An intelligent substation path controllable network system and method based on IPv6 and white box exchange

Abstract

The application relates to the technical field of smart grid communication, and discloses a smart substation path controllable network system and method based on IPv6 and white box switching. A centralized controller is used to gather link state information of each white box switch to construct a global topology view. Based on the view and a service demand vector, network service flow is optimized and evaluated to generate a path planning table. Then, the path planning table is translated and coded by a rule conversion module, and is sent to a programmable data plane of the switch through a southbound interface and is solidified into a hardware forwarding table. According to the path planning table, an access side switch matches a received original service packet, generates an IPv6 encapsulated packet with a path strategy expansion header, and finally performs destination address overwrite and hop-by-hop controllable forwarding along the path indicated by the expansion header to obtain a target service packet. In this way, the closed architecture of traditional underlying equipment is broken, and accurate control and high-reliability deterministic flexible transmission of the end-to-end path of massive service in a substation are realized.

Description

Technical Field

[0001] This application relates to the field of smart grid communication technology, and more specifically, to a path-controllable network system and method for smart substations based on IPv6 and white-box switching. Background Technology

[0002] Smart substations are a crucial foundation and support for smart grids and a significant direction for the development and construction of the power Internet of Things (IoT). They not only undertake various business types such as protection and control, measurement and data acquisition, status monitoring, and auxiliary operation and maintenance, but also need to meet key performance indicators such as real-time performance, reliability, and security. However, traditional substations generally adopt a black-box, closed, dedicated structure with deep hardware and software integration, making it difficult to achieve flexible configuration and rapid iteration of network functions. To address the white-box problem, some research has broken through the traditional integrated hardware and software design of equipment, adopting a white-box open interface design, establishing unified standardized interfaces, and using a dynamic protocol library management method in program software design to achieve flexible system management. Other research has proposed a general SDN system solution based on Open Daylight and white-box switches to support network virtualization of heterogeneous cloud platforms, and provided corresponding system implementations. However, existing white-box technology solutions still face challenges in path controllability, programmable execution, and cross-layer collaboration, especially in scenarios with complex business types and continuously expanding scale. Therefore, it is necessary to conduct overall architecture research based on white-box switching technology.

[0003] To address IPv6 issues, some research has proposed an IPv6+-based intelligent vehicle-to-everything (V2X) computing network scheduling scheme. Through modular design of computing network access, perception, and scheduling, it achieves precise perception and efficient scheduling of computing power, networks, and applications across the "end-edge-cloud" ecosystem. Combined with SRv6 programmable routing technology, it autonomously selects the optimal transmission path, realizing IPv6+-based collaborative optimization of the computing network. Addressing the problems of high encapsulation overhead, limited networking range, and weak data plane capabilities in traditional virtual Ethernet networking technologies, some research has proposed EVN6, a multi-site Ethernet virtual private network technology based on the IPv6 data plane. This technology directly encapsulates the Ethernet data frames to be transmitted within IPv6 data packets, improving encapsulation efficiency and maintainability. With its large address space, automatic configuration, and security advantages, IPv6 technology has gradually become a crucial infrastructure for smart grid communication, capable of meeting the needs of large-scale equipment deployment and multi-service identifier management. However, traditional IPv4 addresses are limited, complex to configure, and difficult to support unified management of massive numbers of devices and dynamic service requirements. At the same time, as the scale of substation equipment increases and the network topology expands, traditional routing algorithms may not be able to perform path calculation and selection efficiently, which can easily lead to problems such as increased network latency, increased packet loss rate, and unbalanced link load. Therefore, it is of great significance to study a path-controllable network mechanism based on the deep integration of IPv6 and white-box switching. Summary of the Invention

[0004] To address the technical shortcomings of existing technologies, such as the rigid hardware and software caused by the black-box closed architecture of traditional substations, the lack of cross-layer collaboration between the advantages of the IPv6 protocol and white-box hardware, and the high risk of network congestion and packet loss in large-scale topologies, this application proposes a path-controllable network system and method for smart substations based on IPv6 and white-box switching. The aim is to break down the closed barriers at the device level and construct a full-stack management and control mechanism that accurately distributes high-level business requirements and integrates them into the physical hardware forwarding logic. This will meet the flexible scheduling needs of environments with massive ubiquitous device access and completely solve key pain points such as routing performance degradation, uncontrollable latency, and link load imbalance faced during substation expansion.

[0005] According to one aspect of this application, a path controllability method for smart substations based on IPv6 and white-box switching is provided, comprising: Step 1, periodically sending link detection commands to each white-box switch node in the substation through a centralized controller, and aggregating and associating the link status information reported by each node to obtain a global topology view; Step 2, generating, evaluating, and selecting the optimal path for each service flow on the network topology based on the global topology view and service demand vector to obtain a path planning table; Step 3, translating and encoding forwarding rules in the path planning table through a rule conversion module, and sending the obtained forwarding rules to the programmable data plane of the corresponding white-box switch node via the southbound interface for writing and solidification to obtain a hardware forwarding table; Step 4, performing packet feature matching and forwarding action query on the original service packets received by the access-side white-box switch based on the hardware forwarding table to obtain encapsulated packets; Step 5, sequentially overwriting the destination address and performing hop-by-hop controllable forwarding of the encapsulated packets along the hop-by-hop forwarding path indicated by the path policy extension header in the encapsulated packets to obtain the target service packets.

[0006] According to another aspect of this application, a smart substation path controllable network system based on IPv6 and white-box switching is provided, comprising: a global topology awareness module, used to periodically send link detection commands to each white-box switch node in the substation through a centralized controller, and to aggregate and correlate the link status information reported by each node to obtain a global topology view; an intelligent path planning module, used to generate, evaluate and select the optimal path for each service flow on the network topology based on the global topology view and service demand vector to obtain a path planning table; and a rule compilation and distribution module, used to compile and distribute rules. The conversion module translates and encodes the forwarding rules in the path planning table, and sends the obtained forwarding rules to the programmable data plane of the corresponding white-box switch node via the southbound interface for writing and solidification to obtain the hardware forwarding table; the packet feature matching module is used to perform packet feature matching and forwarding action query on the original service packets received by the access-side white-box switch based on the hardware forwarding table to obtain the encapsulated packet; the hop-by-hop controllable forwarding module is used to perform destination address overwriting and hop-by-hop controllable forwarding on the encapsulated packet in sequence along the hop-by-hop forwarding path indicated by the path policy extension header in the encapsulated packet to obtain the target service packet.

[0007] Compared with existing technologies, this application provides a smart substation path-controllable network system and method based on IPv6 and white-box switching. This system breaks free from the deep hardware and software integration of traditional equipment, endowing the substation's communication infrastructure network with programmable expansion capabilities that allow dynamic reconfiguration as services iterate. By explicitly mapping service requirements to IPv6 path extension headers and executing this mapping with white-box lookup, the network achieves extremely high end-to-end path control accuracy and service flow isolation guarantees. Especially when facing substation expansion and sudden surges in traffic, the unified scheduling of optimal path allocation effectively prevents cascading congestion and bottleneck resonance caused by blind contention in local links, and prevents backpressure caused by core buffer queue overflow. The entire network's link resources are utilized in an extremely balanced manner, significantly reducing end-to-end transmission latency and jitter for latency-sensitive critical services (such as GOOSE / SV relay protection data), and controlling packet loss rate to an extremely low level. This provides the highest level of determinism and high reliability for service transmission in new smart substations. Attached Figure Description

[0008] The above and other objects, features, and advantages of this application will become more apparent from the more detailed description of the embodiments of this application in conjunction with the accompanying drawings. The drawings are provided to further illustrate the embodiments of this application and form part of the specification. They are used together with the embodiments of this application to explain the application and do not constitute a limitation thereof. In the drawings, the same reference numerals generally represent the same components or steps.

[0009] Figure 1This is a three-layer network architecture according to an embodiment of this application.

[0010] Figure 2 This is a schematic flowchart illustrating a smart substation path controllability method based on IPv6 and white-box switching according to an embodiment of this application.

[0011] Figure 3 This is a comparison chart showing the time delay as a function of the number of nodes according to an embodiment of this application.

[0012] Figure 4 This is a comparison chart showing the packet loss rate as a function of the number of nodes according to an embodiment of this application.

[0013] Figure 5 This is a comparison chart showing the change in link utilization with the number of nodes according to an embodiment of this application.

[0014] Figure 6 This is a comparison chart showing the change in latency as a function of the number of candidate paths according to an embodiment of this application.

[0015] Figure 7 This is a comparison chart showing the packet loss rate as a function of the number of candidate paths according to an embodiment of this application.

[0016] Figure 8 This is a schematic block diagram of a smart substation path controllable network system based on IPv6 and white-box switching, according to an embodiment of this application. Detailed Implementation

[0017] Hereinafter, exemplary embodiments according to this application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this application, and not all embodiments of this application. It should be understood that this application is not limited to the exemplary embodiments described herein.

[0018] Before detailing the specific steps of the path controllability method proposed in this application, we will first combine... Figure 1 The overall network architecture on which the method of this application is based is described.

[0019] like Figure 1 As shown, Figure 1 This application presents a three-layer network architecture according to its embodiments. To meet the core requirements of decoupling and flexible management of smart substation network systems, this application constructs a novel physical architecture in the smart substation communication system that features layered decoupling, centralized control, and path planning. This network architecture is strictly composed of three layers from top to bottom: the service application layer, the network control layer, and the data forwarding layer.

[0020] Specifically, the business application layer directly faces various power business systems within the substation (including relay protection, measurement and control, monitoring systems, and other business-end equipment), serving as the input end for network architecture requirements. This layer does not participate in specific data packet forwarding actions but is responsible for submitting business requests (such as submitting business requirements and performance constraints) to the network control layer, as well as receiving network operating status and service feedback.

[0021] The downward-facing network control layer consists of a centralized controller, which serves as the intelligent hub of the new architecture, responsible for global topology maintenance, path calculation, and policy distribution. Combined with... Figure 2 As can be seen, this control layer deploys four core logical units: a network topology awareness module, a network status monitoring module, a path planning and scheduling module, and a policy management and rule distribution module. It serves as a bridge and makes decisions, receiving instructions from the application layer and uniformly distributing forwarding rules downwards.

[0022] The lowest data forwarding layer consists of clusters of white-box switching devices deployed in various bay layers and station control layers of the substation. This layer is responsible for receiving various hardware forwarding policies issued by the network control layer, and relying on the programmable data plane inside the white-box switches to perform multi-path forwarding table matching and load balancing, forming a highly reliable, low-latency data transmission corridor.

[0023] The business request-driven processing logic of this architecture is as follows: First, the business application layer extracts the business requirements and submits them downwards. Then, the network control layer performs path planning and policy calculation based on the global perception results and physically distributes the forwarding rules to the underlying white-box switching device cluster. Finally, the data forwarding layer executes deterministic and controllable forwarding according to the rules.

[0024] Figure 2 This is a schematic flowchart illustrating a path controllability method for smart substations based on IPv6 and white-box switching, according to an embodiment of this application. Figure 2 As shown, based on the above three-layer network architecture, this application provides a method for path controllability in smart substations based on IPv6 and white-box switching, including: Step 1: The centralized controller periodically sends link probe commands to each white-box switch node within the substation. The link status information reported by each node is aggregated and topologically correlated to obtain a global topology view. It can be understood that the network control layer, composed of the centralized controller, is the intelligent hub of the new architecture, responsible for global topology maintenance, path calculation, and policy distribution, undertaking the key functions of global perception, unified decision-making, and policy distribution. To achieve this decoupling and intelligent allocation of underlying devices, the network control layer mainly relies on two core modules to obtain real-world operating parameters: one is a network topology perception module that must acquire the topology structure and link status of white-box switching devices in real time; the other is a network status monitoring module that must monitor key indicators such as link latency, utilization, and packet loss rate. Based on this, the control layer constructs a global network view that accurately reflects physical connections through a combination of periodic and event-triggered methods. This solves problems such as disordered state and congestion oversight in traditional substation network systems, providing an indispensable high-precision data source for the next step of candidate path generation and service scheduling from a global perspective.

[0025] In one embodiment, step one includes: issuing and triggering link detection commands to each white-box switch node in the substation by responding to a timed trigger signal through the centralized controller; each white-box switch node encapsulating the adjacency relationship with the device interface status data collected at its own end and reporting it to the centralized controller to obtain link status information; parsing the link status information and calculating the weights to obtain a set of nodes and edges and a set of link weights; mapping and binding the set of nodes and edges and the set of link weights and constructing a graph structure, wherein the switch nodes are mapped as vertices and the physical links are mapped as weighted edges to obtain a global topology view.

[0026] Specifically, firstly, the centralized controller responds to periodically triggered signals to send and trigger link probe commands to each white-box switch node within the substation. Upon receiving the periodically generated timed trigger signals, the centralized controller sends link-layer probe commands to all white-box switches within the substation via its southbound interface. Each underlying white-box switch responds to the probe commands, sending physical probe signals to adjacent device ports and combining this with device interface status data extracted from its own hardware, including available interface bandwidth, real-time load, and basic interface transmit / receive latency. Based on this, each white-box switch node encapsulates the adjacency relationships using the device interface status data collected locally and reports them to the centralized controller to obtain link status information.

[0027] Next, the link status information is parsed and weights are calculated to obtain the node and edge set and the link weight set. The central controller's status monitoring module unpacks and extracts features from the aggregated link status information across the entire network, parsing out the white-box switch nodes currently alive in the substation network and the physical links interconnected with them, generating the node and edge set. For each valid physical link parsed out, the central controller extracts the available bandwidth, link latency, and load level parameters of that link, and generates a corresponding attribute vector for each link through feature combination. After aggregation, a link weight set covering the entire network is obtained. The internal weight vector calculation formula for this process is expressed as: in, This represents a weight attribute vector belonging to a specific link in the link weight set. This represents a communication link in the network topology. This indicates the available bandwidth of the link in its current state, reflecting the remaining transmission capacity of the link; This represents the basic physical transmission delay of the link, used to measure the propagation time of photoelectric signals between two nodes; This indicates the current load level of the link, representing the percentage of bandwidth being used or the degree of congestion. (In the global infrastructure reconstruction model.)

[0028] Finally, the system maps and binds the set of nodes and edges to the set of link weights and constructs the graph structure. Specifically, switch nodes are mapped to vertices, and physical links are mapped to weighted edges to obtain a global topology view. The topology planning engine of the centralized controller maps and binds the set of nodes and edges to the corresponding set of link weights one-to-one. Using graph theory modeling methods, switch nodes are mapped to vertices of the undirected graph, and physical links are mapped to edges, with the corresponding link attribute vectors assigned as edge weights. Through this hardware-software association, a complete undirected graph model is finally constructed, persisted to the graph database, and a unified global topology view containing static connection relationships and dynamic load characteristics is generated. The overall undirected graph modeling formula for this step is expressed as: in, This represents the system-level mathematical expression of the completed global network undirected graph model, i.e., the global topology view; This represents the white-box swapping node set, which is objectively mapped from the node portion of the node and edge set. , The number of nodes; It represents the set of physical communication links, which is mapped from the boundary and medium portions of the node and edge set; It corresponds to the set of link weights for the entire network, which includes the weight attribute values ​​of all valid interconnected links in the network.

[0029] It is worth noting that in the specific scenario of smart substations, where safety and continuity requirements are extremely high, in addition to the periodic detection mechanism relying on commands issued by the centralized controller, the network control layer of this application also supports dynamic state awareness based on a combination of periodicity and event triggering. Specifically, when unexpected faults occur in the underlying substation network, such as sudden cascading congestion, abnormal fiber optic link failure, or hardware failure of white-box switching equipment, the underlying white-box switching equipment cluster uses the system's built-in event triggering mechanism to proactively and immediately report abnormal state indicators to the centralized controller, bypassing the detection cycle. After capturing the abnormal event, the network control layer immediately re-prunes the topology based on the latest network state information and triggers the subsequent path planning module to dynamically recalculate and allocate safe paths for the affected power services. By dynamically adjusting and issuing new white-box switching rules, the system can guide traffic to bypass congested or faulty nodes within millisecond delays, thereby achieving intelligent network self-healing of the communication network and ensuring the ultimate survivability of the overall network architecture in the face of physical damage.

[0030] Step 2: Based on the global topology view and service demand vector, candidate paths are generated, evaluated, and optimal paths are selected for each service flow on the network topology to obtain a path planning table. It is understandable that when the scale of substation equipment increases, the network topology expands, and service types become more complex and intertwined, traditional routing algorithms such as the Shortest Path First (SPF) algorithm or the Equal Cost Multi-Path Load Sharing (ECMP) mechanism are prone to increasing network latency, packet loss rate, and link load imbalance. More importantly, traditional routing planning implicitly assumes that the performance of each link is independent and does not interfere with each other when calculating path costs, completely unaware of the congestion propagation coupling relationship that objectively exists between adjacent links based on common switching nodes. In a real network environment, this blind spot can cause buffer backpressure propagation, bottleneck resonance amplification, and path cost scoring distortion, which may misschedule high-priority protection services to paths with potential cascading congestion risks. Therefore, by introducing a refined adaptive cost evaluation mechanism that includes scale awareness and link congestion coupling factors, service modeling and multi-path cost optimization are used to achieve service scheduling with extremely low latency and high reliability requirements in substations.

[0031] In one embodiment, step two includes: extracting and standardizing the latency constraints, bandwidth requirements, reliability levels, and service priorities carried by the original service messages submitted by the service application layer to obtain a service requirement vector.

[0032] Specifically, firstly, the traffic classifier at the application layer or network control layer receives raw service packets from secondary equipment in the substation, such as relay protection devices and measurement and control devices. Using deep packet inspection technology or a packet header feature matching mechanism, it reads the Ethernet type, application identifier (e.g., GOOSE or SV application identifier), VLAN priority tag, and source or destination IP address of the packet. The extracted discrete feature parameters are summarized and structured for storage, generating a raw service attribute set describing the communication characteristics of the packet. Next, the network control layer retrieves a pre-installed smart substation service classification template library, which covers standardized templates for protection, measurement, monitoring, and auxiliary services. The input raw service attribute set is compared and mapped with the service classification template library to identify the specific category of the service and its corresponding communication assurance baseline. Based on the mapping results, the qualitative requirements of the service type are converted into quantitative network constraint values. The maximum acceptable end-to-end latency, latency jitter constraint, reliability level, and minimum bandwidth requirement of the service flow are calculated and output, packaged into standardized performance indicators in a unified format. Finally, the abstract modeling module of the network control layer receives the discrete constraint values ​​from the standardized performance indicators. To achieve complete decoupling between substation power services and the underlying communication network transmission logic, the modeling module encapsulates these performance indicators into a highly abstract four-tuple mathematical model structure. The data sequence is organized according to the four-tuple structure, ultimately outputting a unified data interface encapsulation that can be directly called by subsequent routing algorithms, i.e., the service requirement vector. The specific vectorization transformation and calculation formulas are expressed as follows: in, This represents the generated unified data interface encapsulation, i.e., the business requirement vector. This indicates the maximum acceptable end-to-end delay. This represents the delay jitter constraint threshold. Indicates the first The reliability level of a service is usually mapped to the number of redundant paths or the upper limit of packet loss rate. This indicates the minimum bandwidth requirement.

[0033] For illustration, a specific embodiment is provided below. When a relay protection device deployed in the substation bay layer needs to send a GOOSE message to the station control layer, the message is first captured by the service application layer. The traffic classifier initiates a deep packet inspection mechanism, reads the type field from the Ethernet frame header of the GOOSE message, and identifies it as a GOOSE message type defined by IEC 61850-8-1. Further, the application identifier field is extracted from the message, and its value is found to be 0x0001, corresponding to a transformer protection trip signal. Simultaneously, the priority field in the VLAN tag is read, and its value is 7, indicating the highest priority. The source IPv6 address is extracted as 2001:db8:1::10, and the destination IPv6 address as 2001:db8:1::20. These characteristic parameters are summarized to generate the original service attribute set. Subsequently, the system retrieves the service classification template library and compares the original service attribute set with the protection service templates in the library. In the template library, protection-type services are defined by an application identifier range of 0x0000 to 0x00FF, a VLAN priority of 6 or 7, and a successful match. Based on the protection-type service template, the system extracts the performance requirements for this type of service: a maximum acceptable end-to-end latency of 4 milliseconds, a latency jitter constraint of 1 millisecond, a reliability level requirement of a packet loss rate of less than one ten-thousandth, and a minimum bandwidth requirement of 10 Mbps. These quantitative indicators are packaged into standardized performance indicators. Finally, the abstract modeling module encapsulates these indicators into a four-tuple structure to generate a service requirement vector. The vector is then passed to the path planning module of the network control layer as an input constraint for subsequent candidate path generation and cost evaluation, ensuring that an optimal transmission path that meets stringent latency and reliability requirements is planned for the GOOSE message, thereby realizing a standardized transformation process from specific power services to abstract network requirements.

[0034] In one embodiment, step two further includes: filtering all physical links in the global topology view for bandwidth availability based on the source and destination node identifiers and minimum bandwidth constraints extracted from the business demand vector to obtain a pruned topology map; calling a multi-path search algorithm on the pruned topology map to generate candidate paths to obtain a candidate path set; performing cost quantification evaluation on each candidate path in the candidate path set to obtain a path comprehensive cost set; performing minimum cost location and optimization on the path comprehensive cost set, locking the path with the minimum cost from the candidate path set and converting it into an ordered node forwarding sequence to obtain a path planning table.

[0035] Specifically, firstly, the path planning module of the control layer parses the service demand vector, extracting the source node identifier, destination node identifier, and minimum bandwidth requirement constraint for the service flow. Based on the minimum bandwidth requirement constraint, it iterates and compares all physical links recorded in the global topology view, directly eliminating links whose current available bandwidth is less than the minimum bandwidth requirement, forming a pruned topology that meets the bandwidth minimum. On the pruned topology, using the extracted source node as the starting point and the destination node as the ending point, it calls the K-shortest path algorithm to search for the top K available transmission paths with connectivity, and packages the node sequences of these K paths to generate a candidate path set.

[0036] Next, the control layer counts the total number of currently active network nodes from the global topology view, and, combined with a preset baseline network size constant, calculates a size-aware factor that dynamically reflects the degree of network expansion using a logarithmic function. The calculation formula is expressed as follows: in, As a baseline network size, the absolute number of nodes Convert to relative ratio To eliminate the influence of absolute values ​​and define it in logarithmic form, on the one hand, it ensures that it monotonically increases with network size, thus strengthening the constraint on path length in large-scale networks; on the other hand, its sublinear growth characteristic avoids excessive penalty for path length, allowing the load-aware term to still play a role in the cost function, and the logarithmic function The base value is a constant parameter greater than 1 (usually the natural constant e or 10). In this embodiment, there is no unique restriction on the specific base value. It can be empirically optimized according to the range of network node scale changes of the specific substation.

[0037] In the first embodiment of this application, a cost quantification evaluation is performed on each candidate path in the candidate path set to obtain a path comprehensive cost set, including: extracting the basic latency and real-time load level of each link from the global topology view; and determining the path comprehensive cost of each candidate path based on the basic latency item and the link utilization load penalty item to obtain a path comprehensive cost set.

[0038] For each candidate path in the candidate path set, the control layer further extracts the basic propagation delay and real-time load level (i.e., link utilization) of each underlying physical link contained in that path from the global topology view. The basic propagation delay is weighted using the scale-aware factor calculated above, while a load-sensitive weight coefficient is introduced to amplify the real-time load level of the link as a load penalty term for link utilization. The weighted delay and the link utilization load penalty term are summed to calculate the overall comprehensive cost of each candidate path, and all calculation results are summarized into a path comprehensive cost set. For each candidate path... Comprehensive cost The calculation formula is expressed as follows: Among them: the first item The second term represents the base latency for scale awareness. It reflects the link load pressure, and the link utilization rate reflects the degree of congestion, penalizing high-load links. This is a size-aware factor used to adjust the impact of network size on path selection. Load-sensitive weights are used to adjust the penalty for link utilization load in the cost function. They can be dynamically adjusted based on real-time network conditions. For example, =0.3~0.5, suitable for critical protection services requiring extremely low latency (<4ms), such as GOOSE messages and SV sample values. Indicates the basic propagation delay. Indicates available bandwidth. Indicates link load. Indicates communication link The link utilization rate, i.e., the real-time load level of the link.

[0039] However, the linear summation operation in the above embodiments implicitly assumes that all links are independent of each other, which in turn implies that the performance of each link on the path is independent and does not interfere with each other. But in the physical networking environment of a smart substation, the end-to-end transmission path is carried by multiple white-box switches connected in series, and any two adjacent links... and They inevitably share a common intermediate switching node, but the message buffer queue resources, switching matrix scheduling bandwidth, and flow table query processing capabilities of this node are all limited and shared hardware resources. As a result, there is an objective congestion propagation coupling relationship between adjacent links, and the cost function of the first embodiment is completely unaware of this relationship.

[0040] The harm caused by this defect in substation communication scenarios manifests in three specific aspects.

[0041] Regarding the failure to detect backpressure propagation in the buffer zone, when the link When congestion occurs in the outbound port queue, the buffer pool of the common switching node is filled more quickly, the backpressure mechanism is triggered and acts in reverse on the link. Inbound message backlog, causing The actual latency experienced by the device is not solely determined by its own traffic load, but is also influenced by downstream factors. The congestion state is deeply coupled. For GOOSE relay protection messages that need to meet a 4ms transmission time limit, the additional queue delay caused by congestion propagation can easily lead to misjudgment or even failure of protection action.

[0042] Regarding the bottleneck resonance amplification misjudgment, when and When both links are under high load, the actual performance damage caused by the joint competition for common node resources between the two links far exceeds the theoretical expectation of the arithmetic superposition when they are under high load independently, forming a bottleneck resonance amplification effect of nonlinear deterioration. This process is completely invisible under the original linear summation framework.

[0043] Regarding path cost scoring distortion, the link utilization distribution is as follows: and The two paths may achieve similar scores under the original summation cost formula, but the former carries a much higher risk of cascading congestion due to the high-load link and adjacent links sharing a common node than the latter. The original function cannot distinguish between these two load distribution patterns, and is highly likely to schedule high-priority protection services to paths with potential congestion cascading risks, resulting in implicit service quality degradation.

[0044] To address the aforementioned technical deficiencies, in the second embodiment of this application, a cost quantification evaluation is performed on each candidate path in the candidate path set to obtain a comprehensive path cost set. This includes: traversing all links contained in each candidate path in the candidate path set, extracting the basic propagation delay and real-time load level of each link from the global topology view to obtain a link attribute parameter set; based on the link attribute parameter set, quantifying the link pair congestion coupling factor of all adjacent link pairs within each candidate path in the candidate path set to obtain a path coupling cost set; and based on the link attribute parameter set and the path coupling cost set, performing a coupling-aware comprehensive cost evaluation on each candidate path in the candidate path set to obtain a comprehensive path cost set.

[0045] First, iterate through all links contained in each candidate path in the candidate path set, and extract the basic propagation delay of each link one by one from the global topology view. With real-time load level The scale-aware factor is combined with the attribute parameters of each link. Unified encapsulation generates a set of link attribute parameters.

[0046] In smart substation scenarios, substation expansion, bay additions, and monitoring equipment deployment all lead to... Dynamically added, Real-time computation ensures that the cost function continuously senses the evolution of network size; simultaneously, The real-time extraction captures the instantaneous load status differences of the interval layer communication links at different time periods (such as the peak period when the protection system is synchronously started and the low period when the equipment monitoring data is aggregated), so that the path planning decision is accurately synchronized with the actual operation status of the substation. This provides a reliable unified data input for subsequent congestion coupling factor calculation and comprehensive cost assessment, and achieves the goal of incorporating network scale awareness and link status awareness into a unified path planning decision system.

[0047] Then, considering that the link-independent summation structure of the original cost function makes it unable to perceive the congestion propagation coupling relationship generated by adjacent links through common white-box switching nodes, this deficiency is the root cause of path cost scoring distortion. Therefore, for each pair of adjacent links in the candidate path, a coupling factor that explicitly quantifies the intensity of their congestion propagation interaction must be established to fill the perception blind spot of the original cost function.

[0048] For each candidate path in the candidate path set Arrange its contained links into an ordered sequence according to topological order. Extract all adjacent link pairs in sequence. Read from the link attribute parameter set and Substitute the link into the formula for the congestion coupling factor to perform quantitative modeling: in, For adjacent link pairs The congestion coupling factor quantifies the intensity of congestion propagation interaction between two links through a common intermediate white-box switching node. and Links and Current real-time load level; Take the larger value between the load levels of two adjacent links to identify the bottleneck side at the common node.

[0049] The above formula has undergone specific consideration of the white-box switching and forwarding characteristics of substations: numerator The joint competition strength is characterized by the product of the load levels of the two links. When either link approaches no load, the coupling effect naturally approaches zero, which aligns with physical intuition; the denominator... Using the remaining capacity of the bottleneck-side link as the amplification basis, when the higher load value approaches 1, the denominator approaches 0, and the coupling factor exhibits a non-linear explosive growth. This accurately captures the physical scenario where the public node buffer resources are consumed by destructive competition and the message queuing delay deteriorates sharply under near-saturation conditions.

[0050] After calculating the coupling factor for each link, the candidate paths are... The path-level coupling aggregation cost is obtained by summing and aggregating the congestion coupling factors of all adjacent link pairs: in, Candidate paths The path-level coupling aggregation cost comprehensively reflects the overall congestion propagation risk of the path; For path Total number of links included; index Traversing from 1 to Covers all within the path A pair of adjacent links.

[0051] Taking a candidate path carrying a relay protection GOOSE signal as an example, if the path passes through switches A→B→C sequentially, with a real-time load u=0.80 for the A→B link and u=0.85 for the B→C link, then the corresponding coupling factor γ=(0.80×0.85) / (1-0.85)=0.68 / 0.15≈4.53, which is significantly higher than expected. This accurately reflects the high-risk state of switch B under double-sided high voltage, where the buffer pool is facing overflow and is very likely to introduce sudden delay spikes to GOOSE messages. On the other hand, the load of the other path A→B is 0.60 and the load of B→C is 0.30, with a corresponding coupling factor γ=(0.60×0.30) / (1-0.60)=0.18 / 0.40=0.45. Switch B has sufficient operating margin and stable forwarding is fully guaranteed. The coupling and aggregation costs of each candidate path are summarized to generate a path coupling cost set, which provides a quantitative basis for path congestion risk with scenario discrimination for subsequent comprehensive cost assessment steps. This enables accurate completion of the congestion propagation blind spot of the original cost function, and makes the cost assessment results truly reflect the differences in congestion stability of different paths under the actual flow load scenario of the substation.

[0052] Finally, the proven effective original independent link cost mechanism and the newly introduced adjacent link congestion penalty mechanism are organically integrated into a unified coupling-aware cost function. The final cost quantification of each candidate path in the candidate path set is completed, and the path comprehensive cost set is output, providing an evaluation basis with congestion coupling awareness for the optimal path selection step.

[0053] For each candidate path in the candidate path set Read the scale-aware factor from the link attribute parameter set and each link and And read the corresponding path-level coupling aggregation cost from the path coupling cost set. Substitute the values ​​into the coupled-aware synthesis cost function to complete the evaluation: in, Candidate paths The cost of coupled sensing and integrated transmission; For scale-aware factor; For link Basic propagation delay; The load-sensitive weight coefficient controls the intensity of load penalty on a per-link basis. For link Real-time load level; The coupling-sensitive weight coefficient controls the contribution of adjacent link congestion penalty terms to the overall cost; The path-level coupling aggregation cost comes from the path coupling cost set.

[0054] The formula consists of two complementary terms. The first two terms... The original cost function's link-by-link scale-aware latency weighting and load penalty logic are fully preserved, allowing this embodiment to innovatively extend the original path selection capabilities while maintaining backward compatibility; the third item Introducing path-level congestion coupling penalty, where the scale-aware factor The introduction of this feature ensures that in substation expansion scenarios (where the average number of hops per path increases and the number of adjacent link pairs increases), the congestion coupling penalty is adjusted synchronously with the network size. This avoids the problem of excessive amplification of the coupling penalty due to the accumulation of hops in large-scale networks or insufficient perception in small-scale networks, thus maintaining the stability and balance of the cost function under different network sizes.

[0055] Taking two candidate paths carrying protection and control services with similar initial costs as an example: For a path with a link utilization distribution of [0.10, 0.90, 0.10], the coupling factors of their adjacent link pairs are (0.10 × 0.90) / (1 - 0.90) = 0.90 and (0.90 × 0.10) / (1 - 0.90) = 0.90, respectively. The path coupling aggregation cost... For paths with link utilization distribution of [0.40, 0.50, 0.40], the corresponding coupling factors are (0.40×0.50) / (1-0.50)=0.40 and (0.50×0.40) / (1-0.50)=0.40, respectively. The path coupling aggregation cost is... The coupling-aware cost function is achieved through... The significant cost difference will accurately guide the path planning decision to the second path with a more even load distribution and more sufficient working margin of the common intermediate node. This will effectively avoid scheduling high-priority protection services to paths with potential cascading congestion risks on adjacent high-load links. The comprehensive cost set of the path will truly carry the accurate quantification of the congestion coupling risk of each candidate path, thereby supporting the optimal path selection step to make scheduling decisions that are more in line with the substation's service protection needs.

[0056] In summary, the second embodiment of this application, while retaining the original scale awareness and load penalty mechanism, incorporates the physical coupling relationship between links in the substation communication network into the path cost evaluation system in the mathematical form of the adjacent link congestion coupling factor. This fundamentally fills the blind spot in congestion propagation perception caused by the link independence assumption in the original function. The improved cost function can effectively identify candidate paths with similar total link utilization but drastically different load distribution patterns. It applies a nonlinear penalty proportional to the actual cascade congestion probability to high-risk paths containing adjacent high-load link pairs. In path planning decisions for high-priority, delay-sensitive services such as relay protection and measurement and control, it prioritizes transmission paths with uniform link load distribution and sufficient operational margin for common intermediate nodes. This fundamentally reduces the probability that critical messages such as GOOSE will exceed the transmission time limit due to delay spikes caused by congestion propagation. It significantly improves the path quality stability and service deterministic transmission guarantee capability of the smart substation communication network under dynamic traffic load scenarios, achieving the goal of comprehensively improving the overall service quality of the substation communication network through a refined path cost awareness strategy without increasing any network hardware investment.

[0057] Finally, in the optimal path selection and planning table generation stage, the optimization decision engine receives and iterates through all cost values ​​in the comprehensive cost set of paths. It performs a minimum value search within the finite cost set, identifies the element with the minimum cost value, and finds the corresponding candidate path from the candidate path set as the globally optimal path. The solution formula is expressed as: in, This represents the globally locked optimal path with the absolute minimum comprehensive cost after the optimization decision has been executed. The locked globally optimal path is abstracted into a strictly ordered sequence of white-box switch node IDs and egress identifiers for forwarding, and then encapsulated and converted into a standardized forwarding form format. The final output is a path planning table that can be distributed.

[0058] Furthermore, the performance of the scale-aware adaptive path planning algorithm (SR-APP) used in this embodiment is verified from the perspectives of time and space complexity of the computational system. The total number of white-box switch nodes in the entire network is set to N, the total number of interconnected communication links to E, and the number of candidate paths required for a single service to be generated by the system to be K. If the underlying layer adopts the optimized shortest path algorithm logic, the computation time complexity of a single basic shortest path can be reduced to... At the level of [unclear], in a scenario with multi-attribute constraints including scale factor and congestion coupling calculations, the overall limit execution complexity for generating and optimizing K candidate paths is only [unclear]. The rigorous complexity analysis results show that even in a large-scale networking environment with massive access to various secondary equipment in substations and rapid expansion of network physical topology, the computation process integrating multidimensional penalty algebra designed in this application still exhibits extremely high polynomial efficiency, fundamentally solving the computational bottleneck and time overload problem of traditional computation mechanisms when facing complex power grid communication.

[0059] Step 3: The rule conversion module translates and encodes the forwarding rules in the path planning table, and then sends the resulting forwarding rules to the programmable data plane of the corresponding white-box switch node via the southbound interface for writing and solidification to obtain the hardware forwarding table. It should be understood that traditional substations generally adopt a black-box, closed, dedicated structure with deep hardware and software integration, making it difficult to achieve flexible configuration and rapid iteration of network functions. After completing the global optimization of the control layer, the policy management and rule distribution module must convert the planning results into executable forwarding rules and distribute them to the data plane. Only through unified configuration, by distributing the aforementioned IPv6 identification information and forwarding table entries to the white-box switching device, can the white-box switching data plane be programmed to parse specific extension header fields and map the path policy information therein to specific forwarding actions. This enables the data forwarding layer to completely break down device blockades and ensure deterministic transmission of critical substation services by executing multi-path forwarding and load-sharing mechanisms.

[0060] In one embodiment, step three includes: translating the path policy and encoding the abstract node forwarding sequence in the path planning table using a rule conversion module to obtain a programmable forwarding instruction set; serializing and distributing the programmable forwarding instruction set via the controller's southbound interface security channel; receiving the programmable forwarding instruction set and calling the switch's abstract interface component to perform chip-level protocol translation on the programmable forwarding instruction to obtain the hardware abstraction layer call payload; and writing and embedding the hardware abstraction layer call payload into storage space using the underlying chip driver to obtain the hardware forwarding table.

[0061] Specifically, firstly, the rule conversion module translates the abstract node forwarding sequences in the path planning table into path policies and encodes instructions to obtain a programmable forwarding instruction set. In practice, the rule conversion module inside the controller reads the abstract node forwarding sequence information carried in the path planning table. For the IPv6 data plane architecture used in the substation network, the rule conversion module maps the above abstract node sequences into strictly ordered IPv6-based segment routing identifier sequences. Then, the generated identifier sequences are combined and encoded with corresponding flow classification matching fields, such as the destination IPv6 address and flow label, converting them into a "match-action" format conforming to the standard southbound interface protocol, ultimately generating a structured programmable forwarding instruction set. The instruction mapping set generation logic in this process is internally represented by a transformation formula consisting of set union operations: in, This represents the overall framework of the programmable forwarding instruction set output after translation and encoding; This represents the received overall route planning table; This represents the total number of independent business flow path mapping entries contained in the above path planning table; This represents the set traversal index value that represents the continuous iterative recursion during the sequence translation phase. This indicates the first index corresponding to this index. The rules extract and encode a set of matching fields specifically for traffic identity verification. This indicates the set of action instructions, such as port specification, that is bound to the matching field parameter; and This indicates that the corresponding objective matching conditions and system command execution actions are wrapped and recombined in a specific format to generate a single flow table instruction mapping pairing group.

[0062] Next, the programmable forwarding instruction set is serialized, transmitted, and distributed via the controller's southbound interface security channel. Upon receiving the instruction, the network operating system of each white-box switch invokes the switch abstraction interface component to perform chip-level protocol translation of the programmable forwarding instruction to obtain the hardware abstraction layer (HAL) call payload. The controller serializes the encapsulated instruction set via the established southbound control channel and sends it to the network operating system inside the target white-box switch node. After receiving and parsing the instruction, the network operating system invokes the internally deployed switch abstraction interface component to mask the hardware read / write differences between the underlying integrated circuit chips. This translates the software-level flow table instructions into read / write bitstream data that the underlying chip can directly acquire and execute—the HAL call payload.

[0063] Finally, through the underlying chip driver, the hardware abstraction layer call payload is written to and solidified in the storage space to obtain the hardware forwarding table. After obtaining this payload, the underlying chip driver, based on the hardware bus flow, directly writes the matching rule part of the payload into the high-speed space of the tri-state addressable memory of the switching application-specific integrated circuit chip, and simultaneously writes the corresponding action rule part of the payload, which is used to perform the operation, into the static random access memory space. After confirming that the physical level of the internal operation address of the chip has been completed and verified to be correct, the control strategy that was previously detached from the system software level is completely presented as the solidified chip-level circuit execution logic state, that is, a unique hardware forwarding table that directly guides the line-speed forwarding of messages in the data plane is constructed.

[0064] To illustrate the above process, a specific embodiment is provided. When the control layer in a smart substation generates a communication task plan for a certain sampling monitoring service via a specific link, the module responsible for system translation instantly rewrites the abstract link into a program segment that conforms to the OpenFlow standard protocol or P4 runtime code. The system serially packages and distributes this program code to the white-box switch operating system of a certain node through a pre-built gRPC high-security control channel. At this time, the operating system, acting as the control medium, starts the switch abstract interface middleware process inside the machine, generates chip-level control microcode, and transmits it to the main switching chip along the internal serial high-level physical bus. The driving process then forcefully intervenes and uses electrical signals to solidify the task of extracting the specific monitoring service feature identifier into the gate circuit register area with extremely high verification speed, i.e., the three-state content addressable memory block; while the instruction that, once the monitoring service identifier is matched, needs to be forced to push a specific IPv6 identifier header and forcibly arranged to be delivered outward from the No. 8 fiber downlink interface is written into and deadlocked in the corresponding static random access memory space.

[0065] Step 4: Based on the hardware forwarding table, perform packet feature matching and forwarding action query on the original service packets received by the access-side white-box switch to obtain the encapsulated packets. It should be understood that after completing the global path planning of the control layer and the solidification of the underlying hardware forwarding table, the data forwarding layer consists of white-box switching devices deployed in the substation, performing multi-path forwarding and load balancing to form a highly reliable, low-latency data forwarding network. Each white-box switching device follows a unified forwarding execution logic, first parsing the key fields of the IPv6 packet, then matching flow table or forwarding table entries, and finally completing the execution of forwarding, queue selection, and scheduling operations. For different service types, the data forwarding layer ensures that critical services still meet latency requirements under high load conditions through queue isolation and priority scheduling mechanisms. This application proposes an IPv6 and white-box switching integration mechanism, fully leveraging the unified identification and semantic expression capabilities of IPv6, as well as the programmable data plane advantages of white-box switching, and globally coordinated by the control layer to achieve a service-driven, policy-driven, distributed, and controllable path network.

[0066] In one embodiment, step four includes: parsing the protocol header and extracting features from the original service packet using the data plane parsing engine of the access-side white-box switch to obtain a packet feature vector; performing a precise matching query on the packet feature vector in the hardware forwarding table, and reading the port forwarding policy and path identifier sequence from the associated storage area of ​​the hit entry to obtain the forwarding action and extension header parameters; constructing and layering the header of the forwarding action and extension header parameters with the original service packet, and sequentially pushing the IPv6 basic header and path policy extension header outside the original service packet to obtain the encapsulated packet.

[0067] Specifically, firstly, the access-side white-box switch's data plane parsing engine performs protocol header parsing and feature extraction on the original service packets to obtain packet feature vectors. The parsing engine, through deep packet inspection logic, reads the five-tuple from the packet header, including source IP, destination IP, protocol type, source port, destination port, and service flow classification identifier. It then extracts key fields using hashing and bit-truncation algorithms, aggregating them into a fixed-length query keyword to generate the packet feature vector. Simultaneously, the received complete packet body is stored in the switch's data plane high-speed buffer, generating a cached original packet. The input macro-level lookup reference object hardware forwarding table remains unchanged and is passed through to the next flow table query stage to prevent data flow interruptions. The logical description and calculation formula for this process are expressed as follows: in, This indicates the source data payload introduced through the external communication interface, which corresponds to the original business message of the input data object. This represents the feature extraction and message cloning buffer synthesis function executed by the message parsing engine, which mainly performs protocol parsing and field truncation. This represents the key features for stream classification extracted from the message, which corresponds to the generated data object message feature vector. This represents the message body data stored in the data plane high-speed buffer, which is the original message cached by the corresponding data object.

[0068] Next, a precise matching query is performed on the packet feature vector in the hardware forwarding table. The port forwarding policy and path identifier sequence are read from the associated storage area of ​​the matched entry to obtain the forwarding action and extended header parameters. The hardware forwarding table refers to the hardware-level retrieval space resource in the underlying application-specific integrated circuit (ASIC) chip. It contains a three-state content-addressable memory (TPM) for storing matching rules and a static random access memory (SRAM) for storing execution actions. The switch's lookup engine uses the packet feature vector as a retrieval index, directly inputting it into the three-state TPM space of the underlying hardware forwarding table for high-speed matching. After a unique matching entry is found in the hardware space, the specific execution policy corresponding to the service flow is read from the associated SRAM instruction space, such as specifying the physical outgoing port and the list of IPv6-based segment routing path identifiers to be pushed in. These two are then packaged together to generate the forwarding action and extended header parameters. The original packet, which does not need to participate in the current lookup logic, is smoothly passed through to the next-level encapsulation processing module. The logical description and matching formula of this process are as follows: in, This refers to the hardware-level retrieval space resources in the underlying application-specific integrated circuit (ASIC) chip, which corresponds to the hardware forwarding table of the input data object. This represents a hardware-level fast matching function based on a three-state content-addressable memory mechanism, which uses feature vectors to find exact matches or longest prefix matches. This indicates the specific data plane processing action contained in the hit hardware entry, such as sending to a specified physical exit or performing traffic shaping. This indicates the segment routing parameter information to be carried by the IPv6 extension header, which includes a list of segment identifiers and service tags, etc. This represents the combined payload mapping body consisting of actions and extended parameters, i.e., the forwarding actions and extended header parameters of the corresponding data object.

[0069] Finally, the forwarding action and extension header parameters are combined with the original service packet to construct headers and perform layered encapsulation. An IPv6 base header and a path policy extension header are sequentially pushed onto the outside of the original service packet to obtain the encapsulated packet. The forwarding action and extension header parameters refer to the combined payload mapping body of the specific data plane processing action contained in the hit hardware entry and the segment routing parameter information carried by the IPv6 extension header, including a segment identifier list and service tags. The switch's data plane rewriting and encapsulation engine reads the path policy carried in the forwarding action and extension header parameters, especially the segment identifier list. Based on these parameters, a completely new IPv6 base header and a segment routing extension header containing complete source routing path information are constructed. The original packet in the buffer is extracted as the internal payload, and the constructed IPv6 base header and extension header are sequentially pushed onto it. Packet concatenation is performed, and finally, an encapsulated packet carrying an end-to-end path controllability flag is output. The concatenation logic formula for this process is expressed as: in, This indicates a brand-new external IPv6 base header that is constructed in real time during the encapsulation phase according to the protocol stack specifications. Operators for sequential concatenation and push operations on byte or bit streams. This indicates that the final composite data packet contains a complete IPv6 routing protocol structure nested outside the original payload, which corresponds to the final output encapsulation message of the system.

[0070] Furthermore, to achieve explicit and lossless mapping of service characteristics to path policies at the data plane, this application extends the functionality of the IPv6 address structure and flow label mechanism in the encapsulated packets. Specifically, for the IPv6 address structure, while retaining the standard format, the address field is logically divided into a site identifier field, a device type field, a service function field, and a node identifier field. The site identifier field distinguishes different substations or physical areas within a substation; the device type field identifies whether the endpoint is a primary device, secondary device, or network device; and the service function field reflects the service type of the packet, thus directly mapping the service type, service domain, and security level to the IPv6 address prefix, providing a foundation for rapid classification at the data plane. In addition, this application establishes a flow label mapping mechanism. When the control layer performs the aforementioned end-to-end path planning, it not only generates a specific sequence of forwarding nodes but also assigns a unique or semi-unique IPv6 flow label value to service flows with different specific performance requirements and binds this label value to the issued forwarding rules. Thus, when the encapsulation engine constructs a new IPv6 basic header, it can directly push this label value into the flow label field. Therefore, white-box switching devices can directly perform queue selection and differentiated forwarding based on flow labels in the subsequent data plane, successfully decoupling path selection and service characteristics at the network layer and avoiding the introduction of additional complex tunneling mechanisms.

[0071] To illustrate the above process, a specific embodiment is provided. When a relay protection device deployed in the substation bay layer needs to send a GOOSE message to the monitoring host in the station control layer, the message first arrives at the access-side white-box switch. The switch's data plane parsing engine immediately initiates a deep packet inspection mechanism, extracting the source MAC address and destination MAC address from the Ethernet frame header of the GOOSE message, the source IPv6 address and destination IPv6 address from the IPv6 base header, and the protocol type and port number from the transport layer header. These fields are then compressed into a 128-bit fixed-length lookup key, i.e., the message feature vector, using a CRC32 hash algorithm. Simultaneously, the switch copies and stores the entire byte content of the complete GOOSE message to a free slot in the on-chip high-speed buffer, generating a cached original message. Subsequently, the lookup table engine uses the aforementioned 128-bit message feature vector as an index to perform parallel comparisons in the three-state content-addressable memory space of the hardware forwarding table, matching a pre-issued matching rule within nanoseconds. The rule instructs the system to read two key pieces of information from the associated static random access memory: first, the GOOSE packet should be forwarded from physical port 5 of the switch; second, a segment routing extension header containing three segment identifiers needs to be pushed onto the packet, with these three identifiers pointing sequentially to the three intermediate switch nodes along the path. After obtaining these forwarding actions and extension header parameters, the encapsulation engine immediately constructs a new IPv6 base header, setting its version field to 6, its flow label field to the unique identifier of the service flow, and its next header field to 43 to indicate that a segment routing extension header follows. Next, the segment routing extension header is constructed, filling the segment list with the three segment identifiers in sequence and setting the remaining segment numbers to 2. Finally, the encapsulation engine retrieves the previously stored complete GOOSE packet from the buffer, uses it as the payload, and sequentially concatenates the constructed IPv6 base header and segment routing extension header before it to form a complete encapsulated packet. The encapsulated message is then pushed to the transmission queue of physical port 5, ready for hop-by-hop controlled forwarding along the explicit path planned by the controller, thus completing the critical transformation process from the original service message to the encapsulated message carrying the path policy.

[0072] Step 5: Following the hop-by-hop forwarding path indicated by the path policy extension header in the encapsulated packet, the encapsulated packet is sequentially overwritten with the destination address and forwarded in a controlled hop-by-hop manner to obtain the target service packet. It should be understood that the white-box switching data plane first parses the IPv6 basic header and extension header fields, extracts service-related information such as address fields, flow labels, and extension headers, and matches them with the local forwarding table. Then, based on the matching results, path selection is performed. Differentiated queue scheduling, traffic shaping, or rate limiting is performed on services of different priorities to ensure deterministic transmission of critical services. Link utilization, queue length, and other operational status are collected and periodically fed back to the control layer for dynamic optimization of the path policy. In this way, deep collaboration between the IPv6 extension header and programmable forwarding logic is achieved, enabling efficient execution of the path policy in the data plane. This application utilizes the IPv6 extension header to provide a flexible mechanism for carrying additional information at the network layer, and designs an extension header format for expressing path policies to carry auxiliary information related to path planning, such as service priority, path category, or forwarding constraints. The generation strategy and identifier mapping relationship of this extension header are issued by the control layer during the service establishment phase and pushed in by the access-side forwarding layer device. Subsequent forwarding devices execute corresponding forwarding strategies based on the information after parsing. The control layer sends the above IPv6 identifier information together with the forwarding table entries to the white-box switching device. The data plane of the white-box switching device can be programmed to parse specific extension header fields and map the path policy information therein to specific forwarding actions.

[0073] In one possible implementation, step five includes: parsing and reading the path policy extension header of the encapsulated packet through the data plane engine of the intermediate white-box switch to obtain the remaining segment count flag and active segment identifier; based on the remaining segment count flag and active segment identifier, overwriting the destination address and encoding the hop-by-hop forwarding packet to obtain the hop-by-hop forwarding packet; and through the decapsulation module of the tail white-box switch, when it is detected that the remaining segment value of the hop-by-hop forwarding packet is zero or the destination address matches the device identifier, stripping the extension header and restoring the payload of the hop-by-hop forwarding packet to obtain the target service packet.

[0074] Specifically, firstly, the data plane engine of the intermediate white-box switch parses and reads the path policy extension header of the encapsulated packet to obtain the remaining segment count flag and the active segment identifier. The data plane engine of the intermediate forwarding white-box switch in the substation network receives the encapsulated packet transmitted from the previous hop. The parsing engine reads the IPv6 segment routing extension header of the packet's outer layer, extracts the remaining segment count value, and determines the packet's progress along the predetermined path. Based on the read remaining segment count pointer position, it locates and reads the IPv6 address value that the next hop must strictly point to from the segment identifier list carried in the extension header, and outputs it as the active segment identifier. Simultaneously, the parsed value is output as an independent remaining segment count flag, and the packet body, stripped of its current layer logic operations and awaiting overwriting, is passed to the next stage as the encapsulated packet to be updated. The calculation and extraction logic formula for this process is expressed as: in, This refers to the message entity that is actually transmitted in the network and carries extended encapsulation information, i.e., the message encapsulated in the corresponding input data object. This refers to the packet parsing function in the data plane of a white-box switch, which is responsible for quickly identifying the IPv6 base header and segment routing extension header at the hardware level and extracting characteristic parameters. This indicates the specific IPv6 address value currently pointed to by the pointer in the segment identifier list, which corresponds to the active segment identifier of the data object. It indicates the number of segments remaining when the message arrives at this node, and is used to indicate the position of the active pointer in the path list, i.e., the remaining segment count marker of the corresponding data object. This represents the memory image of the message structure that is waiting to be updated, including the target address and segment pointer; that is, the encapsulated message containing the corresponding data object to be updated.

[0075] Next, based on the remaining segment count flag and the active segment identifier, the encapsulated packet undergoes destination address overwriting and hop-by-hop forwarding encoding to obtain a hop-by-hop forwarding packet. The switch's header rewriting engine receives the remaining segment count flag and performs a strict decrement operation to update the path progress of end-to-end communication. The updated remaining segment value is written back to the corresponding extended header field of the encapsulated packet to be updated. Simultaneously, the input active segment identifier is extracted and directly overwritten to the destination address field of the outer IPv6 base header of the packet. According to the local hardware routing table of the white-box switch, the packet with the updated header information is pushed to the corresponding physical output port for deterministic forwarding, generating a hop-by-hop forwarding packet flowing to the next-hop switch. The packet state update formula for this process is expressed as: in, This function represents the message header rewrite and update function, which performs pointer decrement and IPv6 destination address overwrite operations to guide the message hardware forwarding behavior. This indicates that the packet has completed the header update operation and is already in the outgoing port's sending queue, meaning the corresponding data object is forwarded hop-by-hop. This indicates that after the decrement operation, i.e. The new remaining segment value that will soon be sent to the downstream node.

[0076] Finally, through the decapsulation module of the tail-end white-box switch, when it detects that the remaining segment value of the hop-by-hop forwarding packet is zero or the destination address matches the local device identifier, it strips the extended header and restores the payload to obtain the target service packet. The tail-end white-box switch, i.e., the network egress node of the substation service data, receives the hop-by-hop forwarding packets transmitted from the network link and performs endpoint verification. When it detects that the remaining segment value in the packet extended header has decreased to 0, or the outer IPv6 destination address matches the local identifier of the tail-end device itself, it is determined that the packet has completed all explicit paths allocated by the control layer. The decapsulation module of the tail-end switch triggers the stripping mechanism, performing the popping operation of the IPv6 segment routing extended header and the outer IPv6 base header of the packet. The outer network tunnel encapsulation is completely removed, accurately restoring it to the original Ethernet payload sent by the communication, outputting the final target service packet, and delivering it to the destination power secondary equipment to complete the data loop. The decapsulation logic formula of this process is expressed as: in, This represents the decapsulation function, used to strip all outer encapsulation headers introduced by IPv6 and IPv6-based segment routing protocols at the tail node, retaining only the internal payload. $P_{target}$ represents the clean power service data frame finally exposed after the decapsulation and restoration operation, corresponding to the target service message of the macroscopic output object.

[0077] It should be further explained that, in the aforementioned hop-by-hop forwarding process, in addition to the parsing of the packet header and the precise overwriting of the target address, the data plane engine of each white-box switch also relies on the underlying hardware characteristics to execute deep queue scheduling and closed-loop status monitoring strategies. Specifically, after each switch matches the local port forwarding strategy based on the remaining segment count, it can perform differentiated queue scheduling, traffic shaping, or strict rate limiting at the physical egress point in real time according to the flow label and service priority carried by the packet. This is to cope with sudden micro-burst traffic and absolutely guarantee the deterministic transmission lower limit of key substation services such as protection control and continuous sampling under local high load environments. At the same time, the underlying monitoring probes of each white-box device continuously and frequently collect the current link utilization value of each port and the real-time backlog length of each port queue, etc., to obtain deep operational status. Through a combination of timed active uploading and congestion threshold-triggered alarms, the above status parameters are periodically fed back to the network control layer. This not only realizes the closed-loop visibility of the underlying execution environment, but also constitutes the most direct data source for calculating and continuously correcting the path congestion coupling factor in the aforementioned system embodiments, realizing dynamic self-healing optimization of strategies under the converged architecture.

[0078] To illustrate the above process, a specific embodiment is provided. When an encapsulated packet carrying a segment routing extension header is sent from the access-side white-box switch, it is forwarded hop-by-hop along the explicit path planned by the controller. Assume the packet's segment identifier list contains three segment identifiers, SID1, SID2, and SID3, corresponding to the IPv6 addresses of the three intermediate switch nodes along the path. When the packet arrives at the first intermediate switch, its data plane engine immediately initiates the parsing process, reading the remaining segment number from the packet's segment routing extension header. Finding its value to be 2, it indicates that two more hops are needed to reach the destination. Simultaneously, the parsing engine locates the active segment identifier, SID1, in the segment identifier list based on the remaining segment value. The header rewriting engine then overwrites the destination address field of the packet's outer IPv6 base header with the IPv6 address corresponding to SID1 and decrements the remaining segment number by 1, updating it to 1. After the update, the switch queries the local hardware forwarding table for the outgoing port corresponding to SID1 and forwards the packet from that port. After the packet reaches the second intermediate switch, the above process is repeated, the remaining segment count decreases from 1 to 0, the destination address is overwritten with the IPv6 address corresponding to SID2, and the packet continues to be forwarded. When the packet reaches the third switch, i.e., the tail white-box switch, the data plane engine of this switch reads the remaining segment number and finds that its value is 0, or finds that the outer IPv6 destination address is exactly its own local identifier SID3. At this time, the tail switch determines that the packet has completed all hop-by-hop forwarding, and the decapsulation module immediately starts, stripping the outer IPv6 base header and segment routing extension header from the packet to extract the original GOOSE packet. This GOOSE packet is then delivered to the station control layer monitoring host connected to the tail switch. The monitoring host successfully receives and parses the packet, completing the end-to-end controllable transmission from the source relay protection device to the destination monitoring host. Throughout the process, messages are forwarded strictly according to the path pre-planned by the controller. The forwarding behavior of each hop is precisely controlled by the segment identifier in the segment routing extension header, realizing path controllability and deterministic service transmission in the smart substation communication network.

[0079] To intuitively verify the effectiveness and performance advantages of the path controllability method proposed in this application and its scale-aware adaptive path planning algorithm (SR-APP), a simulation test environment was established. The basic link bandwidth was 10Mbps, the number of baseline nodes was 10, and the number of candidate paths was 3. The simulation cycle parameter was based on 200 continuous dynamic packet scheduling cycles. The performance differences of this test algorithm with the traditional Shortest Path First (SPF) algorithm and Equivalent Multipath Load Sharing (ECMP) algorithm in terms of core parameters such as latency and packet loss rate were compared.

[0080] Reference Figure 3The comparison of time delay with the number of network topology nodes shows that as the number of access nodes in the substation increases and the topology becomes more complex, the average channel delay under all mechanisms increases to varying degrees, but the differences are extremely significant. This is mainly because the SPF method is limited to selecting the rigid shortest path, leading to a surge in congestion delay; although ECMP has load-sharing capabilities, it lacks awareness in the case of a large number of nodes, still inducing queuing timeouts; while the method in this application, with the strong sensitivity of the scale-aware factor and link congestion coupling factor in its cost function, severely punishes high-utilization public link nodes that may induce cascading reactions due to traffic convergence, and prioritizes and balances the alternative path space. Therefore, it can extremely effectively suppress the phenomenon of rapid increase in congestion delay caused by high load, and has obvious optimal performance.

[0081] Related references Figure 4 Packet loss rate statistics comparison chart and Figure 5 The link utilization extreme value comparison chart shows that in a large environment where the unilateral expansion of network scale induces a surge in regional traffic, the SPF and ECMP packet loss rate data show a monotonically increasing degradation, and after the system has 15 nodes, its maximum single-machine utilization rate quickly saturates and soars to the 95% range. However, thanks to the real-time dynamic adjustment and intelligent recalculation allocation characteristics, this solution can still prevent the lowest level of service data from being lost or damaged, and accurately and steadily control the maximum link utilization rate to always be within the relatively healthy peak range of 85%, completely cutting off the possibility of a crash induced by an overloaded network environment.

[0082] Furthermore, referring to Figure 6 Distribution of latency as a function of the number of candidate paths used in the algorithm Figure 7 As can be seen from the changes in packet loss rate with the number of candidate paths, when the number of candidate paths that the routing engine can explore is manually adjusted, since SPF naturally only selects the single minimum term, its latency index appears as a long straight line in the graph, but its packet loss performance is poor. The blind allocation of cores by ECMP leads to an irreversible sharp decline in various data as the K value increases when a large number of redundant branches that are considered equivalent but are actually suboptimal or even high-risk are introduced. However, this application, by performing deep dynamic prediction of the global network, fundamentally takes into account the real-time coupling load interaction information of multiple transmission links and performs flexible scheduling. This ensures that no matter how the number of candidate paths fluctuates, the system network communication parameters are suppressed and kept within the minimum latency and no data loss red line, providing excellent robustness and service quality control guarantee for the transmission of power safety business flows.

[0083] In summary, the path controllability method for smart substations based on IPv6 and white-box switching provided in this application has been clarified. It constructs a global topology view by aggregating link status information from various white-box switches through a centralized controller. Based on this view and service demand vectors, it optimizes and evaluates network service flows, generating a path planning table. Subsequently, the table is translated and encoded by a rule conversion module, then sent to the programmable data plane of the switch via the southbound interface and solidified into a hardware forwarding table. The access-side switch performs feature matching on the received original service packets, generating IPv6 encapsulated packets with path policy extension headers. Finally, the destination address is overwritten and hop-by-hop controllable forwarding is performed sequentially along the path indicated by the extension header to obtain the target service packet. This breaks the closed architecture of traditional underlying devices, achieving precise end-to-end path control and highly reliable deterministic flexible transmission of massive services in substations.

[0084] Figure 8 This is a schematic block diagram of a smart substation path-controllable network system based on IPv6 and white-box switching, according to an embodiment of this application. Figure 8 As shown, this application also provides a smart substation path controllable network system 800 based on IPv6 and white-box switching, including: a global topology awareness module 810, used to periodically send link detection commands to each white-box switch node in the substation through a centralized controller, and to aggregate and correlate the link status information reported by each node to obtain a global topology view; an intelligent path planning module 820, used to generate, evaluate and select the optimal path for each service flow on the network topology based on the global topology view and service demand vector to obtain a path planning table; and a rule compilation and distribution module 830, used to... The rule conversion module translates and encodes the forwarding rules in the path planning table, and sends the obtained forwarding rules to the programmable data plane of the corresponding white-box switch node via the southbound interface for writing and solidification to obtain the hardware forwarding table; the packet feature matching module 840 is used to perform packet feature matching and forwarding action query on the original service packets received by the access-side white-box switch based on the hardware forwarding table to obtain the encapsulated packet; the hop-by-hop controllable forwarding module 850 is used to sequentially overwrite the destination address and perform hop-by-hop controllable forwarding of the encapsulated packet along the hop-by-hop forwarding path indicated by the path policy extension header in the encapsulated packet to obtain the target service packet.

[0085] It should be noted that the path controllable network system for smart substations based on IPv6 and white-box switching in this application embodiment is similar in principle to the aforementioned path controllable method for smart substations based on IPv6 and white-box switching. Therefore, the implementation process, implementation principle, and beneficial effects of the path controllable network system for smart substations based on IPv6 and white-box switching can be found in the description of the implementation process, implementation principle, and beneficial effects of the aforementioned method, and will not be repeated.

[0086] It should be understood that the various forms of processes shown above can be used to rearrange, add, or delete steps. For example, the steps described in this application can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this application can be achieved, and this is not limited herein.

[0087] Furthermore, 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 at least one of that feature. In the description of this application, "a plurality of" means two or more, unless otherwise explicitly specified.

[0088] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for path controllability in smart substations based on IPv6 and white-box switching, characterized in that, include: Step 1: Periodically send link detection commands to each white-box switch node in the substation through the centralized controller, and aggregate and associate the link status information reported by each node to obtain a global topology view. Step 2: Based on the global topology view and business demand vector, generate, evaluate and select the optimal path for each business flow on the network topology to obtain a path planning table. Step 3: Translate and encode the forwarding rules of the path planning table through the rule conversion module, and send the obtained forwarding rules to the programmable data plane of the corresponding white box switch node via the southbound interface for writing and solidification to obtain the hardware forwarding table. Step 4: Based on the hardware forwarding table, perform packet feature matching and forwarding action query on the original service packets received by the access-side white-box switch to obtain the encapsulated packets. Step 5: Following the hop-by-hop forwarding path indicated by the path policy extension header in the encapsulated message, the destination address of the encapsulated message is overwritten and hop-by-hop controlled forwarding is performed sequentially to obtain the target service message.

2. The method for controllable path in a smart substation based on IPv6 and white-box switching according to claim 1, characterized in that, Step two includes: extracting and standardizing the latency constraints, bandwidth requirements, reliability levels, and service priorities carried by the original business messages submitted by the business application layer to obtain a business requirement vector.

3. The method for controllable path in a smart substation based on IPv6 and white-box switching according to claim 1, characterized in that, Step one includes: The centralized controller responds to the timed trigger signal to send and trigger the link detection command to each white box switch node in the substation. Each white box switch node combines the device interface status data collected at its own end to encapsulate the adjacency relationship and report it to the centralized controller to obtain the link status information. The link state information is parsed and weights are calculated to obtain the set of nodes and edges and the set of link weights; The nodes and edges are mapped and bound to the link weights, and a graph structure is constructed. Specifically, switch nodes are mapped as vertices and physical links are mapped as weighted edges to obtain a global topology view.

4. The method for controllable path in a smart substation based on IPv6 and white-box switching according to claim 2, characterized in that, Step two includes: Based on the source and destination node identifiers and minimum bandwidth constraints extracted from the business demand vector, bandwidth availability filtering is performed on all physical links in the global topology view to obtain a pruned topology map. A multi-path search algorithm is invoked on the trimmed topology graph to generate candidate paths and obtain a set of candidate paths. The cost of each candidate path in the candidate path set is quantitatively evaluated to obtain the comprehensive path cost set. The minimum cost is located and optimized from the set of comprehensive path costs. The path with the minimum cost is locked from the set of candidate paths and transformed into an ordered node forwarding sequence to obtain the path planning table.

5. The method for path controllability in smart substations based on IPv6 and white-box switching according to claim 4, characterized in that, The cost of each candidate path in the candidate path set is quantified and evaluated to obtain a comprehensive path cost set, including: Extract the basic latency and real-time load level of each link from the global topology view; Based on the basic latency term and the link utilization load penalty term, the comprehensive path cost of each candidate path is determined to obtain the comprehensive path cost set.

6. The method for controllable path in a smart substation based on IPv6 and white-box switching according to claim 1, characterized in that, Step three includes: The rule conversion module translates the abstract node forwarding sequence in the path planning table into a path policy and encodes the instruction to obtain a programmable forwarding instruction set. The programmable forwarding instruction set is serialized, transmitted, and distributed through the secure channel of the controller's southbound interface. After receiving the instruction, the network operating system of each white-box switch calls the switch abstract interface component to perform chip-level protocol translation of the programmable forwarding instruction to obtain the hardware abstraction layer call payload. The hardware forwarding table is obtained by writing and fixing the hardware abstraction layer call payload into the storage space through the underlying chip driver.

7. The method for controllable path in a smart substation based on IPv6 and white-box switching according to claim 1, characterized in that, Step four includes: By using the data plane parsing engine of the access-side white-box switch, the original service packets are parsed for protocol headers and features are extracted to obtain packet feature vectors. The message feature vector is precisely matched and queried in the hardware forwarding table. The port forwarding policy and path identifier sequence are read from the associated storage area of ​​the hit entry to obtain the forwarding action and extended header parameters. The forwarding action, extension header parameters, and original service packets are constructed and encapsulated in layers. The IPv6 basic header and path policy extension header are sequentially pushed outside the original service packets to obtain the encapsulated packets.

8. The method for controllable path in a smart substation based on IPv6 and white-box switching according to claim 1, characterized in that, Step five includes: The intermediate white-box switch data plane engine parses and reads pointers from the path policy extension header of the encapsulated packet to obtain the remaining segment count flag and active segment identifier. Based on the remaining segment count marker and the active segment identifier, the encapsulated message is overwritten with the destination address and hop-by-hop forwarding encoding to obtain a hop-by-hop forwarding message; When the remaining segment value of the hop-by-hop forwarding packet is detected to be zero or the destination address matches the device identifier by the decapsulation module of the tail white-box switch, the extended header of the hop-by-hop forwarding packet is stripped and the payload is restored to obtain the target service packet.

9. The method for controllable path in a smart substation based on IPv6 and white-box switching according to claim 4, characterized in that, The cost of each candidate path in the candidate path set is quantified and evaluated to obtain a comprehensive path cost set, including: Traverse the links contained in each candidate path in the candidate path set, and extract the basic propagation delay and real-time load level of each link from the global topology view to obtain the link attribute parameter set; Based on the link attribute parameter set, the link pair congestion coupling factor of all adjacent link pairs in each candidate path in the candidate path set is quantified to obtain the path coupling cost set. Based on the link attribute parameter set and the path coupling cost set, a coupling-aware comprehensive cost evaluation is performed on each candidate path in the candidate path set to obtain the path comprehensive cost set.

10. A smart substation path-controllable network system based on IPv6 and white-box switching, characterized in that, include: The global topology awareness module is used to periodically send link detection commands to each white-box switch node in the substation through the central controller, and to aggregate and associate the link status information reported by each node to obtain a global topology view. The intelligent path planning module is used to generate, evaluate and select the optimal path for each service flow on the network topology based on the global topology view and service demand vector to obtain a path planning table. The rule compilation and distribution module is used to translate and encode the forwarding rules of the path planning table through the rule conversion module, and distribute the obtained forwarding rules to the programmable data plane of the corresponding white box switch node via the southbound interface for writing and solidification to obtain the hardware forwarding table. The message feature matching module is used to perform message feature matching and forwarding action query on the original service messages received by the access-side white-box switch based on the hardware forwarding table to obtain the encapsulated message. The hop-by-hop controllable forwarding module is used to sequentially overwrite the destination address and perform hop-by-hop controllable forwarding of the encapsulated message along the hop-by-hop forwarding path indicated by the path policy extension header in the encapsulated message to obtain the target service message.

Images