: Define a digital twin graph G = (V, E, Λ), where V={v1,...,v6}, E={e1,...,e6}, Λ=320. Each link e has Θ_e_base containing γ_base_e=1.3, amplifier noise figure n_sp=1.5, etc.

[0087] S1.2 Prediction of Fragmentation Damage Increment: At simulation time t=152.3s, the fragmentation feature vector Frag(t) of the entire network spectrum is calculated. For example, for link e1, the occupancy status of its 320 frequency slots is [1: occupied, 2: occupied, ..., 150: idle, 151: idle, ...]. Its fragmentation index FI_e1 = 0.68 is calculated, and the maximum consecutive idle frequency slot MaxGap_e1 = 25.

[0088] The prediction model M_FD (in this example, a pre-trained lightweight GRU network) is invoked. The inputs are the features of link e1 in Frag(t) [FI_e1=0.68, MaxGap_e1=25, ...] and the predicted load L_e1(t)=0.85 for the next 10 seconds.

[0089] The nonlinear enhancement coefficient of the M_FD output prediction is: Δγ_e1(t) = M_FD(L_e1(t), Frag_e1(t)) =0.021 / W / km.

[0090] S1.3 Synthesis of Twin Dynamic Parameters: For link e1, the dynamic nonlinear coefficient used for simulation at time t is: γ_sim_e1(t) = γ_base_e1 + Δγ_e1(t) = 1.3 + 0.021 = 1.321 / W / km. This value will be used for subsequent OSNR calculations.

[0091] S2: Perform multi-strategy parallel simulation and failure mode logging S2.1 Initialization and Driving: Use the random seed seed = 12345 to generate a deterministic service sequence Φ. The k = 1205th event in the sequence is: event_k = (t = 152.3s, type = ARRIVAL, r_k), where r_k = {src = v2, dst = v4, bw = 100GHz (8 frequency slots), class = HIGH, SLA_delay = 5ms, OSNR_Req = 18dB}.

[0092] S2.2 Parallel Execution and Failure Judgment (taking policy π_C as an example): In replica G_C, policy π_C receives r_k. Analyze the current state S_t through its policy network and output the decision action: Try to allocate services in frequency slots c = [151, 158] on path p = {e2, e3}.

[0093] Perform failure judgment check: Path check: Paths(r_k) is not empty, FM_no_path does not hold. <00​​​​​​​​​​​​​​​​​​​​​​​​

[0102] If at the same time, policy π_A cannot find 8 consecutive free resources in another replica due to fragmentation, and the minimum migration cost is 120, which exceeds the cost threshold of 100, it will be judged as FM_cost, and a record will be recorded in Log_A: (152.3, r_k, FM_cost).

[0103] S3: Dynamic Scene Evaluation Based on Mutually Exclusive Network Intent S3.1 Define the scenario and weights: Scenario_Reliability (reliability first): W_rel = [w_no_path=10, w_delay=5, w_cost=1, w_quality=8, w_policy=0]; Scenario_Economy (economic priority): W_eco = [w_no_path=2, w_delay=1, w_cost=10, w_quality=3, w_policy=0].

[0104] S3.2 Calculate the scenario performance score (taking π_C in a reliable scenario as an example within a time window): Within the time period [t, t+1], π_C experienced one service blockage due to no available path (i.e., no available path failure, c_no_path=1), and performed two spectrum migrations with a total migration distance of 15 frequency slots. This migration cost was included in the cost overrun cost (c_cost=15). No delay violations, substandard signal quality, or policy-driven rejections occurred.

[0105] Calculate the instantaneous cost vector: c_C(t) =[c_no_path=1,c_delay=0,c_cost=15,c_quality=0,c_policy=0] The weight vector for the reliability-first scenario (defined according to step S501) is: W_rel=[w_no_path=10, w_delay=5,w_cost=1,w_quality=8, w_policy=0] The weighted immediate cost is: =W_rel·c_C(t)=10×1+5×0+1×15+8×0+0×0=25 Discrete cumulative summation was performed over the entire simulation period T=10000s (approximately 5000 business events), and a time discount factor γ=0.99 was added to obtain the final score of π_C in the reliable scenario: ≈2345.6 Where M is the total number of events within the evaluation period. Let m be the time when the m-th event occurs. This represents the time interval since the last event.

[0106] S4: Online Self-Evolution and Evaluation System Calibration S4.1 Data Collection: After this round of evaluation, a total of M=245 failure records (π_A, π_B, π_C) were collected and stored in the evolutionary dataset D_evo. Each record contains FM, C_t (e.g., [load=0.82, FI=0.55, service level=HIGH]) and a preset initial impact value (1.0 for high-level service failures).

[0107] S4.2 Online Weight Calibration (Calculation Example): Regression analysis was performed on all FM_quality records in the evolutionary dataset. It was found that when "load > 0.8 and FI > 0.5" in C_t, the predicted value of Impact, P_impact, was 2.3 times that of other scenarios.

[0108] Calibration rule generation: Based on this, the internal weight calibration function is updated. The rule is: if load > 0.8 and FI > 0.5: w_quality(t) = w_quality_base × 2.3. This rule will take effect in step S3 of the next round of evaluation.

[0109] S4.3 Typical Failure Scenario Mining (Calculation Example): Perform DBSCAN clustering on all FM_delay records in the evolutionary dataset. Set the neighborhood radius eps=0.5 and the minimum number of samples min_samples=5.

[0110] The algorithm outputs 3 clusters. The feature center of cluster 1 (12 records in total) is decoded as follows: between nodes v3 and v4, when the total network load is >0.9, a delay violation occurs due to an excessively long rerouting path for a high-level, high-bandwidth (200GHz) service request.

[0111] Save this pattern as a new test template T_new_delay for use in generating targeted test traffic during subsequent evaluations.

[0112] S5: Generate a selection report with insights into fusion and evolution. Data Integration: The system integrates S3 scores and S2 failure statistics. The final calculation yields: Scenario ranking (ascending order): Rank_rel = [π_C (2345.6), π_B (3012.3), π_A(4500.8)]; Rank_eco = [π_B (1987.2), π_A (2100.5), π_C (3254.1)].

[0113] Failure spectrum example: P_C(FM) = {FM_quality: 5%, FM_delay: 10%, FM_cost: 80%, FM_policy: 5%}.

[0114] S5.3 Report Generation: Generates a report containing the following core content: Tabular results: Clearly displaying the above rankings and failure spectrum.

[0115] Quantitative comparison: ΔScore_rel(π_C vs π_A) = (4500.8 - 2345.6) / 4500.8 ≈ 48%, indicating that π_C improves performance by 48% in reliable scenarios.

[0116] Evolutionary hints: Dynamic weight adjustment has been enabled: the new round of evaluation will apply a 2.3x penalty weight to FM_quality failures in "high load and high fragmentation" scenarios.

[0117] Added robustness test item: The test case library has added the template T_new_delay, which is used to specifically test the "large business throughput capability of the strategy under ultra-high load between nodes (v3,v4)".

[0118] Warning about the shortcomings of the strategy: The data from this evaluation shows that 80% of the failures of the π_C strategy are due to FM_cost (cost exceeding the limit). It is recommended to choose this strategy with caution in scenarios with strict budget constraints.

[0119] In summary, the optical network spectrum fragmentation management strategy evaluation method proposed in this invention solves the core problems of distortion in traditional strategy evaluation and high risk of trial and error in the live network by constructing a predictive digital twin and parallel simulation environment. By introducing root cause analysis of failure modes, it can deeply reveal the capability boundaries and shortcomings of each strategy. Through dynamic scenario evaluation based on mutually exclusive network intentions, it can provide operators with quantitative and precise decision support on which strategy to choose under what objectives. Particularly noteworthy is that this invention, by introducing an online self-evolving closed loop, enables the evaluation system to continuously learn from historical data, dynamically calibrate evaluation standards, and automatically discover potential risk scenarios. This achieves self-improvement in the discrimination and predictive power of the evaluation system, giving it long-term adaptability and intelligence that becomes more accurate with use. Ultimately, it realizes a fundamental shift from static experience-based decision-making to a data-driven and continuously evolving intelligent decision-making paradigm, significantly improving network resource utilization efficiency and the scientific and forward-looking nature of operation and maintenance.

[0120] On the other hand, embodiments of the present invention also provide an evaluation system for optical network spectrum fragmentation management strategies, such as... Figure 8 As shown, the optical network spectrum fragmentation management strategy evaluation system 800 includes: Digital twin building module 801 is used to construct a predictive digital twin that couples spectral fragmentation state with physical layer damage evolution; The strategy parallel simulation module 802 is used to simulate multiple optical network spectrum fragmentation strategies in parallel using deterministic service event sequences in a predictive digital twin, and to record the state evolution trajectory of each optical network spectrum fragmentation strategy and failure logs containing preset failure mode classifications. The strategy evaluation module 803 is used to evaluate multiple optical network spectrum fragmentation strategies based on state evolution trajectories and failure logs, and output the evaluation results. The evaluation results are used to characterize the performance differences of each optical network spectrum fragmentation strategy under different network states.

[0121] The optical network spectrum fragmentation strategy evaluation system 800 provided in the above embodiments can implement the technical solutions described in the above embodiments of the optical network spectrum fragmentation strategy evaluation method. The specific implementation principles of each module or unit can be found in the corresponding content in the above embodiments of the optical network spectrum fragmentation strategy evaluation method, and will not be repeated here.

[0122] Those skilled in the art will understand that all or part of the processes of the methods described in the above embodiments can be implemented by a computer program instructing related hardware (such as a processor, controller, etc.), and the computer program can be stored in a computer-readable storage medium. The computer-readable storage medium may be a disk, optical disk, read-only memory, or random access memory, etc.

[0123] The above provides a detailed description of the optical network spectrum fragmentation strategy evaluation method and system provided by the present invention. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A method for evaluating optical network spectrum fragmentation management strategies, characterized in that, include: Construct a predictive digital twin that couples spectral fragmentation states with physical layer damage evolution; In the predictive digital twin, multiple optical network spectrum fragmentation strategies are simulated in parallel using deterministic service event sequences, and the state evolution trajectory of each optical network spectrum fragmentation strategy and failure logs including preset failure mode classifications are recorded. The multiple optical network spectrum fragmentation strategies are evaluated based on the state evolution trajectory and the failure log, and the evaluation results are output. The evaluation results are used to characterize the performance differences of each optical network spectrum fragmentation strategy under different network states.

2. The method of evaluating an optical network spectrum fragmentation polishing strategy according to claim 1, wherein, The construction of a predictive digital twin coupling spectral fragment states and physical layer damage evolution includes: The physical optical network is parametrically abstracted to establish a static network model that includes parameters of nodes, links, spectrum resources, and basic link impairments. A fragmentation damage prediction model is constructed, which takes the predicted service load and current spectrum fragmentation characteristics as input and the physical layer damage increment caused by fragmentation as output. The network static model is coupled with the fragmentation damage prediction model to generate the predictive digital twin, which is used to output dynamic damage parameters for physical layer performance calculation of each link in real time.

3. The method for evaluating optical network spectrum fragmentation strategies according to claim 2, characterized in that, The current spectrum fragmentation characteristics include the fragmentation index and the normalized bandwidth of the largest consecutive idle frequency slot.

4. The method of evaluating an optical network spectrum fragmentation polishing strategy according to claim 2, wherein, In the predictive digital twin, multiple optical network spectrum fragmentation strategies are simulated in parallel using deterministic service event sequences. The state evolution trajectory of each optical network spectrum fragmentation strategy and a failure log containing a preset failure mode classification are recorded, including: For each of the optical network spectrum fragmentation strategies, a digital twin copy with the same initial state is cloned, and the digital twin copy is driven to run synchronously by a deterministic service event sequence with a fixed seed. When the optical network spectrum fragmentation strategy is unable to process the incoming service requests, the failure modes are determined sequentially according to a preset order. Record the state evolution sequence and failure log of each optical network spectrum fragmentation strategy. The failure log includes the failure occurrence time, service request information and failure mode.

5. The method of evaluating an optical network spectrum fragmentation polishing strategy according to claim 4, wherein, The failure modes include no available path, delay violation, cost exceeding limits, substandard signal quality, and policy-driven rejection; the step of determining the failure modes in a preset order includes: Determine if there is a physical path in the current optical network that meets the bandwidth requirements; If there is no physical path that meets the bandwidth requirements, the failure mode is determined to be no available path. If a physical path exists that meets the bandwidth requirements, then determine whether the transmission latency of the physical path that meets the bandwidth requirements is greater than the latency limit threshold specified in the service level agreement corresponding to the current service request. If the transmission delay is greater than the delay limit threshold specified in the service level agreement corresponding to the current service request, the failure mode is a delay violation. If the transmission delay is less than or equal to the delay limit threshold specified in the service level agreement corresponding to the current service request, then determine whether the minimum cleanup and migration cost corresponding to the current service request is greater than the cost threshold. If the minimum cleanup and migration cost corresponding to the current business request is greater than the cost threshold, the failure mode is cost overrun. If the minimum cleanup and migration cost corresponding to the current business request is less than or equal to the cost threshold, then obtain the predicted physical layer damage increment determined based on the predictive digital twin, and determine the current optical signal-to-noise ratio based on the predicted physical layer damage increment, and determine whether the current optical signal-to-noise ratio is less than the required signal-to-noise ratio value. If the current optical signal-to-noise ratio is less than the required signal-to-noise ratio value, the failure mode is that the signal quality is substandard. If the current optical signal-to-noise ratio is greater than or equal to the required signal-to-noise ratio value, the failure mode is policy-driven active rejection.

6. The method for evaluating optical network spectrum fragmentation strategies according to claim 1, characterized in that, The evaluation of the multiple optical network spectrum fragmentation strategies based on the state evolution trajectory and the failure log, and the output of the evaluation results, include: Based on multiple predefined evaluation scenarios and corresponding dynamic weighting functions, the weighted performance score of each optical network spectrum fragmentation strategy under different evaluation scenarios is calculated, and the weighted performance score is used as the evaluation result.

7. The method of evaluating an optical network spectrum fragmentation polishing strategy according to claim 6, wherein, The calculation of the weighted performance score of each optical network spectrum fragmentation strategy under different evaluation scenarios, based on multiple predefined evaluation scenarios and corresponding dynamic weight functions, includes: Multiple evaluation scenarios are defined, and each evaluation scenario corresponds to a dynamic weight function that changes dynamically with time and network state. For each of the optical network spectrum fragmentation strategies and each of the evaluation scenarios, the instantaneous cost vector is determined based on the state evolution trajectory, and the weighted instantaneous cost is determined based on the dynamic weighting function. The weighted instantaneous costs are accumulated over time throughout the entire evaluation period to obtain the weighted performance score of each optical network spectrum fragmentation strategy in each evaluation scenario.

8. The method of evaluating an optical network spectrum fragmentation polishing strategy according to claim 6, wherein, The method further includes: An evolutionary dataset is constructed based on failure logs, network status snapshots, and business contexts from historical evaluation periods. The parameters in the dynamic weight function are dynamically calibrated based on the evolutionary dataset and the online learning mechanism. Unsupervised clustering analysis is performed on the evolutionary dataset to identify new typical failure scenarios, and the test case library is updated based on the new typical failure scenarios.

9. The method of evaluating an optical network spectrum fragmentation polishing strategy of claim 7, wherein, The method further includes: The frequency of occurrence of various failure modes is statistically analyzed based on the failure logs of each optical network spectrum fragmentation strategy, and a strategy failure mode spectrum is constructed based on the frequency of occurrence. For each of the aforementioned evaluation scenarios, a ranking list of scenario strategies is generated based on the weighted performance score of each optical network spectrum fragmentation management strategy. For each evaluation scenario, the optimal recommended strategy for that evaluation scenario is determined based on the scenario strategy ranking list, and the failure mode distribution of the optimal strategy is extracted from the strategy failure mode spectrum. Calculate the relative difference in performance scores between the optimal recommendation strategy and the preset benchmark strategy in each evaluation scenario; A structured strategy selection report is generated based on the optimal recommendation strategy, the failure mode distribution of the optimal strategy, and the relative difference in performance scores.

10. An optical network spectrum fragmentation grooming strategy evaluation system, characterized by, include: A digital twin building block is used to construct predictive digital twins that couple spectral fragment states with physical layer damage evolution; The strategy parallel simulation module is used to simulate multiple optical network spectrum fragmentation strategies in parallel using deterministic service event sequences in the predictive digital twin, and to record the state evolution trajectory of each optical network spectrum fragmentation strategy and failure logs including preset failure mode classifications. The strategy evaluation module is used to evaluate the multiple optical network spectrum fragmentation strategies based on the state evolution trajectory and the failure log, and output the evaluation results. The evaluation results are used to characterize the performance differences of each optical network spectrum fragmentation strategy under different network states.

Images