A cross-domain quality of service mapping method and apparatus

Abstract

The application relates to the technical field of power system communication, and discloses a cross-domain service quality mapping method and device, which realizes full-process collaborative optimization through a power heterogeneous cross-domain transmission scene, generates a service standard demand vector through message QoS label mapping, accurately matches transmission requirements such as power service time delay and bandwidth, calculates a nonlinear congestion penalty value relying on queue operation indexes, quantifies transmission loss caused by queue congestion, constructs a global path standard state matrix in combination with a multi-path state, and completely represents real-time working conditions of each transmission link. A path comprehensive adaptation score is obtained through weight self-adaptive adjustment, fuzzy vector inner product matching and congestion attenuation correction, so that the flow proportion is dynamically allocated to complete scheduling. The method effectively balances multi-link loads, reduces power data transmission time delay and packet loss risk, takes into account the service quality demands of different power services, and improves cross-domain power data flow transmission stability and network resource utilization.

Description

Technical Field

[0001] This invention relates to the field of power system communication technology, and specifically to a cross-domain quality of service mapping method and apparatus. Background Technology

[0002] With the accelerated implementation of new power systems, the high proportion of renewable energy grid connection, and the massive distributed source-load access, the power grid control boundary has extended from the traditional grid connection point to the user-side network. Multi-level collaborative control services place stringent constraints on the latency, jitter, and packet loss indicators of transmission links. Power communication is no longer limited to traditional dispatch message transmission; massive real-time services such as protection, energy storage, and distributed power source control are emerging. End-to-end deterministic transmission across the entire link has become a rigid requirement for the safe and stable operation of the power grid. In practice, power transmission links are mostly heterogeneous cross-domain networks integrating multiple standards such as TSN, 5G slicing, and SPN. The industry generally adopts technologies such as time slot scheduling, queue priority, network slicing, and explicit routing to ensure deterministic data forwarding.

[0003] Existing cross-domain QoS guarantee schemes use static mapping tables to achieve network priority interoperability, which has obvious shortcomings. The main problem is that fixed mapping rules cannot perceive the real-time load of the link. When there is a traffic surge or network congestion, the high-priority queue resources are over-reserved, resulting in waste. Critical services are difficult to match with suitable queues, and the transmission accuracy drops significantly. There is an urgent need for a dynamic QoS mapping mechanism that is adapted to cross-domain power scenarios and can achieve transparent protocol alignment. Summary of the Invention

[0004] This invention provides a cross-domain QoS mapping method and apparatus to solve the problems that existing fixed mapping rules for cross-domain QoS cannot perceive real-time link load, and that excessive reservation of high-priority queue resources leads to waste during traffic bursts or network congestion, making it difficult for critical services to match and adapt to the appropriate queue, resulting in a significant decrease in transmission accuracy.

[0005] In a first aspect, the present invention provides a cross-domain quality of service mapping method, the method comprising: When the original power data stream enters the cross-domain gateway, the heterogeneous network service quality tag in the message corresponding to the original power data stream is extracted and mapped to generate a business standard requirement vector. Collect local queue operation metrics of each candidate sending queue within the cross-domain gateway, and use the local queue operation metrics to calculate a nonlinear congestion penalty value; Evaluate the path status of multiple candidate transmission paths of the cross-domain gateway, and construct a global path standard state matrix based on the evaluation results; The business standard demand vector, the nonlinear congestion penalty value, and the global path standard state matrix are used to perform joint adaptation calculation to generate a comprehensive path adaptation score. Based on the comprehensive adaptation score of the path, the elastic allocation ratio of path traffic for each transmission path is determined, and traffic scheduling is performed according to the elastic allocation ratio of path traffic.

[0006] This invention achieves end-to-end collaborative optimization in heterogeneous cross-domain power transmission scenarios. It generates service standard demand vectors through message QoS tag mapping, accurately matching differentiated transmission requirements such as latency and bandwidth for power services. It calculates nonlinear congestion penalty values ​​based on queue operation indicators to quantify transmission losses caused by queue congestion. A global path standard state matrix is ​​constructed by combining multi-path states to fully characterize the real-time operating conditions of each transmission link. A comprehensive path adaptation score is obtained through adaptive weight adjustment, fuzzy vector inner product matching, and congestion attenuation correction, which is used to dynamically allocate traffic proportions for scheduling. This effectively balances the load across multiple links, reduces power data transmission latency and packet loss risks, accommodates different power service quality requirements, and improves the stability of cross-domain power data stream transmission and network resource utilization.

[0007] In one optional implementation, when the original power data stream enters the cross-domain gateway, the heterogeneous network service quality label in the packet corresponding to the original power data stream is extracted and mapped to generate a service standard requirement vector, including: When the raw power data stream enters the cross-domain gateway, the heterogeneous network service quality tag in the message corresponding to the raw power data stream is extracted. A multi-dimensional general service quality requirement model is constructed by adopting the end-to-end latency requirements, latency jitter requirements, maximum allowable packet loss rate, and service guarantee bandwidth requirements of the power business. Based on the aforementioned multidimensional business service quality general requirement model, a heterogeneous semantic mapping matrix is ​​constructed. The heterogeneous network service quality labels are input into the heterogeneous semantic mapping matrix for mapping, generating an initial service requirement vector. The initial business requirement vector is normalized to generate a standard business requirement vector.

[0008] This invention establishes a standardized conversion process to address the semantic incompatibility of QoS tags in heterogeneous power networks. Based on latency, jitter, packet loss rate, and guaranteed bandwidth, a multi-dimensional universal requirement model is constructed. A corresponding heterogeneous semantic mapping matrix completes the unified parsing and mapping of different network tags, which is then normalized to obtain a standardized business requirement vector. This unifies the expression of transmission requirements for various power services, eliminates cross-network semantic barriers, and outputs requirement data with a unified format and aligned dimensions. This provides a regularized input for subsequent path matching and adaptive weight adjustment, improving the accuracy and stability of cross-domain power transmission scheduling calculations.

[0009] In one optional implementation, the step of collecting local queue operation metrics of each candidate sending queue within the cross-domain gateway and using the local queue operation metrics to calculate a nonlinear congestion penalty value includes: The local queue operation indicators of each candidate sending queue within the cross-domain gateway are collected in parallel with a microsecond-level sampling period; wherein, the local queue operation indicators include instantaneous load ratio, queue depth change rate and scheduling time slot utilization rate; The business standard demand vector is used to adjust the preset Gaussian membership function, and the instantaneous load ratio, the queuing depth change rate and the scheduling time slot utilization rate are fuzzified based on the adjusted Gaussian membership function to generate corresponding fuzzy membership degrees. Substitute each of the fuzzy membership degrees into the preset multidimensional fuzzy condition control rules to perform fuzzy deduction and output a fuzzy set of queue congestion status. The fuzzy set of queue congestion status is defuzzified using the centroid method to generate the effective congestion pressure value of the corresponding queue. The effective congestion pressure value is input into a preset dual-threshold nonlinear hardening congestion penalty function to generate a nonlinear congestion penalty value corresponding to each candidate sending queue.

[0010] This invention employs microsecond-level parallel sampling to accurately capture multi-dimensional operational indicators of queues. It adaptively adjusts the Gaussian membership function to fuzzify these indicators based on power business needs. A congestion fuzzy set is derived using multi-dimensional fuzzy rules, and the fuzzy output is quantified using the centroid method to defuzzify the fuzzy set. Furthermore, a congestion penalty value is calculated using a dual-threshold nonlinear hardening function. This allows for precise differentiation of different levels of queue congestion, mitigating the impact of minor congestion while amplifying the constraints of severe congestion. It accurately quantifies queue transmission losses, providing a reliable basis for subsequent path adaptation scoring and enhancing the congestion perception sensitivity of cross-domain power traffic scheduling.

[0011] In one optional implementation, the evaluation of the path states of multiple candidate transmission paths of the cross-domain gateway, and the construction of a global path standard state matrix based on the evaluation results, includes: Collect raw transmission metrics of multiple candidate transmission paths of the cross-domain gateway; wherein, the raw transmission metrics include path latency metrics, data packet loss rate metrics, and available guaranteed bandwidth metrics; The path delay index, data packet loss rate index and available guaranteed bandwidth index of each candidate transmission path are preprocessed using a preset normalization function to generate standardized path status data. The path status of the corresponding candidate transmission path is evaluated based on the path latency index, the data packet loss rate index, and the available guaranteed bandwidth index. Based on the path status evaluation results, a global path standard status matrix is ​​constructed using the standardized path status data corresponding to each candidate transmission path.

[0012] This invention collects three types of raw transmission metrics—latency, packet loss rate, and available bandwidth—from each candidate path. After unified normalization preprocessing to obtain standardized path data, a multi-dimensional comprehensive evaluation of path status is performed, ultimately constructing a global path standard status matrix. This unifies the differences in the dimensions of various link metrics, comprehensively and systematically stores the real-time operating status of multiple transmission paths, intuitively quantifies the transmission carrying capacity of each link, eliminates calculation biases caused by inconsistent metric dimensions, and provides a unified and reliable link data foundation for subsequent fuzzy vector inner product matching and comprehensive adaptation score calculation.

[0013] In one optional implementation, the step of jointly adapting the business standard demand vector, the nonlinear congestion penalty value, and the global path standard state matrix to generate a comprehensive path adaptation score includes: Extract the basic transmission dimension requirements information corresponding to the end-to-end latency requirements, latency jitter requirements, maximum allowable packet loss rate, and service guarantee bandwidth requirements from the business standard requirement vector. The global path standard state matrix and the basic requirement information of each transmission dimension are input into a preset business scenario decoupling adaptive adjustment model, and the initial dimension weight coefficients corresponding to the basic requirement information of each transmission dimension are output. Based on the nonlinear congestion penalty value, the initial dimension weight coefficients are dynamically corrected and normalized to generate adaptive business demand weights. The macroscopic theoretical matching degree of a single transmission path is calculated by using the adaptive service demand weights and the standardized path state data of each candidate transmission path in the global path standard state matrix. The nonlinear congestion penalty value is converted into a link pressure exponential decay factor, and the link pressure exponential decay factor is used to attenuate and correct the macroscopic theoretical matching degree of each single transmission path, thereby generating a path comprehensive adaptation score for each candidate transmission path.

[0014] This invention generates initial dimensional weights based on a business scenario decoupling model, and dynamically corrects and normalizes these weights using queue nonlinear congestion penalty values ​​to obtain adaptive weights. It calculates the macroscopic matching degree of the path through fuzzy vector inner product, and then introduces a pressure attenuation factor corresponding to congestion to complete the score correction, taking into account both the differentiated needs of power services and the real-time congestion status of links. It can dynamically adjust the weights of dimensions such as latency and bandwidth, accurately quantifying the comprehensive adaptability of each link, avoiding the allocation of excessive power service traffic to congested links, and providing an objective and reliable scoring basis for multi-path elastic traffic scheduling.

[0015] In one optional implementation, determining the path traffic elastic allocation ratio for each transmission path based on the path comprehensive adaptation score, and performing traffic scheduling according to the path traffic elastic allocation ratio, includes: Based on the preset variable structure fuzzy membership function, the path comprehensive adaptation score of each candidate transmission path is converted into the corresponding path traffic elastic allocation ratio. Traffic scheduling is performed according to the elastic traffic allocation ratio for each path. Real-time monitoring of service transmission performance deviations after traffic scheduling, and correction of the parameters of the heterogeneous semantic mapping matrix based on the service transmission performance deviations.

[0016] This invention utilizes a variable-structure fuzzy membership function to convert path adaptation scores into elastic traffic allocation ratios, enabling dynamic distribution of power data streams based on link adaptation capabilities. It synchronously monitors transmission performance deviations after scheduling online and iteratively corrects the parameters of the heterogeneous semantic mapping matrix, forming a closed-loop mechanism of demand mapping, path scoring, traffic scheduling, and parameter self-correction. This continuously adapts to dynamic changes in power services and network operating condition fluctuations, constantly reducing transmission performance errors and improving the balance, reliability, and scheduling adaptability of cross-domain power data transmission in the long term.

[0017] Secondly, the present invention provides a cross-domain quality of service mapping apparatus, the apparatus comprising: The mapping module is used to extract the heterogeneous network service quality tags from the packets corresponding to the original power data stream when the original power data stream enters the cross-domain gateway, and then map them to generate a business standard requirement vector. The acquisition module is used to acquire local queue operation indicators of each candidate sending queue within the cross-domain gateway, and to calculate nonlinear congestion penalty values ​​using the local queue operation indicators. The evaluation module is used to evaluate the path status of multiple candidate transmission paths of the cross-domain gateway and construct a global path standard status matrix based on the evaluation results. The calculation module is used to perform joint adaptation calculation using the business standard demand vector, the nonlinear congestion penalty value, and the global path standard state matrix to generate a comprehensive path adaptation score. The execution module is used to determine the path traffic elastic allocation ratio corresponding to each transmission path based on the path comprehensive adaptation score, and to perform traffic scheduling according to the path traffic elastic allocation ratio.

[0018] In one optional implementation, the mapping module includes: The extraction unit is used to extract the heterogeneous network service quality tag from the packet corresponding to the original power data stream when the original power data stream enters the cross-domain gateway. The model building unit is used to construct a multi-dimensional general service quality requirement model by adopting the end-to-end latency requirements, latency jitter requirements, maximum allowable packet loss rate, and service guarantee bandwidth requirements of power services. The matrix construction unit is used to construct a heterogeneous semantic mapping matrix based on the multidimensional business service quality general requirement model. The mapping unit is used to input the heterogeneous network service quality label into the heterogeneous semantic mapping matrix for mapping, and generate an initial service requirement vector. The transformation unit is used to normalize the initial business requirement vector to generate a standard business requirement vector.

[0019] In one optional implementation, the acquisition module includes: The acquisition unit is used to collect local queue operation indicators of each candidate sending queue inside the cross-domain gateway in parallel with a microsecond-level sampling period; wherein, the local queue operation indicators include instantaneous load ratio, queue depth change rate and scheduling time slot utilization rate; The adjustment unit is used to adjust the preset Gaussian membership function using the business standard demand vector, and to perform fuzzification processing on the instantaneous load ratio, the queuing depth change rate and the scheduling time slot utilization rate based on the adjusted Gaussian membership function, so as to generate the corresponding fuzzy membership degree. The deduction unit is used to substitute each of the fuzzy membership degrees into the preset multi-dimensional fuzzy condition control rules to perform fuzzy deduction and output the fuzzy set of queue congestion status. The processing unit is used to defuzzify the fuzzy set of queue congestion states using the centroid method to generate the effective congestion pressure value of the corresponding queue. The input unit is used to input the effective congestion pressure value into a preset dual-threshold nonlinear hardened congestion penalty function to generate a nonlinear congestion penalty value corresponding to each candidate sending queue.

[0020] In one optional implementation, the evaluation module includes: The data acquisition unit is used to collect the original transmission metrics of multiple candidate transmission paths of the cross-domain gateway; wherein, the original transmission metrics include path latency metrics, data packet loss rate metrics, and available guaranteed bandwidth metrics. The preprocessing unit is used to preprocess the path delay index, data packet loss rate index and available guaranteed bandwidth index of each candidate transmission path using a preset normalization function to generate standardized path status data. The evaluation unit is used to evaluate the path status of the corresponding candidate transmission path based on the path delay index, the data packet loss rate index and the available guaranteed bandwidth index. The combining unit is used to combine the path state evaluation results and construct a global path standard state matrix using the standardized path state data corresponding to each candidate transmission path. Attached Figure Description

[0021] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0022] Figure 1 This is a schematic diagram of an application system corresponding to the cross-domain quality of service mapping method according to an embodiment of the present invention; Figure 2 This is a logical architecture diagram of the application system corresponding to the cross-domain quality of service mapping method according to an embodiment of the present invention. Figure 3 This is a schematic diagram of the first type of cross-domain quality of service mapping method according to an embodiment of the present invention; Figure 4 This is a schematic diagram of the second process of the cross-domain quality of service mapping method according to an embodiment of the present invention; Figure 5 This is a structural block diagram of a cross-domain quality of service mapping device according to an embodiment of the present invention. Detailed Implementation

[0023] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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, 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.

[0024] It is understood that before using the technical solutions disclosed in the various embodiments of the present invention, users should be informed of the types, scope of use, and usage scenarios of the personal information involved in the present invention and their authorization should be obtained in accordance with relevant laws and regulations through appropriate means.

[0025] The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0026] This invention provides a cross-domain Quality of Service (QoS) mapping method. It achieves end-to-end collaborative optimization in heterogeneous cross-domain power transmission scenarios. By mapping message QoS tags, it generates a service standard demand vector, accurately matching the differentiated transmission requirements of power services, such as latency and bandwidth. It calculates nonlinear congestion penalty values ​​based on queue operation indicators to quantify transmission losses caused by queue congestion. A global path standard state matrix is ​​constructed by combining multi-path states to fully characterize the real-time operating conditions of each transmission link. A comprehensive path adaptation score is obtained through adaptive weight adjustment, fuzzy vector inner product matching, and congestion attenuation correction, which is used to dynamically allocate traffic proportions for scheduling. This effectively balances the load on multiple links, reduces power data transmission latency and packet loss risk, takes into account the different QoS requirements of different power services, and improves the stability of cross-domain power data stream transmission and network resource utilization.

[0027] This method is applied to a cross-domain Quality of Service (QoS) mapping system based on fuzzy membership functions. This system is deployed at the gateway node of a cross-domain network. A system diagram is shown below. Figure 1 As shown. The system includes various service terminals, TSN network, 5G access network, SPN network, cross-domain deterministic gateway, etc., among which the cross-domain deterministic gateway realizes functions such as cross-domain interconnection, protocol conversion, and service mapping of different networks.

[0028] System logical architecture diagram as follows Figure 2 As shown, it is divided into three layers: Network Status and Service Awareness Layer: Deployed on the gateway's input interface and transmission link, this layer serves as the underlying data source for the entire system, used to capture service characteristics and network physical status in real time and in multiple dimensions. This layer specifically includes the following modules: Business tag extraction module: Deployed on the input interface side of the gateway, it uses technologies such as deep packet inspection to parse the header of the input power data stream and dynamically extract the original heterogeneous protocol tags to achieve a preliminary understanding of the original business requirements.

[0029] Local queue status sampling module: Deployed inside the gateway, it samples key operational indicators such as instantaneous load percentage, queuing depth change rate, and scheduling time slot utilization rate of each candidate sending queue inside the gateway at a high frequency of microseconds.

[0030] Global network telemetry probe: Based on in-band network telemetry (INT) technology, it is deployed on cross-domain transmission links to dynamically collect and sense global deterministic state indicators on each multi-hop candidate path and construct a macro-state matrix.

[0031] Multidimensional Fuzzy Mapping Decision Layer: Deployed in the core processing unit of the gateway, serving as the central computing logic core of the system, it is responsible for transforming the discrete, multidimensional physical indicators perceived by the lower layers into service-aligned quantitative decision scores. This layer specifically includes the following modules: Demand Normalization Matrix Transformation Module: This module converts discrete network service protocol labels into continuous physical boundary values ​​through a heterogeneous semantic mapping matrix, and performs a dimensionless, decoupled, nonlinear normalization transformation to generate a standard spatial demand vector.

[0032] Queue congestion fuzzy inference engine: Taking the standard spatial demand vector as the base input, it uses the Gaussian membership function to dynamically modulate the local queue index, and outputs the effective congestion pressure value and nonlinear congestion penalty value based on the multi-dimensional fuzzy condition control rule base and defuzzification calculation.

[0033] Dynamic mutation triggering and adaptive weight adjustment module: Based on time sliding window differential gradient, it captures the time delay mutation characteristics of the global path and introduces local congestion penalty value as reverse constraint, thereby adaptively reconfiguring and adjusting the demand weight vector of the business scenario.

[0034] The comprehensive adaptation score generation module introduces the Euclidean proximity model from fuzzy mathematics to calculate the macro-theoretical matching degree between the actual capability of the path and the ideal business requirements, and couples the pressure exponential decay factor to generate the final comprehensive adaptation score for each candidate path.

[0035] Dynamic traffic splitting and multi-queue execution layer: Deployed on the gateway's output interface side, it serves as the closed-loop implementation and execution mechanism for system decisions, responsible for elastically and adaptively distributing traffic based on the adaptation score output by the decision layer. This layer specifically includes the following modules: Variable structure membership traffic ratio generation module: Introduces an adaptive variable structure adjustment factor, performs variable structure intervention according to the network load status (light load stable period or sudden congestion period), and transforms the continuous comprehensive adaptation score into the traffic elasticity ratio vector that each candidate path should carry.

[0036] Hash / Dynamic Token Bucket Flow Splitting Executor: Receives the elastic proportional vector of traffic, uses a hash scheduling algorithm to dynamically split the aggregated power business flow into multiple sub-flows and delivers them to the corresponding queues. At the same time, it configures a dynamic token bucket mechanism to constrain the traffic throughput rate, ensures that data packets are transmitted in order, and uses online closed-loop correction mapping matrix parameters based on the actual performance deviation at the egress point.

[0037] This embodiment provides a cross-domain quality of service mapping method. Figure 3 This is a flowchart of a cross-domain quality of service mapping method according to an embodiment of the present invention, such as... Figure 3 As shown, the process includes the following steps: Step S101: When the original power data stream enters the cross-domain gateway, extract the heterogeneous network service quality label from the message corresponding to the original power data stream and map it to generate a business standard requirement vector.

[0038] In this embodiment of the invention, after the original power data stream is accessed by the cross-domain gateway, the gateway parses the data packet header and extracts the built-in heterogeneous network service quality label. It then calls the pre-built multi-dimensional business service quality general requirement model and heterogeneous semantic mapping matrix to perform cross-network semantic unified transformation on the label to obtain the initial business requirement vector. The initial business requirement vector is then normalized and transformed to unify the dimensions of each requirement dimension, thereby generating a business standard requirement vector with regular dimensions and uniform format.

[0039] Step S102: Collect local queue operation indicators of each candidate sending queue within the cross-domain gateway, and use the local queue operation indicators to calculate the nonlinear congestion penalty value.

[0040] In this embodiment of the invention, three types of local queue operation indicators—instantaneous load ratio, queue depth change rate, and scheduling time slot utilization rate—are collected in parallel with a microsecond-level sampling period for all candidate sending queues within the gateway. The indicators are fuzzified by adjusting the Gaussian membership function in conjunction with the business standard demand vector. The effective congestion pressure value is obtained through multi-dimensional fuzzy rule deduction and centroid method defuzzification. Then, the nonlinear congestion penalty value corresponding to each candidate sending queue is obtained by inputting a double-threshold nonlinear hardening congestion penalty function.

[0041] Step S103: Evaluate the path status of multiple candidate transmission paths of the cross-domain gateway, and construct a global path standard state matrix based on the evaluation results.

[0042] In this embodiment of the invention, the path latency, data packet loss rate, and available guaranteed bandwidth of all candidate transmission paths are collected in real time. The dimensions of each indicator are uniformly processed by a preset normalization function to obtain standardized path status data. Based on the three types of standardized indicators, the transmission status of each path is comprehensively evaluated. The standardized path status data corresponding to all candidate paths are integrated and uniformly arranged to form a global path standard status matrix.

[0043] Step S104: Joint adaptation calculation is performed using the business standard demand vector, nonlinear congestion penalty value, and global path standard state matrix to generate a comprehensive path adaptation score.

[0044] In this embodiment of the invention, basic requirement information for each transmission dimension is extracted from the business standard requirement vector. The initial dimension weight coefficients are obtained by combining the global path standard state matrix with the business scenario decoupling adaptive adjustment model. The weight correction and normalization are completed using nonlinear congestion penalty values ​​to obtain adaptive business requirement weights. The macroscopic theoretical matching degree is calculated based on the adaptive weights and the standardized state data of each path. The nonlinear congestion penalty value is converted into a pressure exponential decay factor to correct the matching degree. Finally, the comprehensive path adaptation score corresponding to each candidate transmission path is obtained.

[0045] Step S105: Based on the path comprehensive adaptation score, determine the path traffic elastic allocation ratio corresponding to each transmission path, and perform traffic scheduling according to the path traffic elastic allocation ratio.

[0046] In this embodiment of the invention, based on the path comprehensive adaptation score of each candidate transmission path, the path traffic elastic allocation ratio corresponding to each path is obtained by converting it through a preset variable structure fuzzy membership function. Based on the allocation ratio, the total traffic of power services is dynamically diverted and coordinated for scheduling. At the same time, the service transmission performance deviation after scheduling is monitored in real time, and the heterogeneous semantic mapping matrix parameters are iteratively corrected based on the performance deviation to realize the dynamic optimization closed loop of the traffic scheduling strategy.

[0047] A cross-domain gateway refers to a forwarding device deployed at the boundary of a heterogeneous power network. It is responsible for receiving various types of raw power data streams and completing data forwarding between different networks, parsing business requirements, and unified scheduling of multi-path traffic.

[0048] Raw power data streams refer to the real-time acquisition, control, and remote signaling / telemetry data generated by power business equipment such as substations, distribution terminals, and dispatching master stations.

[0049] Heterogeneous network service quality labels refer to QoS identifiers carried in messages of different communication network protocols. The semantics and value rules of each network label are not uniform.

[0050] The business standard requirement vector refers to a multi-dimensional standardized vector formed after heterogeneous semantic unified mapping and normalization processing. The dimensions include latency, jitter, packet loss rate, and guaranteed bandwidth.

[0051] Candidate transmission queues refer to multiple buffer queues within a cross-domain gateway used to cache power data streams to be forwarded, with different queues corresponding to different transmission paths.

[0052] Local queue operation metrics refer to real-time queue operation parameters obtained by microsecond-level sampling, including instantaneous load percentage, queue depth change rate, and scheduling slot utilization.

[0053] The nonlinear congestion penalty value refers to the numerical value quantized based on fuzzy inference and a nonlinear hardening function.

[0054] Candidate transmission paths refer to multiple optional communication links between the cross-domain gateway and the peer device that can be used to transmit power data.

[0055] The global path standard state matrix refers to a two-dimensional standardized matrix constructed by integrating the normalized preprocessing of the latency, packet loss rate, and available bandwidth indicators of each link.

[0056] The path comprehensive adaptation score refers to the quantitative score calculated by integrating business needs, real-time link status, and queue congestion level. The higher the score, the better the link matches the current power business transmission needs.

[0057] The path traffic elastic allocation ratio refers to the traffic weight ratio obtained by converting the path comprehensive adaptation score through a variable structure fuzzy membership function.

[0058] Traffic scheduling refers to a closed-loop management process that distributes power data streams to various candidate transmission paths based on the elastic allocation ratio of traffic, and iteratively optimizes the front-end semantic mapping parameters based on feedback from transmission performance deviations.

[0059] This embodiment provides a cross-domain quality of service mapping method. Figure 4 This is a flowchart of a cross-domain quality of service mapping method according to an embodiment of the present invention, such as... Figure 4 As shown, the process includes the following steps: Step S201: When the original power data stream enters the cross-domain gateway, extract the heterogeneous network service quality label from the message corresponding to the original power data stream and map it to generate a business standard requirement vector.

[0060] Specifically, step S201 includes: Step S2011: When the original power data stream enters the cross-domain gateway, extract the heterogeneous network service quality label from the message corresponding to the original power data stream.

[0061] In this embodiment of the invention, when the original power data stream (such as 5G slicing service or TSN service) enters the cross-domain gateway, the gateway parses the header fields of the data packet and extracts the heterogeneous network service quality label used to identify the service transmission performance requirements.

[0062] Step S2012: Using the end-to-end latency requirements, latency jitter requirements, maximum allowable packet loss rate, and service guarantee bandwidth requirements of the power business, a multi-dimensional general service quality requirement model is constructed.

[0063] In this embodiment of the invention, a multi-dimensional service QoS general requirement model is constructed: an N-dimensional continuous orthogonal geometric space is constructed, called the general requirement feature space. This is used to represent the communication needs of different power services. In this embodiment, N=4 is set, and the core parsing dimension is defined as:

[0064] In the formula, This indicates the end-to-end latency requirements of the business. This indicates the latency jitter requirement. Indicates the maximum allowable packet loss rate. This indicates the bandwidth requirements for service assurance. The gateway's protocol feature extraction module uses Deep Packet Inspection (DPI) technology to parse the header of the input data packet and dynamically extract its original heterogeneous protocol tags. T(Such as the 5QI identifier for 5G networks, the PCP priority field for TSN networks, and the SFI identifier for SPN).

[0065] Step S2013: Based on the multidimensional business service quality general requirement model, construct a heterogeneous semantic mapping matrix.

[0066] In this embodiment of the invention, in order to convert discrete network service protocol tags that cannot be directly mathematically calculated into continuous physical boundary values ​​in a general requirement feature space, a heterogeneous semantic set T is defined:

[0067] This set contains all heterogeneous priority tags that may enter the gateway on cross-domain links. (Set elements...) Indicates the first k A specific type of online tag.

[0068] Based on this, and according to the Service Level Agreements (SLAs) committed to each label in international standards and specifications (including 3GPP specifications, IEEE standards, and power service protocols), a [system / mechanism] is constructed. Fixed heterogeneous semantic mapping matrix of dimension :

[0069] Each row in the matrix corresponds to a protocol label of a heterogeneous network, and each column corresponds to a communication performance requirement dimension. The element values ​​represent the worst-case tolerance limit declared by the label to the underlying network.

[0070] Heterogeneous semantic mapping matrix The construction method is as follows: Assume that the input heterogeneous tag set contains three typical power tags: .

[0071] Mapping matrix The mathematical structure is constructed as follows:

[0072] Step S2014: Input the heterogeneous network service quality labels into the heterogeneous semantic mapping matrix for mapping to generate the initial service requirement vector.

[0073] In this embodiment of the invention, when a data stream with a specific label arrives, the data stream with the specific label is converted into a one-hot encoded feature vector. (Where the label corresponds to position 1), the corresponding explicit multidimensional original demand vector is calculated through mapping matrix transformation. (i.e., the initial business requirement vector):

[0074] In the formula, This is the one-hot encoded feature vector corresponding to the original heterogeneous protocol label. For pre-injection based on multi-domain collaboration standards A fixed-dimensional heterogeneous semantic mapping matrix, where the element values ​​represent the worst-case tolerance limits declared by the labels to the underlying network.

[0075] When with When the power control flow of the tag enters the gateway, its one-hot encoded vector is: .

[0076] The mapping result is:

[0077] In this way, the matrix accurately transforms discrete labels into specific physical requirement numbers.

[0078] By transforming the mapping matrix, discrete network protocol labels are mapped to cross-domain, continuous service performance requirement boundary characteristics. The parameters in the mapping matrix are pre-injected by multi-domain collaborative standards, enabling cross-domain gateways to have transparent awareness of underlying service performance requirements without disrupting the original network protocol stack.

[0079] It is worth mentioning that the mapping matrix is ​​a "computable eigenvector transformation" rather than a "fixed-value correspondence table," which is significantly different from the traditional hard lookup table mapping. It forms the basis of the fuzzy comprehensive evaluation model, realizing semantic decoupling and mapping from discrete protocol labels to continuous physical space, and solving the problem that hard lookup table methods cannot handle real-time dynamic environmental deviations.

[0080] Step S2015: Normalize the initial business requirement vector to generate a standard business requirement vector.

[0081] In this embodiment of the invention, since the mapped original physical indicators have different physical dimensions and orders of magnitude, they cannot be directly subjected to joint fuzzy logic operations. Therefore, the system must process the original feature vectors. The dimensionless requirement vector is then mapped to a dimensionless standard spatial requirement vector (i.e., a business standard requirement vector). Its value is strictly limited to the interval [0, 1].

[0082] To address the physical evolution characteristics of different indicators, the system employs a piecewise nonlinear mapping function for normalization.

[0083] Normalization of cost indicators: Cost metrics include latency Shaking Packet loss rate The physical law governing these indicators is that the smaller the value, the higher the requirement for business certainty. The system uses a negative exponential normalization function:

[0084] In the formula, and These represent the absolute maximum tolerable failure limit and the ideal physical transmission limit for power services in the current heterogeneous network domain, respectively. Through this nonlinear transformation, the more stringent the service's requirements for latency or packet loss, the better. The closer the value is to 1.

[0085] Normalization of benefit-type indicators: Benefit-oriented indicators refer to bandwidth For these types of metrics, higher values ​​indicate greater throughput resource consumption. The system employs a logarithmic saturation normalization function:

[0086] In the formula, The maximum rated bandwidth for the physical interface of the cross-domain egress link is determined. Utilizing the non-linear smoothing properties of the logarithmic function, it can sensitively amplify the characteristic differences of low-bandwidth core control flows while preventing the abrupt long-tail effect of characteristic values ​​caused by large-volume data (such as video surveillance).

[0087] After the above normalization transformation, a general requirement feature vector is finally generated in the unified standard space of [0,1]. This serves as the baseline input for subsequent multidimensional fuzzy membership degree joint adaptation algorithms.

[0088] Step S202: Collect local queue operation indicators of each candidate sending queue within the cross-domain gateway, and use the local queue operation indicators to calculate the nonlinear congestion penalty value.

[0089] Specifically, step S202 includes: Step S2021: Collect local queue operation indicators of each candidate sending queue inside the cross-domain gateway in parallel with a microsecond sampling period; among which, the local queue operation indicators include instantaneous load ratio, queuing depth change rate and scheduling time slot utilization rate.

[0090] In this embodiment of the invention, to prevent high-priority queues from being blindly filled by critical boundary services, leading to hard packet loss of sudden critical service traffic, the invention introduces a queue load perception and penalty mechanism within the gateway node. This is achieved by non-linearly coupling multi-dimensional queue status indicators through fuzzy logic. Specifically: The network status and service awareness layer uses in-band network telemetry probes to sample each candidate transmission queue in the current gateway in parallel at microsecond intervals. Three core operational indicators: Instantaneous load percentage (Value range [0,1]): Represents the ratio of the memory space currently occupied by the queue buffer to the total allocated space.

[0091] Queue depth change rate (Value range [-1, 1]): Represents the slope of the queue length change within adjacent sampling periods. This indicates that traffic is backlogged. This indicates that the flow is dissipating.

[0092] Scheduling slot utilization (Value range [0,1]): This indicates the proportion of transmission time actually occupied by the queue within the current scheduling period (such as the TDM slot of SPN or the gated period of TSN).

[0093] Step S2022: Adjust the preset Gaussian membership function using the business standard demand vector, and perform fuzzification processing on the instantaneous load ratio, queuing depth change rate and scheduling slot utilization rate based on the adjusted Gaussian membership function to generate the corresponding fuzzy membership.

[0094] In this embodiment of the invention, the measured hard physical indicators are mapped to corresponding low, medium and high fuzzy label membership values ​​using Gaussian membership functions, which are then input into the fuzzy inference engine to activate the corresponding collaborative control rules. Finally, the centroid method is used to defuzzify the output fuzzy set to obtain a continuous and smooth effective congestion pressure value.

[0095] Specifically, given the nonlinear collaborative influence of the three-dimensional metrics of the aforementioned candidate sending queues on the actual risk of queue collapse, and the limitations of traditional fixed-weight linear weighting algorithms, this invention constructs a fuzzy logic-based reasoning mechanism to calculate the effective congestion pressure value. : The system is , and Three fuzzy set labels are defined: Low / Negative, Medium, and High / Positive. A Gaussian membership function is used to map the measured deterministic physical indicators to the corresponding membership degree values. To ensure that the effective congestion pressure inference of the local sending queue can accurately perceive the Service Level Agreement (SLA) requirements of multi-dimensional services, a unified standard spatial demand vector output in step S201 is used. (Corresponding to latency, jitter, packet loss, and bandwidth requirements) is the dynamic modulation factor used as a Gaussian membership function.

[0096] In terms of physical mechanisms, different communication needs have different sensitivities to queue collapse risk. This invention decouples them into two sets of adjustment paths: Latency and jitter sensitive paths (adjustment) and (Fuzzy boundaries): When the business is sensitive to latency and shaking When the requirements are extremely high, even a slight backlog in the local buffer can lead to end-to-end deterministic timeouts or jitter exceeding limits. Therefore, the load metric must be made more sensitive to triggering the "High" state.

[0097] Packet loss and bandwidth throughput path (adjustment) (Fuzzy boundary): When the business has a high packet loss rate and ensure bandwidth When the requirements are extremely high, it means that if the scheduling time slots are preempted or the utilization rate is too high, hard packet loss will occur. Therefore, the time slot indicators must be made more sensitive to trigger the "High" state.

[0098] Based on the above collaborative mechanism, the modified Gaussian membership function is constructed as follows: 1) Regarding the percentage of instantaneous load and queue depth change rate Functional correction mapping to the fuzzy set "High":

[0099]

[0100] In the formula, and The standard deviation is the preset Gaussian function. This is called the latency-jitter demand constraint sensitivity factor, and its mathematical expression is:

[0101] In the formula, This is the sensitivity adjustment coefficient (value range [0.1, 0.25]). , The relative correlation weight between latency and jitter satisfies... .

[0102] When the power service input in step S201 belongs to the protection and control category or other streams with extremely stringent requirements for latency and jitter... and The value approaches 1, leading to Significantly reduced (width shrinkage of Gaussian distribution).

[0103] At this time, even the real-time load collected by the hardware layer or the rate of change of backlog If it is in the physical medium range, it will be judged to have a very high "High" membership degree during the fuzzing stage due to the centripetal contraction of the curve, thus activating the congestion avoidance rules in Table 1 in advance.

[0104] 2) Regarding the utilization rate of scheduling time slots Functional correction mapping to the fuzzy set "High":

[0105] In the formula, The standard deviation is the preset Gaussian standard deviation. This is called the packet loss-throughput requirement constraint sensitivity factor, and its mathematical expression is:

[0106] In the formula, This is the adjustment coefficient (range [0.1, 0.2]). , As the relative weight of packet loss rate and bandwidth requirement, satisfying .

[0107] If the current business is extremely sensitive to packet loss ( ) or belongs to high-throughput bandwidth services ( ), This will decrease, leading to a reduction in scheduling slot utilization. The "High" membership curve contracts synchronously. This design ensures that even when high-bandwidth or high-reliability services encounter minor transmission slot constraints, the system can readily detect the risk and improve the final effective congestion pressure value. .

[0108] Step S2023: Substitute each fuzzy membership degree into the preset multidimensional fuzzy condition control rules to perform fuzzy deduction and output the fuzzy set of queue congestion state.

[0109] In this embodiment of the invention, to reflect the dynamic synergy between indicators, the fuzzy inference engine constructs a multi-dimensional fuzzy conditional control rule base based on the principle of "trend prediction and collaborative punishment". Its core control logic matrix is ​​shown in Table 1: Table 1. Effective congestion pressure fuzzy inference rule matrix

[0110] Based on the above fuzzy deduction results, the fuzzy set of queue congestion status, which integrates multi-dimensional queue operation status information, is output to represent the degree of queue congestion. This facilitates subsequent defuzzification processing using the centroid method to obtain a quantified effective congestion pressure value.

[0111] Step S2024: Defuzzify the fuzzy set of queue congestion status using the centroid method to generate the effective congestion pressure value of the corresponding queue.

[0112] In this embodiment of the invention, after the inference engine activates the corresponding rule, it uses the centroid method to defuzzify the output fuzzy set, and finally outputs an effective congestion pressure value that is non-linearly decoupled and can truly reflect the future deterministic boundary risk of the queue. The centroid method is used here for defuzzification, aiming to achieve a weighted synthesis of the outputs of each fuzzy rule, thereby avoiding control abrupt changes caused by a single rule trigger. This ensures the continuity and smoothness of the effective congestion pressure value under small perturbations of the input indicators, which is crucial for ensuring the deterministic stability of power services during cross-domain forwarding.

[0113] Step S2025: Input the effective congestion pressure value into the preset dual-threshold nonlinear hardening congestion penalty function to generate the nonlinear congestion penalty value corresponding to each candidate sending queue.

[0114] In this embodiment of the invention, after obtaining the effective congestion pressure Subsequently, in order to provide a gradual early warning in the early stages of queue deterioration and a sharp suppression when overflow is imminent, a nonlinear hardening congestion penalty function based on dual thresholds was constructed. :

[0115] In the formula, The set soft safety load threshold (in this embodiment, the value range is [0.5, 0.65]) is used. Below this threshold, the network is in a light load stable period, and the penalty term is 0. A hard crash threshold is set for the system (in this embodiment, the value ranges from [0.85, 0.95]). Once the effective congestion pressure exceeds this value, the penalty score is directly locked to the maximum saturation limit of 1, forcibly rejecting the injection of any new service flow. This is the soft safety load threshold. and hard crash threshold Pre-configure the static rated capacity based on the current gateway physical port's cache queue depth; n It is a non-linear penalty exponent (in this embodiment, it is set to a cubic leap penalty of n=3).

[0116] This step distinguishes between mild, moderate, and severe congestion states of the queue using a dual-threshold-based nonlinear hardening congestion penalty function and performs nonlinear hardening correction. It accurately quantifies the congestion suppression degree of different queues and ultimately outputs the nonlinear congestion penalty value corresponding to each candidate sending queue. This will be used as a reverse constraint term and directly input into the subsequent step S203 to participate in the multi-dimensional quality adaptation decision calculation, thereby dynamically reducing the comprehensive adaptation weight of the congestion queue and forcing traffic to elastically overflow and load-sharing to the less loaded suboptimal queue.

[0117] Step S203: Evaluate the path status of multiple candidate transmission paths of the cross-domain gateway, and construct a global path standard state matrix based on the evaluation results.

[0118] Specifically, step S203 includes: Step S2031: Collect the original transmission metrics of multiple candidate transmission paths of the cross-domain gateway; wherein, the original transmission metrics include path latency metrics, data packet loss rate metrics, and available guaranteed bandwidth metrics.

[0119] In this embodiment of the invention, for the M multi-hop candidate transmission paths existing in an end-to-end cross-domain heterogeneous network, the system evaluates the state matrix of each path in real time and adaptively adjusts the weights of each dimension according to the dynamic evolution characteristics of the service, specifically as follows: The gateway's perception layer uses in-band network telemetry (INT) technology to collect real-time data on each multi-hop candidate path. N Construct a set of global status metrics (such as path latency, data packet loss rate, and available guaranteed bandwidth). The original global state matrix of dimension 1 .OK M Represents the path selection space, columns N This represents the state characteristics of the path.

[0120] Step S2032: The path delay index, data packet loss rate index and available guaranteed bandwidth index of each candidate transmission path are preprocessed using a preset normalization function to generate standardized path status data.

[0121] In this embodiment of the invention, in order to mitigate the interference of different physical dimensions and extreme long-tailed samples on subsequent decision-making, a nonlinear metric normalization function based on the expected value of the power SLA benchmark is introduced to normalize the original matrix. X Transform into a standard matrix Its value is strictly limited to [0,1]: 1) Elastic normalization of cost-related metrics (metrics such as latency and packet loss rate, where lower is better): The system no longer uses the traditional linear range transform, but instead introduces an inverse Sigmoid saturation normalization function:

[0122] In the formula, This serves as a pre-planned baseline for the expected latency / packet loss rate for this type of service. This is a nonlinear scaling factor that controls the sensitivity of the control function. When the actual measured value... When it is better than the baseline, It rapidly approaches 1; however, once it begins to deviate from the baseline, the score will exhibit a sharp, non-linear decline, thus becoming extremely sensitive to link deterioration.

[0123] 2) Elastic normalization of benefit-oriented indicators (indicators where larger available bandwidth is better): Employing a saturation-type fine-tuning function:

[0124] In the formula, This represents the maximum bandwidth requirement for the aggregated power service flows currently entering the gateway. This is the shape adjustment coefficient. This function ensures that once the available bandwidth of a path meets the basic business requirements, the marginal score benefit diminishes, preventing the high bandwidth of a single path from blindly masking its latency degradation.

[0125] Traditional Min-Max normalization is linear and cannot reflect the varying sensitivities of power grid operations to latency jitter. In contrast, the elastic normalization function of this invention (such as the inverse Sigmoid transform) introduces a baseline expected value and a nonlinear sensitivity factor, achieving a threshold amplification effect on critical grid control flows near critical thresholds. This enables the algorithm not only to identify changes in network indicators but also to determine the severity of the threat these changes pose to grid operation security.

[0126] Step S2033: Evaluate the path status of the corresponding candidate transmission path based on path delay index, data packet loss rate index, and available guaranteed bandwidth index.

[0127] In this embodiment of the invention, based on three types of indicators—path latency, data packet loss rate, and available guaranteed bandwidth—the real-time operating status and transmission performance of candidate transmission paths are comprehensively evaluated from multiple dimensions, including transmission delay reliability, data integrity, and service carrying capacity. This completes the quantitative evaluation of single-path status and provides a reliable evaluation basis for subsequent integration of standardized path data and construction of a global path standard status matrix.

[0128] Step S2034: Combining the path status evaluation results, construct a global path standard status matrix using the standardized path status data corresponding to each candidate transmission path.

[0129] In this embodiment of the invention, the real-time status evaluation results of each candidate transmission path are combined, and the standardized path status data such as latency, packet loss rate, and available guaranteed bandwidth of all paths after normalization preprocessing are summarized. The data are then arranged and integrated in a unified manner according to the path dimension and the indicator dimension to construct a global path standard status matrix with unified dimensions and regular data. .

[0130] Step S204: Joint adaptation calculation is performed using the business standard demand vector, nonlinear congestion penalty value, and global path standard state matrix to generate a comprehensive path adaptation score.

[0131] Specifically, step S204 includes: Step S2041: Extract the basic transmission dimension requirements information corresponding to the end-to-end latency requirements, latency jitter requirements, maximum allowable packet loss rate, and service guarantee bandwidth requirements from the service standard requirement vector.

[0132] In this embodiment of the invention, the basic requirement information of four core transmission dimensions—end-to-end latency, latency jitter, maximum allowable packet loss rate, and service guarantee bandwidth—is accurately extracted from the business standard requirement vector. The standardized requirement parameters of each dimension are retained, providing accurate dimensional requirement input conditions for subsequent input of the business scenario decoupling adaptive adjustment model and solving the initial weight coefficients of each dimension.

[0133] Step S2042: Input the global path standard state matrix and the basic requirement information of each transmission dimension into the preset business scenario decoupling adaptive adjustment model, and output the initial dimension weight coefficients corresponding to the basic requirement information of each transmission dimension.

[0134] In this embodiment of the invention, the extracted basic requirement information of each transmission dimension and the constructed global path standard state matrix are jointly input into a pre-built business scenario decoupling adaptive adjustment model. By decoupling the business constraints and path state relationships of each transmission dimension through the model, the influence of different dimensions on power business transmission is adaptively analyzed, and the initial dimension weight coefficients corresponding to the basic requirement information of each transmission dimension are output accordingly, providing basic weight parameters for subsequent dynamic weight correction.

[0135] Specifically, a dynamic weight vector of business requirements is introduced. Its satisfaction The weight vector is adaptively reconfigured based on the current scheduling policy and the service flow's Service Level of Control (SLA). For protection and control-type service flows, which are extremely sensitive to deterministic latency, the adaptive configuration weight vector is as follows: .

[0136] For high-bandwidth data streams such as video, which have higher requirements for throughput and bandwidth guarantees, the weight vector is adaptively adjusted as follows: In the formula, This indicates the weight of end-to-end latency requirements for the business. Indicates the demand weight for latency jitter. The weight represents the maximum allowable packet loss rate. This indicates the weight of bandwidth requirements for business assurance.

[0137] To address the instantaneous degradation of link quality in heterogeneous power networks caused by topology switching, fiber optic outages, or malicious traffic attacks, a mutation trigger based on time sliding window differential gradient was deployed within the decision-making layer.

[0138] The system constructs a time-sliding window of length H (the value of H is usually configured according to the sampling period of the service and the sensitivity to latency changes; in actual engineering implementation, this value is usually set as a periodic multiple of the service's critical control frames) for the latency index of each path, and calculates the differential latency evolution gradient between adjacent windows in real time. :

[0139] Establish a dynamic mutation triggering equation:

[0140] In the formula, The variance of the delay within the current window. , The preset joint sensitivity coefficient, Control the sensitivity to the rate of change of the average time delay (gradient); Controlling the sensitivity to the degree of time delay fluctuation (variance), This represents the local congestion constraint penalty coefficient.

[0141] When the calculated path status trigger value When the communication performance fluctuation threshold is exceeded, the system automatically initiates a route state reassessment mechanism: at this time, even if the latency gradient of the remote path is exceeded... No drastic changes have occurred yet, once the nonlinear congestion penalty value of the local queue in step S202... Crossing the soft safety load threshold And exhibiting a surge during a cubic leap. This will act as a reverse constraint force, forcing... The closer it gets to exceeding the threshold, the more the system will adjust the weight of the latency metric corresponding to the deteriorated path. The adaptive weight vector W is adjusted to the preset maximum intervention weight value (e.g., 0.95). By significantly increasing the latency weight ratio in the comprehensive adaptation score calculation in step S204, the theoretical quality score of the path is strongly suppressed, guiding business traffic away from the degraded path and dynamically migrating to other available paths, so as to achieve elastic avoidance of path failures at the second or millisecond level.

[0142] Step S2043: Dynamically correct the weight coefficients of each initial dimension based on the nonlinear congestion penalty value and perform normalization processing to generate adaptive business demand weights.

[0143] In this embodiment of the invention, a nonlinear congestion penalty value is used. The equation is introduced as a reverse constraint term to reduce or adjust the latency weight of the corresponding path when the state is re-evaluated, so as to dynamically correct the weight coefficients of each initial dimension and perform normalization processing to generate the final adaptive business requirement weight.

[0144] Step S2044: The macroscopic theoretical matching degree of a single transmission path is calculated by using the adaptive service demand weight and the standardized path state data of each candidate transmission path in the global path standard state matrix.

[0145] In this embodiment of the invention, the multidimensional fuzzy mapping decision layer is used to evaluate the first... i To determine the degree to which the multi-hop candidate paths meet the current business requirements, we introduce the Euclidean proximity model from fuzzy mathematics. Specifically: The current business standard requirement vector calculated in step S201 Treating it as an ideal target fuzzy set, the first digit of the standard matrix R in step S203 is... i row vector Considered as the first i A fuzzy set of the actual capabilities provided by each path. Combined with a weight vector of business requirements. W Calculate the first i The theoretical matching score of each path to this power business :

[0146] In the formula, the actual path capability vector With business demand vector The closer the geometric distance in the weighted space, the higher the degree of fit between the resource supply and business needs of the path, and the higher the theoretical matching score. The closer it gets to 1.

[0147] Step S2045: Convert the nonlinear congestion penalty value into a link pressure exponential decay factor, and use the link pressure exponential decay factor to attenuate and correct the macroscopic theoretical matching degree of each single transmission path, thereby generating a path comprehensive adaptation score for each candidate transmission path.

[0148] In this embodiment of the invention, a comprehensive adaptation score is introduced by incorporating a link pressure exponential attenuation factor. calculate: To prevent end-to-end deterministic collapse due to local congestion within the gateway, the system utilizes the local candidate queue calculated in step S022. Non-linear congestion penalty value The theoretical matching degree is nonlinearly reduced. A comprehensive fit score based on an exponential decay factor is constructed. Calculation equation:

[0149] That is:

[0150] In the formula, The micro-constraint enhancement coefficient is generated by linking and mapping the original service labels extracted in step S201. For high-security services of protection and control type, The static configuration is set to an upper limit of 3.0 to enhance the sensitivity of the reverse constraint.

[0151] As can be seen from the above formula, the final comprehensive score is... Strictly limited to the continuous interval (0,1). When the local queue is extremely smooth (penalty value) When the local queue's effective congestion pressure exceeds the safety threshold, the overall score equals the path-theoretical matching degree; however, once the local queue's effective congestion pressure exceeds the safety threshold, the penalty value... Surge, exponential decay term This will result in a precipitous drop, forcibly reducing the overall adaptation score of the path, thereby achieving real-time reverse constraint of local load status on global forwarding decisions.

[0152] Step S205: Based on the path comprehensive adaptation score, determine the path traffic elastic allocation ratio corresponding to each transmission path, and perform traffic scheduling according to the path traffic elastic allocation ratio.

[0153] Specifically, step S205 includes: Step S2051: Based on the preset variable structure fuzzy membership function, the path comprehensive adaptation score of each candidate transmission path is converted into the corresponding path traffic elastic allocation ratio.

[0154] In this embodiment of the invention, in order to effectively improve the performance degradation and resource waste of boundary critical traffic caused by traditional hard threshold classification, the decision layer introduces a variable structure fuzzy membership function, which transforms the path comprehensive adaptation score of each consecutive candidate transmission path into the traffic elasticity ratio that each candidate path / queue should carry. :

[0155] In the formula, Topology adaptive variable structure adjustment factor set for the system Its dynamic evolution control logic is as follows: Network light load period: Decision-makers will Set the baseline value to 1. At this point, the traffic distribution ratio becomes proportional to the overall score, and the aggregated traffic is smoothly and complementaryly distributed into each available candidate path, maximizing the utilization of the surplus bandwidth resources of the multi-path heterogeneous network.

[0156] Sudden surges in traffic or periods of severe localized congestion: When the gateway detects any path... When the activation or local queue penalty value is greater than 0.5, the multidimensional fuzzy mapping decision layer initiates variable structure intervention, which will... Adaptive increase (e.g., through step control) (Instantly boosted to 3 or 5). At this point, the membership probability distribution function rapidly and steeply concentrates on the highest-scoring path (the higher the score, the higher the score, and the lower the score), forcing the vast majority of core power control flows to absolutely concentrate on the optimal deterministic queue, thereby squeezing non-core overflow traffic to suboptimal links and ensuring the deterministic bottom line of critical business.

[0157] Step S2052: Perform traffic scheduling according to the elastic allocation ratio of traffic for each path.

[0158] In this embodiment of the invention, the gateway receives the path traffic elasticity vector output in step S2051 and performs traffic scheduling.

[0159] Specifically, the gateway uses the obtained traffic elasticity ratio vector to dynamically split the aggregated power service flow into multiple sub-flows. For each newly arriving micro-flow, the gateway maps it to the corresponding queue and path based on a hash scheduling algorithm. At the same time, the gateway configures a dynamic token bucket mechanism for each queue, ensuring that the traffic throughput rate of each path strictly corresponds to the ratio calculated in step S2051, thus ensuring macro-load balancing while strictly guaranteeing the ordered transmission of data packets within the same service flow.

[0160] Step S2053: Monitor the service transmission performance deviation after traffic scheduling in real time, and correct the parameters of the heterogeneous semantic mapping matrix based on the service transmission performance deviation.

[0161] In this embodiment of the invention, adaptive correction is implemented based on the actual effect of traffic scheduling: To address mapping deviations caused by fluctuations in cross-domain network parameters, the gateway monitors the actual performance of service flows at the heterogeneous network egress point in real time. The gateway compares the actual service performance with the original target demand defined in step S201 and calculates the deviation value. When a long-term service transmission performance deviation (the difference between the average cumulative delay over the sliding window time and the expected demand) is detected to exceed the safety tolerance range, the gateway utilizes a gradient feedback mechanism to perform lightweight adaptive incremental corrections to the parameters of the heterogeneous semantic mapping matrix. This ensures that the system can dynamically track network topology evolution and maintain long-term high accuracy in cross-domain mapping.

[0162] The proposed cross-domain QoS mapping method, by constructing a general multi-dimensional demand feature space and a heterogeneous semantic mapping matrix, transforms discrete nonlinear network labels in different deterministic networks into unified, continuous physical boundary feature vectors. This significantly reduces the risk of performance jitter caused by priority resolution differences during cross-domain forwarding. The invention also uses real-time sensing of multi-dimensional operational indicators of the gateway's local queues and calculates effective congestion pressure using fuzzy logic and a double-threshold penalty function. This pressure value serves as a reverse constraint, dynamically adjusting the comprehensive adaptation score of paths through an exponential decay factor. This effectively limits excessive crowding of high-priority queues during sudden network congestion, ensuring core services while reducing performance degradation of critical boundary services. Furthermore, the method dynamically calculates path delay mutation states using time-sliding window differential gradients and variance, and generates traffic ratios using a variable-structure fuzzy membership function, achieving smooth load balancing of multi-path traffic during stable network periods. Finally, when link failures or sudden congestion occur, the method adaptively increases the adjustment factor, causing core control flow to rapidly concentrate on the optimal deterministic path, improving the network's ability to avoid failures and mutations.

[0163] This embodiment also provides a cross-domain quality of service mapping device, which is used to implement the above embodiments and preferred embodiments; details already described will not be repeated. As used below, the term "module" can be a combination of software and / or hardware that implements a predetermined function. Although the device described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0164] This embodiment provides a cross-domain quality of service mapping device, such as... Figure 5 As shown, it includes: The mapping module 301 is used to extract the heterogeneous network service quality label from the message corresponding to the original power data stream and map it to generate a business standard requirement vector when the original power data stream enters the cross-domain gateway. The acquisition module 302 is used to acquire the local queue operation indicators of each candidate sending queue inside the cross-domain gateway, and to calculate the nonlinear congestion penalty value using the local queue operation indicators. Evaluation module 303 is used to evaluate the path status of multiple candidate transmission paths of the cross-domain gateway and construct a global path standard status matrix based on the evaluation results. Calculation module 304 is used to perform joint adaptation calculation using business standard demand vector, nonlinear congestion penalty value and global path standard state matrix to generate path comprehensive adaptation score; The execution module 305 is used to determine the elastic allocation ratio of path traffic for each transmission path based on the path comprehensive adaptation score, and to perform traffic scheduling according to the elastic allocation ratio of path traffic.

[0165] In some alternative implementations, the mapping module 301 includes: The extraction unit is used to extract the heterogeneous network service quality label from the message corresponding to the original power data stream when the original power data stream enters the cross-domain gateway. The model building unit is used to construct a multi-dimensional general service quality requirement model by adopting the end-to-end latency requirements, latency jitter requirements, maximum allowable packet loss rate, and service guarantee bandwidth requirements of power services. The matrix construction unit is used to construct a heterogeneous semantic mapping matrix based on a multi-dimensional business service quality general requirement model. The mapping unit is used to input heterogeneous network service quality labels into the heterogeneous semantic mapping matrix for mapping, and generate an initial business requirement vector. The transformation unit is used to normalize the initial business requirement vector to generate a standard business requirement vector.

[0166] In some alternative implementations, the acquisition module 302 includes: The acquisition unit is used to collect local queue operation indicators of each candidate sending queue within the cross-domain gateway in parallel with a microsecond sampling period; among which, the local queue operation indicators include instantaneous load ratio, queue depth change rate and scheduling time slot utilization rate; The adjustment unit is used to adjust the preset Gaussian membership function using the business standard demand vector, and to perform fuzzification processing on the instantaneous load ratio, queuing depth change rate and scheduling slot utilization rate based on the adjusted Gaussian membership function to generate the corresponding fuzzy membership. The deduction unit is used to substitute each fuzzy membership degree into the preset multi-dimensional fuzzy condition control rules to perform fuzzy deduction and output the fuzzy set of queue congestion status. The processing unit is used to defuzzify the fuzzy set of queue congestion status using the centroid method to generate the effective congestion pressure value of the corresponding queue. The input unit is used to input the effective congestion pressure value into a preset dual-threshold nonlinear hardened congestion penalty function to generate the nonlinear congestion penalty value corresponding to each candidate sending queue.

[0167] In some alternative implementations, the evaluation module 303 includes: The data collection unit is used to collect the raw transmission metrics of multiple candidate transmission paths of the cross-domain gateway; among which, the raw transmission metrics include path latency metrics, data packet loss rate metrics, and available guaranteed bandwidth metrics. The preprocessing unit is used to preprocess the path delay index, data packet loss rate index and available guaranteed bandwidth index of each candidate transmission path using a preset normalization function to generate standardized path status data. The evaluation unit is used to evaluate the path status of the corresponding candidate transmission path based on path delay indicators, data packet loss rate indicators, and available guaranteed bandwidth indicators. The combination unit is used to combine the path state evaluation results and construct a global path standard state matrix using the standardized path state data corresponding to each candidate transmission path.

[0168] In some alternative implementations, the computing module 304 includes: The information extraction unit is used to extract the basic transmission dimension requirements information corresponding to the end-to-end latency requirements, latency jitter requirements, maximum allowable packet loss rate, and service guarantee bandwidth requirements in the business standard requirement vector. The solution unit is used to input the global path standard state matrix and the basic requirement information of each transmission dimension into the preset business scenario decoupling adaptive adjustment model, and output the initial dimension weight coefficients corresponding to the basic requirement information of each transmission dimension. The correction unit is used to dynamically correct the weight coefficients of each initial dimension based on the nonlinear congestion penalty value and perform normalization processing to generate adaptive business demand weights. The calculation unit is used to calculate the macroscopic theoretical matching degree of a single transmission path by using adaptive service demand weights and standardized path state data of each candidate transmission path in the global path standard state matrix. The attenuation correction unit is used to convert the nonlinear congestion penalty value into a link pressure exponential attenuation factor, and use the link pressure exponential attenuation factor to attenuate and correct the macroscopic theoretical matching degree of each single transmission path, thereby generating a path comprehensive adaptation score for each candidate transmission path.

[0169] In some alternative implementations, execution module 305 includes: The conversion unit is used to convert the path comprehensive adaptation score of each candidate transmission path into the corresponding path traffic elastic allocation ratio based on the preset variable structure fuzzy membership function. The execution unit is used to perform traffic scheduling according to the elastic allocation ratio of traffic for each path. The parameter correction unit is used to monitor the service transmission performance deviation after traffic scheduling in real time, and correct the parameters of the heterogeneous semantic mapping matrix based on the service transmission performance deviation.

[0170] The cross-domain quality of service mapping apparatus provided in this embodiment of the invention can execute the cross-domain quality of service mapping method provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects for executing the method. Further functional descriptions of the various modules and units described above are the same as in the corresponding embodiments described above, and will not be repeated here.

[0171] Although embodiments of the invention have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of the invention, and such modifications and variations all fall within the scope defined by the appended claims.

Claims

1. A cross-domain quality of service mapping method, characterized in that, The method includes: When the original power data stream enters the cross-domain gateway, the heterogeneous network service quality tag in the message corresponding to the original power data stream is extracted and mapped to generate a business standard requirement vector. Collect local queue operation metrics of each candidate sending queue within the cross-domain gateway, and use the local queue operation metrics to calculate a nonlinear congestion penalty value; Evaluate the path status of multiple candidate transmission paths of the cross-domain gateway, and construct a global path standard state matrix based on the evaluation results; The business standard demand vector, the nonlinear congestion penalty value, and the global path standard state matrix are used to perform joint adaptation calculation to generate a comprehensive path adaptation score. Based on the comprehensive adaptation score of the path, the elastic allocation ratio of path traffic for each transmission path is determined, and traffic scheduling is performed according to the elastic allocation ratio of path traffic.

2. The method according to claim 1, characterized in that, When the original power data stream enters the cross-domain gateway, the heterogeneous network service quality label in the corresponding packet of the original power data stream is extracted and mapped to generate a service standard requirement vector, including: When the raw power data stream enters the cross-domain gateway, the heterogeneous network service quality tag in the message corresponding to the raw power data stream is extracted. A multi-dimensional general service quality requirement model is constructed by adopting the end-to-end latency requirements, latency jitter requirements, maximum allowable packet loss rate, and service guarantee bandwidth requirements of the power business. Based on the aforementioned multidimensional business service quality general requirement model, a heterogeneous semantic mapping matrix is ​​constructed. The heterogeneous network service quality labels are input into the heterogeneous semantic mapping matrix for mapping, generating an initial service requirement vector. The initial business requirement vector is normalized to generate a standard business requirement vector.

3. The method according to claim 1, characterized in that, The step of collecting local queue operation metrics of each candidate sending queue within the cross-domain gateway and calculating nonlinear congestion penalty values ​​using these local queue operation metrics includes: The local queue operation indicators of each candidate sending queue within the cross-domain gateway are collected in parallel with a microsecond-level sampling period; wherein, the local queue operation indicators include instantaneous load ratio, queue depth change rate and scheduling time slot utilization rate; The business standard demand vector is used to adjust the preset Gaussian membership function, and the instantaneous load ratio, the queuing depth change rate and the scheduling time slot utilization rate are fuzzified based on the adjusted Gaussian membership function to generate corresponding fuzzy membership degrees. Substitute each of the fuzzy membership degrees into the preset multidimensional fuzzy condition control rules to perform fuzzy deduction and output a fuzzy set of queue congestion status. The fuzzy set of queue congestion status is defuzzified using the centroid method to generate the effective congestion pressure value of the corresponding queue. The effective congestion pressure value is input into a preset dual-threshold nonlinear hardening congestion penalty function to generate a nonlinear congestion penalty value corresponding to each candidate sending queue.

4. The method according to claim 1, characterized in that, The evaluation assesses the path status of multiple candidate transmission paths of the cross-domain gateway, and constructs a global path standard state matrix based on the evaluation results, including: Collect raw transmission metrics of multiple candidate transmission paths of the cross-domain gateway; wherein, the raw transmission metrics include path latency metrics, data packet loss rate metrics, and available guaranteed bandwidth metrics; The path delay index, data packet loss rate index and available guaranteed bandwidth index of each candidate transmission path are preprocessed using a preset normalization function to generate standardized path status data. The path status of the corresponding candidate transmission path is evaluated based on the path latency index, the data packet loss rate index, and the available guaranteed bandwidth index. Based on the path status evaluation results, a global path standard status matrix is ​​constructed using the standardized path status data corresponding to each candidate transmission path.

5. The method according to claim 1, characterized in that, The process of jointly adapting the path using the business standard demand vector, the nonlinear congestion penalty value, and the global path standard state matrix to generate a comprehensive path adaptation score includes: Extract the basic transmission dimension requirements information corresponding to the end-to-end latency requirements, latency jitter requirements, maximum allowable packet loss rate, and service guarantee bandwidth requirements from the business standard requirement vector. The global path standard state matrix and the basic requirement information of each transmission dimension are input into a preset business scenario decoupling adaptive adjustment model, and the initial dimension weight coefficients corresponding to the basic requirement information of each transmission dimension are output. Based on the nonlinear congestion penalty value, the initial dimension weight coefficients are dynamically corrected and normalized to generate adaptive business demand weights. The macroscopic theoretical matching degree of a single transmission path is calculated by using the adaptive service demand weights and the standardized path state data of each candidate transmission path in the global path standard state matrix. The nonlinear congestion penalty value is converted into a link pressure exponential decay factor, and the link pressure exponential decay factor is used to attenuate and correct the macroscopic theoretical matching degree of each single transmission path, thereby generating a path comprehensive adaptation score for each candidate transmission path.

6. The method according to claim 2, characterized in that, The step of determining the elastic allocation ratio of path traffic for each transmission path based on the comprehensive path adaptation score, and performing traffic scheduling according to the elastic allocation ratio of path traffic, includes: Based on the preset variable structure fuzzy membership function, the path comprehensive adaptation score of each candidate transmission path is converted into the corresponding path traffic elastic allocation ratio. Traffic scheduling is performed according to the elastic traffic allocation ratio for each path. Real-time monitoring of service transmission performance deviations after traffic scheduling, and correction of the parameters of the heterogeneous semantic mapping matrix based on the service transmission performance deviations.

7. A cross-domain quality of service mapping device, characterized in that, The device includes: The mapping module is used to extract the heterogeneous network service quality tags from the packets corresponding to the original power data stream when the original power data stream enters the cross-domain gateway, and then map them to generate a business standard requirement vector. The acquisition module is used to acquire local queue operation indicators of each candidate sending queue within the cross-domain gateway, and to calculate nonlinear congestion penalty values ​​using the local queue operation indicators. The evaluation module is used to evaluate the path status of multiple candidate transmission paths of the cross-domain gateway and construct a global path standard state matrix based on the evaluation results. The calculation module is used to perform joint adaptation calculation using the business standard demand vector, the nonlinear congestion penalty value, and the global path standard state matrix to generate a comprehensive path adaptation score. The execution module is used to determine the path traffic elastic allocation ratio corresponding to each transmission path based on the path comprehensive adaptation score, and to perform traffic scheduling according to the path traffic elastic allocation ratio.

8. The apparatus according to claim 7, characterized in that, The mapping module includes: The extraction unit is used to extract the heterogeneous network service quality tag from the message corresponding to the original power data stream when the original power data stream enters the cross-domain gateway. The model building unit is used to construct a multi-dimensional general service quality requirement model by adopting the end-to-end latency requirements, latency jitter requirements, maximum allowable packet loss rate, and service guarantee bandwidth requirements of power services. The matrix construction unit is used to construct a heterogeneous semantic mapping matrix based on the multidimensional business service quality general requirement model. The mapping unit is used to input the heterogeneous network service quality label into the heterogeneous semantic mapping matrix for mapping, and generate an initial service requirement vector. The transformation unit is used to normalize the initial business requirement vector to generate a standard business requirement vector.

9. The apparatus according to claim 7, characterized in that, The acquisition module includes: The acquisition unit is used to collect local queue operation indicators of each candidate sending queue inside the cross-domain gateway in parallel with a microsecond-level sampling period; wherein, the local queue operation indicators include instantaneous load ratio, queue depth change rate and scheduling time slot utilization rate; The adjustment unit is used to adjust the preset Gaussian membership function using the business standard demand vector, and to perform fuzzification processing on the instantaneous load ratio, the queuing depth change rate and the scheduling time slot utilization rate based on the adjusted Gaussian membership function, so as to generate the corresponding fuzzy membership degree. The deduction unit is used to substitute each of the fuzzy membership degrees into the preset multi-dimensional fuzzy condition control rules to perform fuzzy deduction and output the fuzzy set of queue congestion status. The processing unit is used to defuzzify the fuzzy set of queue congestion states using the centroid method to generate the effective congestion pressure value of the corresponding queue. The input unit is used to input the effective congestion pressure value into a preset dual-threshold nonlinear hardened congestion penalty function to generate a nonlinear congestion penalty value corresponding to each candidate sending queue.

10. The apparatus according to claim 7, characterized in that, The evaluation module includes: The data acquisition unit is used to collect the original transmission metrics of multiple candidate transmission paths of the cross-domain gateway; wherein, the original transmission metrics include path latency metrics, data packet loss rate metrics, and available guaranteed bandwidth metrics. The preprocessing unit is used to preprocess the path delay index, data packet loss rate index and available guaranteed bandwidth index of each candidate transmission path using a preset normalization function to generate standardized path status data. The evaluation unit is used to evaluate the path status of the corresponding candidate transmission path based on the path delay index, the data packet loss rate index and the available guaranteed bandwidth index. The combining unit is used to combine the path state evaluation results and construct a global path standard state matrix using the standardized path state data corresponding to each candidate transmission path.

Images