A data reliable delivery method and system in a weak network environment of a power distribution network based on network fusion

Abstract

The application discloses a data reliable delivery method and system in a weak network environment of a power distribution network based on network number fusion, and relates to the technical field of power distribution communication and data transmission. The method generates a network number fusion identifier for a service message, performs data slicing and sets a network location record field, constructs a candidate path set in combination with available links, constructs path weak network state characteristics according to historical confirmation feedback, historical retransmission feedback and path state information, calculates the average loss probability and location consistency score of the candidate path, and determines the bearing path, the coding window and the number of coding redundant slices to be sent according to the score; the candidate path and the sending strategy are updated when the path state is abnormal; the receiving end completes association, decoding recovery, reorganization, integrity check and in-sequence delivery. The application can improve the reliable delivery capability and in-sequence delivery capability of service messages in a weak network environment of a power distribution network.

Description

Technical Field

[0001] This invention relates to the field of power distribution communication and data transmission technology, and in particular to a reliable data delivery method and system based on network-data integration in a weak network environment of a power distribution network. Background Technology

[0002] In distribution network communication scenarios, various types of business data, such as status reporting messages, alarm messages, control confirmation messages, and operation monitoring messages, need to be transmitted between distribution automation master stations, feeder terminals, ring main unit terminals, edge gateways, transformer area acquisition equipment, distribution IoT nodes, and dispatch communication equipment. These services typically have high reliability delivery requirements, some services are sensitive to in-order delivery, and the link environment is complex and variable. In existing distribution communication networks, common data delivery methods mainly include fixed-path transmission, timeout retransmission mechanisms, simple forward error correction mechanisms, and dynamic routing mechanisms based on instantaneous link quality. While these technical solutions can meet certain data transmission requirements in environments with relatively stable link quality, they still have the following problems in weak network environments: When link latency, link jitter, and sudden packet loss change significantly, relying solely on fixed-path transmission or simple timeout retransmission can easily lead to a significant increase in delivery latency, and even situations where multiple retransmissions fail to deliver stably. Existing routing mechanisms typically select paths based on instantaneous link quality or a single metric, making it difficult to consider the risk of data loss within a short future time window, path congestion, and on-order delivery constraints. This can easily lead to out-of-order delivery of data slices along different paths to the receiver. Existing redundancy coding or forward error correction schemes mostly use fixed redundancy ratios, making it difficult to dynamically adjust the amount of redundant transmission based on the actual recovery gap at the receiver. This can easily result in wasted link resources or insufficient redundancy when the recovery gap is large. Existing solutions usually focus on whether the service message content can be recovered, while rarely considering factors such as path state consistency, topology location consistency, and congestion status to constrain the final delivery action. This makes it difficult to meet the comprehensive requirements of stable, orderly, and reliable delivery of distribution services in weak network environments. Therefore, there is an urgent need for a reliable data delivery method and system that can adapt to the weak network environment of distribution networks and improve the reliable delivery capability, on-order delivery capability, and delivery stability of service messages. Summary of the Invention

[0003] The purpose of this invention is to provide a reliable data delivery method and system for distribution networks in weak network environments based on network-data fusion, in order to solve the problems of insufficient path status awareness, lack of foresight in bearer path selection, mismatch between redundant configuration and actual recovery gap, and difficulty in balancing delivery order and status consistency in existing technologies under weak network environments. By combining the service message slicing delivery process with network location recording, path status prediction, in-order delivery cost scheduling, coding redundancy scheduling, and location consistency gating delivery, the reliable delivery capability, in-order delivery capability, and delivery stability of distribution service messages under conditions of link fluctuation, latency jitter, sudden packet loss, and congestion are improved.

[0004] To achieve the above objectives, the present invention employs the following technical solution:

[0005] This invention provides a reliable data delivery method for distribution networks in weak network environments based on network-data integration, comprising the following steps:

[0006] S1. Generate the network-data fusion identifier corresponding to the business message, perform data slicing on the business message, and set a network location record field for each data slice;

[0007] S2. Construct a candidate path set based on the available links between the sender and receiver. Based on the path status information carried by historical confirmation feedback, historical retransmission feedback, and network location record fields, construct the path weak network status characteristics corresponding to each candidate path. Based on the path weak network status characteristics, calculate the average loss probability and location consistency score of each candidate path in the next round-trip prediction window.

[0008] S3. For the data slice to be sent, calculate the in-order delivery cost of each candidate path based on the predicted arrival time slot after the data slice is sent through each candidate path, the in-order delivery constraint of the preceding data slice, the current scheduling time slot, the average loss probability, the path queue occupancy and the location consistency score, select the candidate path with the smallest in-order delivery cost as the carrying path, and send the data slice through the carrying path.

[0009] S4. Based on the average loss probability, path delay, link jitter, and service priority of the current bearer path, determine the slice length, number of slices, and coding window width; generate coded redundant slices for data slices that have not received acknowledgment feedback, and determine the number of coded redundant slices to send based on the average loss probability of the bearer path and the number of linear independent degrees of freedom that have not yet been recovered as reported by the receiver.

[0010] S5. If no acknowledgment feedback is received for a consecutive preset number of transmission cycles on the bearer path, or if the average loss probability of the bearer path is greater than a preset threshold, update the candidate path set and redetermine the number of transmissions of the bearer path and the coded redundant slice.

[0011] S6. The receiving end performs association, decoding recovery, reassembly and integrity verification on the received data slices based on the network data fusion identifier and network location record field. When the preceding data slices to be delivered in sequence have been delivered, the current business message has been successfully decoded and recovered or can be reassembled without decoding, the integrity verification has passed and the location consistency score is not lower than the preset threshold, the in-sequence delivery is performed.

[0012] In a preferred embodiment, the network-data convergence identifier consists of a device node identifier, a service flow identifier, a message identifier, a slice identifier, and a network location identifier;

[0013] The network location identifier consists of a path identifier and a location sequence number;

[0014] The slice identifier is used to reflect the sequential position of the data slice in its respective business message;

[0015] Multiple data slices corresponding to the same business message within the same business flow have the same device node identifier, business flow identifier, and message identifier, but have different slice identifiers;

[0016] The sending and receiving ends perform data slice association, reassembly identification, and retransmission identification based on the network-data fusion identifier.

[0017] In a preferred embodiment, the network location record field is a fixed-length field, which includes a path identifier field, a location sequence number field, a path delay field, a link jitter field, a queue occupancy field, a link packet loss statistics status field, and a path hop number field.

[0018] During the forwarding of data slices along candidate paths, nodes along the route update the path delay field, link jitter field, queue occupancy field, link packet loss statistics status field, and path hop number field based on the link status monitored by the node, while the total length of the network location record field remains unchanged. The receiving end aggregates and analyzes the network location record information carried by multiple data slices corresponding to the same service message to construct the path weak network state characteristics corresponding to each candidate path and calculate the location consistency score.

[0019] In a preferred embodiment, the predicted average loss rate of the candidate paths is calculated according to the following formula:

[0020] ;

[0021] In the formula, Representing a path In scheduling time slots The average loss probability within the next round-trip prediction window; Representing a path In scheduling time slots The number of discrete time slots included in the corresponding round-trip prediction window, and It is a positive integer; This represents a path loss probability prediction model. This represents the set of parameters for the path loss probability prediction model. Representing a path In scheduling time slots The path weak network state feature vector is composed of historical confirmation feedback features, historical retransmission feedback features, and location-related state features extracted from network location record information. This indicates that the path loss probability prediction model predicts the probability of the path loss in the future. The real-valued prediction of each prediction slot output; This represents a probability mapping function that maps real-valued predictions to the probability of loss within the interval [0,1]. This represents the index of the relative prediction step size within the next round-trip prediction window, and .

[0022] In a preferred embodiment, the location consistency score is calculated according to the following formula:

[0023] ;

[0024] In the formula, Representing a path In scheduling time slots Positional consistency score; Representing a path In scheduling time slots The latency matching degree; Representing a path In scheduling time slots Path hop count matching degree; Representing a path In scheduling time slots The packet loss statistics status matching degree; Representing a path In scheduling time slots The queue occupies matching degree; , , , The weight coefficients are non-negative and satisfy the following conditions:

[0025] ;

[0026] The latency matching degree, path hop count matching degree, packet loss statistics status matching degree, and queue occupancy matching degree are all normalized quantities within the interval [0,1].

[0027] In a preferred embodiment, for business messages The A data slice in the path The in-order delivery cost is calculated using the following formula, and the path with the lowest cost is determined as the sending path:

[0028] ;

[0029] In the formula, Indicates business message The A data slice via path The cost of normalized in-order delivery delay during transmission; Represents data slices via path Predicted arrival time slot number after transmission; This indicates the predicted delivery slot number when the previous data slice satisfies the in-order delivery constraint. Indicates the reference time slot number, and It is a positive integer; Representing a path In scheduling time slots The average loss probability within the next round-trip prediction window; Representing a path In scheduling time slots Normalized queue occupancy; Representing a path In scheduling time slots Positional consistency score; , , and Weighting coefficients that are greater than zero; Indicates scheduling time slot The set of candidate paths; when hour, Take the current scheduling slot .

[0030] In a preferred embodiment, the slice length, number of slices, and encoding window width of the service message are determined based on the total length of the service message, the average loss probability of the current bearer path, path latency, link jitter, and service priority.

[0031] The number of coded redundant slices to be sent is determined based on the coding window width, the average loss probability of the current bearer path, and the number of currently unrecovered linear independent degrees of freedom reported by the receiver.

[0032] In a preferred embodiment, the coded redundancy slice is generated by linearly combining unconfirmed data slices within the current coding window over a preset finite domain using preset or randomly selected non-zero coding coefficients.

[0033] The coded redundancy slice carries coded coefficient identification information or seed information for recovering coded coefficients, so that the receiving end can perform decoding and recovery.

[0034] When the number of unconfirmed data slices in the current encoding window is 1, the encoding redundant slice is generated from the unconfirmed data slice;

[0035] When there are at least two unconfirmed data slices in the current encoding window, the redundant encoding slice is generated by a linear combination of multiple unconfirmed data slices.

[0036] In a preferred embodiment, the transmitting end updates the number of coded redundant slices to be sent corresponding to the current coding window based on the number of linear independent degrees of freedom that have not yet been recovered and the average loss probability of the current bearer path, as fed back by the receiving end.

[0037] When the number of linear independent degrees of freedom that have not yet been recovered is zero, stop sending coded redundant slices for the current coding window;

[0038] The receiving end performs in-order delivery only when the current business message meets the following conditions: the number of linear independent degrees of freedom that have not yet been restored corresponding to the current business message is zero, or the current business message can be reassembled without decoding; and the integrity check passes and the position consistency score is not lower than the preset threshold.

[0039] This invention also provides a reliable data delivery system for weak network environments in distribution networks based on network-data fusion, including a sending end, a receiving end, and one or more intermediate nodes located between the sending end and the receiving end, with the sending end and the receiving end communicating through the intermediate nodes; the sending end includes: an identification slicing module, used to generate a network-data fusion identifier corresponding to a service message, perform data slicing on the service message, and set a network location record field for each data slice; a path state construction module, used to construct a candidate path set based on the available links between the sending end and the receiving end, construct path weak network state characteristics corresponding to each candidate path based on historical confirmation feedback, historical retransmission feedback, and path state information carried by the network location record field, and calculate the average loss probability and location consistency score of each candidate path in the next round-trip prediction window based on the path weak network state characteristics; and a path scheduling module, used to schedule the data slices to be sent based on the predicted arrival time after the data slices are sent through each candidate path. The system calculates the in-order delivery cost of each candidate path based on the time slot, the in-order delivery constraints of the preceding data slice, the current scheduled time slot, the average loss probability, the path queue occupancy, and the location consistency score. It selects the candidate path with the lowest in-order delivery cost as the bearer path and sends the data slice through it. A redundancy coding scheduling module determines the slice length, number of slices, and coding window width based on the average loss probability of the current bearer path, path latency, link jitter, and service priority. It generates coded redundancy slices for data slices that have not received acknowledgment feedback and determines the number of coded redundancy slices to send based on the average loss probability of the bearer path and the number of currently unrecovered linear independent degrees of freedom reported by the receiver. A path update module updates the candidate path set and re-determines the number of bearer paths and coded redundancy slices to send when no acknowledgment feedback is received for a preset number of consecutive transmission cycles on the bearer path, or when the average loss probability of the bearer path exceeds a preset threshold.

[0040] The intermediate node includes: a location record update module, used to update the path status information carried in the network location record field according to the link status monitored by this node during the forwarding of data slices along the candidate path;

[0041] The receiving end includes: an association recovery module, used to perform association, decoding recovery, reassembly, and integrity verification on the received data slices based on the network data fusion identifier and network location record field; a delivery determination module, used to perform in-order delivery when the preceding data slices to be delivered in sequence have been delivered, the current business message has been successfully decoded and recovered or can be reassembled without decoding, the integrity verification has passed, and the location consistency score is not lower than a preset threshold; and a degree of freedom feedback module, used to feed back to the sending end the number of linear independent degrees of freedom that have not yet been recovered.

[0042] The beneficial effects of this invention are as follows: By generating a network-data fusion identifier for service messages and setting a network location record field, this invention enables the sending end, receiving end, and intermediate nodes to establish a unified data slice association, path status record, and feedback mechanism around the same service message, providing fundamental support for reliable multi-path delivery in weak network environments. By constructing path weak network state characteristics based on historical confirmation feedback, historical retransmission feedback, and network location record information, and calculating the average loss probability and location consistency score within the next round-trip prediction window based on these characteristics, path selection no longer relies on a single instantaneous quality indicator but can predict and evaluate link risks and state consistency within a short future time window, thereby improving the accuracy of bearer path selection. By comprehensively predicting arrival time slots, on-order delivery constraints of preceding data slices, average loss probability, path queue occupancy, and location consistency score to construct on-order delivery costs, and determining the path with the minimum on-order delivery cost as the bearer path, this invention enables bearer path selection to simultaneously consider reliable recovery, congestion control, on-order delivery, and state consistency requirements, reducing the probability of out-of-order delivery and repeated recovery overhead in weak network environments. By combining the average loss probability of the current bearer path with the number of linear independent degrees of freedom that have not yet been recovered as reported by the receiver, the number of coded redundancy slices to be sent is dynamically determined. This allows the redundancy configuration to adapt to the actual recovery gap at the receiver, avoiding the waste of link resources caused by fixed redundancy configurations and improving the effectiveness of redundant transmission. By using the location consistency score as the final on-order delivery gating condition, the receiver decides whether to execute delivery based on whether the path status meets the predetermined service requirements, provided that the current service message content has been correctly recovered. This improves the stability, consistency, and engineering applicability of power distribution service message delivery results. This invention can improve the reliable delivery capability, on-order delivery capability, and path status adaptation capability of power distribution service messages under weak network conditions such as link fluctuations, latency jitter, sudden packet loss, and local congestion. Attached Figure Description

[0043] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein: Figure 1 This is a flowchart of the method of the present invention; Figure 2 This is a schematic diagram of the modular structure of the system of the present invention. Detailed Implementation

[0044] The present invention will be further described below with reference to embodiments. It should be understood that the following embodiments are for illustrative purposes only and are not intended to limit the scope of protection of the present invention. Where there is no conflict, the technical features of the various embodiments can be combined with each other.

[0045] This invention provides a reliable data delivery method and system for distribution networks in weak network environments based on network-data fusion. It is applicable to service message transmission scenarios between distribution automation master stations, edge gateways, feeder terminals, ring main units, transformer area acquisition devices, and dispatch communication equipment. It is particularly suitable for weak network environments characterized by link fluctuations, latency jitter, sudden packet loss, local congestion, frequent switching of available paths, and unstable acknowledgment feedback. The sending end, the receiving end, and one or more intermediate nodes located between the sending and receiving ends collaborate to complete service message slicing, path evaluation, bearer path selection, redundancy recovery, and gated delivery.

[0046] like Figure 1 As shown, this is an embodiment of the present invention, which provides a reliable data delivery method for a distribution network in a weak network environment based on network-data fusion, including the following steps:

[0047] S1. Generate the network-data fusion identifier corresponding to the business message, perform data slicing on the business message, and set a network location record field for each data slice;

[0048] The network-data fusion identifier consists of a device node identifier, a service flow identifier, a message identifier, a slice identifier, and a network location identifier. Specifically, the device node identifier uniquely identifies the sending device; the service flow identifier distinguishes different service categories or different service logic channels; the message identifier distinguishes different service messages within the same service flow; the slice identifier reflects the sequential position of the data slice within its respective service message; and the network location identifier, composed of a path identifier and a location sequence number, characterizes the path location attribute corresponding to the current data slice.

[0049] The sending end segments the service message according to the total length of the service message and the preset slicing rules. If the length of the service message does not exceed the preset single slice length, a single data slice is generated; if the length of the service message exceeds the preset single slice length, it is divided into multiple data slices in sequence. For multiple data slices corresponding to the same service message within the same service flow, their device node identifier, service flow identifier, and message identifier are the same, but their slice identifiers are different.

[0050] After completing the slicing, the sending end adds a network location record field to each data slice. The network location record field preferably uses a fixed-length structure. In phase S1, the sending end at least writes the path identifier and location sequence number; the remaining links-related fields are initialized to zero, preset default values, or reserved values ​​for subsequent updates by nodes along the path. The receiving end performs data slice association, reassembly identification, and retransmission identification based on the network-data fusion identifier, thereby merging data slices belonging to the same service message into the same message context and identifying their order based on the slice identifier.

[0051] S2. Construct a candidate path set based on the available links between the sender and receiver. Based on the path status information carried by historical confirmation feedback, historical retransmission feedback, and network location record fields, construct the path weak network status characteristics corresponding to each candidate path. Based on the path weak network status characteristics, calculate the average loss probability and location consistency score of each candidate path in the next round-trip prediction window.

[0052] The network location record field is a fixed-length field, including at least a path identifier field, a location sequence number field, a path delay field, a link jitter field, a queue occupancy field, a link packet loss statistics status field, and a path hop count field. During the forwarding of data slices along candidate paths, nodes along the route update the path delay field, link jitter field, queue occupancy field, link packet loss statistics status field, and path hop count field based on the link status monitored by the node, while maintaining the total length of the network location record field. The receiving end aggregates and analyzes the network location record information carried by multiple data slices corresponding to the same service message to obtain path delay statistics, link jitter statistics, queue occupancy statistics, packet loss statistics status, and path hop count statistics, and returns these as path status input information to the sending end.

[0053] The sending end concatenates historical acknowledgment feedback features, historical retransmission feedback features, and location-related state features extracted from network location records in a preset order to form a path-weak network state feature vector. In a preferred embodiment, the feature vector sequentially includes acknowledgment feedback statistics, retransmission feedback statistics, path delay statistics, link jitter statistics, queue occupancy statistics, packet loss statistics, path hop count statistics, and location-related statistics, with the total dimension denoted as [missing information]. .

[0054] The predicted average loss rate for the candidate paths is calculated using the following formula:

[0055] ;

[0056] In the formula, Representing a path In scheduling time slots The average loss probability within the next round-trip prediction window; Representing a path In scheduling time slots The number of discrete time slots included in the corresponding round-trip prediction window, and It is a positive integer; This represents a path loss probability prediction model. This represents the set of parameters for the path loss probability prediction model. This is a probability mapping function. Representing a path In scheduling time slots The path weak network state feature vector is composed of historical confirmation feedback features, historical retransmission feedback features, and location-related state features extracted from network location record information. This indicates that the path loss probability prediction model predicts the probability of the path loss in the future. The real-valued prediction of each prediction slot output; This represents a probability mapping function that maps real-valued predictions to the probability of loss within the interval [0,1]. This represents the index of the relative prediction step size within the next round-trip prediction window, and .

[0057] In a preferred embodiment Based on the estimated round-trip time of the path With a single discrete time slot length The ratio is determined, that is:

[0058] ;

[0059] In a preferred embodiment, the path loss probability prediction model employs a three-layer fully connected network structure, including an input layer, a first hidden layer, a second hidden layer, and an output layer. The input layer receives... The state feature vector of the dimensional path weak network; the first and second hidden layers use the ReLU activation function; the output layer has an output length of... The real-valued prediction vector is used; the probability mapping function is the Sigmoid function. The input of the training samples is the historical path weak network state feature vector, and the labels are the future first to second... The model generates a binary sequence of lost data or an actual loss ratio sequence for each predicted time slot. After training, the model parameters are deployed at the transmitting end or edge computing nodes to output the average loss probability of candidate paths online.

[0060] The location consistency score is calculated according to the following formula:

[0061] ;

[0062] In the formula, Representing a path In scheduling time slots Positional consistency score; Representing a path In scheduling time slots The latency matching degree; Representing a path In scheduling time slots Path hop count matching degree; Representing a path In scheduling time slots The packet loss statistics status matching degree; Representing a path In scheduling time slots The queue occupies matching degree; , , , The weight coefficients are non-negative and satisfy the following conditions:

[0063] ;

[0064] In one embodiment, the latency matching degree can be obtained by mapping the deviation between the current path latency and the target latency baseline; the path hop count matching degree can be obtained by mapping the deviation between the current path hop count and the target hop count template; the packet loss statistics status matching degree can be obtained by mapping the deviation between the current packet loss statistics status value and the preset packet loss template; and the queue occupancy matching degree can be obtained by mapping the deviation between the current queue occupancy statistics value and the preset acceptable occupancy range. All of the above matching degrees are normalized quantities within the interval [0,1].

[0065] S3. For the data slice to be sent, calculate the in-order delivery cost of each candidate path based on the predicted arrival time slot after the data slice is sent through each candidate path, the in-order delivery constraint of the preceding data slice, the current scheduling time slot, the average loss probability, the path queue occupancy and the location consistency score, select the candidate path with the smallest in-order delivery cost as the carrying path, and send the data slice through the carrying path.

[0066] For business messages The A data slice in the path The in-order delivery cost is calculated using the following formula, and the path with the lowest cost is determined as the sending path:

[0067] ;

[0068] In the formula, Indicates business message The A data slice via path The cost of normalized in-order delivery delay during transmission; Represents data slices via path Predicted arrival time slot number after transmission; This indicates the predicted delivery slot number when the previous data slice satisfies the in-order delivery constraint. Indicates the reference time slot number, and It is a positive integer; Representing a path In scheduling time slots The average loss probability within the next round-trip prediction window; Representing a path In scheduling time slots Normalized queue occupancy; Representing a path In scheduling time slots Positional consistency score; , , and Weighting coefficients that are greater than zero; Indicates scheduling time slot The set of candidate paths; when hour, Take the current scheduling slot .

[0069] In one embodiment, the predicted arrival slot number is calculated based on the path delay estimate, link jitter estimate, and queue occupancy. If the path... In scheduling time slots The path delay estimate is The estimated link jitter value is The normalized queue occupancy is The length of a single discrete time slot is Then the predicted transmission delay can be expressed as:

[0070] ;

[0071] The corresponding predicted arrival time slot number is: ;

[0072] Previous data slice predicted delivery slot number Used to embody in-order delivery constraints. When Time to take ;when If a preceding data slice can be delivered without waiting for other preceding slices, its predicted arrival time slot number is used; if a preceding data slice is still constrained by an earlier preceding slice, the larger value between its own predicted arrival time slot number and the predicted delivery time slot number of the earlier preceding slice is used.

[0073] Through the aforementioned cost function, the sender simultaneously considers orderliness, reliability, congestion status, and location consistency, thereby selecting a more suitable bearer path for the current business message slice.

[0074] S4. Based on the average loss probability, path delay, link jitter, and service priority of the current bearer path, determine the slice length, slice quantity, and coding window width; generate coded redundant slices for data slices that have not received acknowledgment feedback, and determine the number of coded redundant slices to send based on the average loss probability of the bearer path and the number of linear independent degrees of freedom that have not yet been recovered as reported by the receiver.

[0075] Let the total length of the business message be The slice length is The number of slices is The encoding window width is The slice length can be set according to the preset base slice length. and adjustment coefficient Sure:

[0076] ;

[0077] In the formula, It decreases as the average loss probability and link jitter increase, and it also decreases as the service priority increases. The number of slices can be expressed as:

[0078] ;

[0079] Encoding window width You can set the minimum window width. and preset maximum window width Adaptively select from among them, and satisfy .

[0080] Unacknowledged data slices within the current encoding window refer to raw data slices that the sender has already transmitted but has not yet received acknowledgment from the receiver. The sender maintains a set of unacknowledged data slices within the current encoding window and dynamically updates this set upon receiving acknowledgment.

[0081] The coded redundancy slice is generated over a preset finite field by linearly combining unconfirmed data slices within the current coding window using preset or randomly selected non-zero coding coefficients. In a preferred embodiment, the preset finite field adopts... or For the set of unconfirmed data slices in the current encoding window ,in , encoding redundant slices It can be represented as: In the formula, These are the non-zero coding coefficients selected from a preset finite field.

[0082] The coded redundancy slice carries coded coefficient identification information or seed information for recovering coded coefficients, which is then used by the receiving end for decoding and recovery. If the coded coefficient identification information is carried directly, the coded coefficient values ​​and their position mapping relationships are written into the header of the coded redundancy slice; if seed information is carried, the sending end and the receiving end pre-agree on pseudo-random number generation rules, and the receiving end recovers the coded coefficients based on the seed information.

[0083] When the number of unconfirmed data slices in the current encoding window is 1, the encoding redundant slice is generated from the unconfirmed data slice; when the number of unconfirmed data slices in the current encoding window is not less than 2, the encoding redundant slice is generated by a linear combination of multiple unconfirmed data slices.

[0084] The receiver maintains the coefficient matrix and reception matrix corresponding to the current coding window and calculates the number of linearly independent degrees of freedom that have not yet been recovered. The number of linearly independent degrees of freedom that have not yet been recovered represents the difference between the number of linearly independent coding equations required to recover all the original data slices within the current coding window and the number of linearly independent coding equations that the receiver has already obtained. If the current coding window width is... The number of linear independent coding equations currently obtained by the receiver is The number of linear independent degrees of freedom that have not yet been recovered can be expressed as: The number of coded redundant slices to be transmitted is determined based on the coding window width, the average loss probability of the current bearer path, and the number of currently unrecovered linear independent degrees of freedom reported by the receiver. Let the number of coded redundant slices be... In one embodiment, it can be determined as follows: In the formula, and This is a non-negative adjustment coefficient. Those skilled in the art can also determine the number of coded redundant slices to send using equivalent methods such as table lookup, piecewise functions, or threshold rules.

[0085] S5. If no acknowledgment feedback is received for a consecutive preset number of transmission cycles on the bearer path, or if the average loss probability of the bearer path is greater than a preset threshold, update the candidate path set and redetermine the number of transmissions of the bearer path and the coded redundant slice.

[0086] In one embodiment, failure to receive acknowledgment feedback for a preset number of consecutive transmission cycles indicates that the sender is continuously... No valid acknowledgment feedback related to the current bearer path was received within a certain transmission cycle, among which It is a positive integer. If the carrying path satisfies: In the formula, If the average loss probability threshold is set, the sender will determine that the current bearer path has a high risk of loss within the next round-trip prediction window, and will also trigger path updates and rescheduling.

[0087] After triggering a path update, the sending end rereads the current link state table, reachable path cache table, and the feedback results of the most recent network location record, and updates the candidate path set. The update methods include: downgrading the current bearer path from the high-priority candidate path list or temporarily removing it from the candidate path set; incorporating recently rediscovered available links into the path combination to supplement and form new candidate paths; recalculating the path's weak network state characteristics, average loss probability, and location consistency score for each path in the original candidate path set; and deleting paths that are no longer reachable, severely congested, or experiencing persistently high packet loss.

[0088] After the candidate path set is updated, the sender re-executes the path evaluation and bearer path selection processes in S2 and S3, and recalculates the number of coded redundant slices to be sent based on the new average loss probability of the bearer path, the current coding window width, and the number of linear independent degrees of freedom that have not yet been recovered as reported by the receiver.

[0089] S6. The receiving end performs association, decoding recovery, reassembly and integrity verification on the received data slices based on the network data fusion identifier and network location record field. When the preceding data slices to be delivered in sequence have been delivered, the current business message has been successfully decoded and recovered or can be reassembled without decoding, the integrity verification has passed and the location consistency score is not lower than the preset threshold, the in-sequence delivery is performed.

[0090] The receiving end, based on the device node identifier, service flow identifier, and message identifier in the network-data fusion identifier, merges data slices belonging to the same service message into the same cache object. It then identifies the sequential position of the current data slice in the original service message based on the slice identifier and records its path source and location attributes according to the path identifier and location sequence number in the network location record field. For duplicate data slices, the receiving end can determine whether it is a duplicate version or a retransmission of a previously received slice, and accordingly perform overwrite, retention, or discard processing.

[0091] The receiving end maintains the coefficient matrix and data matrix corresponding to the received slices within the current encoding window. When the receiving end obtains a sufficient number of linearly independent slices, it performs decoding and recovery. In a preferred embodiment, the receiving end uses Gaussian elimination, finite field matrix inversion, or other linear equation solving methods to solve the encoding matrix within the current encoding window, thereby recovering the original data slices. If all the original data slices within the current encoding window have been received, decoding is unnecessary, and the receiving end can directly enter the reassembly stage.

[0092] The transmitter updates the number of coded redundant slices to be transmitted for the current coding window based on the number of currently unrecovered linear independent degrees of freedom reported by the receiver and the average loss probability of the current bearer path. After receiving a new raw data slice or coded redundant slice, the receiver recalculates the rank of the decoding matrix within the current coding window and obtains the number of currently unrecovered linear independent degrees of freedom. The receiver can [do something] each time. When changes occur, they are returned to the sender via an acknowledgment feedback message or a dedicated feedback message.

[0093] When the number of linear independent degrees of freedom that have not yet been recovered is zero, the sender stops sending coded redundant slices for the current coding window. Specifically, when the sender receives... After receiving feedback, the current encoding window is marked as fully restored, and the generation of new encoding redundancy slices for that encoding window is stopped.

[0094] Once the receiving end has completed decoding and recovery, or has received all the original data slices without needing decoding, it reassembles the original data slices according to their segment identifiers to restore the original business message. After reassembly, the receiving end performs an integrity check on the restored business message. Integrity checks can be performed based on message checksums, CRC checksums, digest values, or hash values. Only when the integrity check passes is the receiving end certain that the current business message has been correctly restored at the content level.

[0095] The receiving end performs in-order delivery only when the current business message meets the following conditions: the number of currently unrecovered linear independent degrees of freedom corresponding to the current business message is zero, or the current business message can be reassembled without decoding; and the integrity check passes, and the position consistency score is not lower than a preset threshold. Specifically, when performing delivery determination, the receiving end checks at least the following conditions: the preceding data slice to be delivered in order has been delivered; the current business message meets the recovery conditions; the current business message has passed the integrity check; and the position consistency score corresponding to the current business message is not lower than a preset threshold.

[0096] If all of the above conditions are met, the delivery determination unit will submit the current business message to the upper-layer business processing module; if any condition is not met, the current business message will continue to be stored in the delivery queue, waiting for the previous message to be delivered, waiting for more slices to arrive, waiting for the integrity verification to pass, or waiting for the location consistency score to rise above the threshold.

[0097] like Figure 2As shown, the present invention also provides a reliable data delivery system for a distribution network in a weak network environment based on network-data integration, including a transmitter, a receiver, and one or more intermediate nodes located between the transmitter and the receiver, wherein the transmitter and the receiver are connected through the intermediate nodes.

[0098] The sending end includes an identification slicing module, a path state construction module, a path scheduling module, a redundancy coding scheduling module, and a path update module. The identification slicing module generates network-data fusion identifiers corresponding to service messages, slices the service messages, and sets a network location record field for each data slice. The path state construction module constructs a set of candidate paths based on available links between the sending and receiving ends. Based on historical acknowledgment feedback, historical retransmission feedback, and path state information carried by the network location record field, it constructs path weak network state characteristics for each candidate path and calculates the average loss probability and location consistency score for each candidate path within the next round-trip prediction window based on these characteristics. The path scheduling module calculates the in-order delivery cost for each candidate path for the data slice to be sent, based on the predicted arrival time slot after the data slice is sent via each candidate path, the in-order delivery constraints of the preceding data slice, the current scheduling time slot, the average loss probability, the path queue occupancy, and the location consistency score. It selects the candidate path with the lowest in-order delivery cost as the bearer path and sends the data slice through this bearer path. The redundancy coding scheduling module determines the slice length, number of slices, and coding window width based on the average loss probability, path delay, link jitter, and service priority of the current bearer path, generates coded redundancy slices, and updates the number of coded redundancy slices to be sent based on the number of linear independent degrees of freedom that have not yet been recovered, as reported by the receiver. The path update module updates the candidate path set and re-determines the number of bearer paths and coded redundancy slices to be sent when no acknowledgment feedback is received for a consecutive preset number of transmission cycles on a bearer path, or when the average loss probability of a bearer path exceeds a preset threshold.

[0099] The intermediate node includes a location record update module. This module updates the path status information carried in the network location record field based on the link status monitored by the node during the forwarding of data slices along candidate paths.

[0100] The receiving end includes an association recovery module, a delivery determination module, and a degree-of-freedom feedback module. The association recovery module performs association, decoding recovery, reassembly, and integrity verification on the received data slices based on the network-data fusion identifier and network location record field. The delivery determination module performs in-order delivery when the preceding data slices to be delivered in sequence have been delivered, the current business message has been successfully decoded and recovered or can be reassembled without decoding, the integrity verification has passed, and the location consistency score is not lower than a preset threshold. The degree-of-freedom feedback module provides feedback to the sending end on the number of currently unrecovered linear independent degrees of freedom.

[0101] In summary, this invention addresses the lack of unified coordination among path state utilization, path scheduling, redundancy control, and delivery determination during service message transmission in weak distribution network environments. It constructs a complete processing link from service message slicing, path state construction, path scheduling, encoding recovery to delivery determination at the receiving end, creating a seamless and cooperative collaborative mechanism between the sender, receiver, and intermediate nodes. The scheme has a clear overall structure and a complete processing flow, suitable for service message transmission between distribution automation master stations, edge gateways, feeder terminals, ring main unit terminals, and dispatch communication equipment. It also facilitates parameter configuration and engineering deployment according to different network conditions and service requirements, demonstrating strong practical application potential and widespread application value.

[0102] The above are merely preferred embodiments of this application and are not intended to limit the scope of protection of this application. All equivalent substitutions, improvements, or modifications made within the spirit and principles of this application shall fall within the scope of protection of this application. The scope of protection of this application is defined by the claims.

Claims

1. A reliable data delivery method for distribution networks in weak network environments based on network-data integration, characterized in that, Includes the following steps: Generate network-data fusion identifiers corresponding to business messages, perform data slicing on business messages, and set network location record fields for each data slice; A candidate path set is constructed based on the available links between the sender and receiver. Based on the path status information carried by historical confirmation feedback, historical retransmission feedback and network location record fields, the path weak network status characteristics corresponding to each candidate path are constructed. Based on the weak network state characteristics of the path, the average loss probability and location consistency score of each candidate path in the next round-trip prediction window are calculated. For the data slice to be sent, the on-order delivery cost of each candidate path is calculated based on the predicted arrival time slot after the data slice is sent through each candidate path, the on-order delivery constraint of the preceding data slice, the current scheduling time slot, the average loss probability, the path queue occupancy and the location consistency score. The candidate path with the smallest on-order delivery cost is selected as the carrying path, and the data slice is sent through the carrying path. The slice length, number of slices, and encoding window width are determined based on the average loss probability, path latency, link jitter, and service priority of the current bearer path. For data slices that have not received acknowledgment feedback, generate coded redundant slices, and determine the number of coded redundant slices to send based on the average loss probability of the bearer path and the number of linear independent degrees of freedom that have not yet been recovered as reported by the receiver. If no acknowledgment is received for a consecutive preset number of transmission cycles on the bearer path, or if the average loss probability of the bearer path is greater than a preset threshold, the candidate path set is updated, and the number of transmissions of the bearer path and the coded redundant slice is re-determined. The receiving end performs association, decoding recovery, reassembly, and integrity verification on the received data slices based on the network data fusion identifier and network location record field; sequential delivery is performed when the preceding data slices to be delivered in sequence have been delivered, the current business message has been successfully decoded and recovered or can be reassembled without decoding, the integrity verification has passed, and the location consistency score is not lower than the preset threshold.

2. The reliable data delivery method for a distribution network in a weak network environment based on network-data fusion as described in claim 1, characterized in that, The network-data convergence identifier consists of a device node identifier, a service flow identifier, a message identifier, a slice identifier, and a network location identifier; The network location identifier consists of a path identifier and a location sequence number; The slice identifier is used to reflect the sequential position of the data slice in its respective business message; Multiple data slices corresponding to the same business message within the same business flow have the same device node identifier, business flow identifier, and message identifier, but have different slice identifiers; The sending and receiving ends perform data slice association, reassembly identification, and retransmission identification based on the network-data fusion identifier.

3. The reliable data delivery method for a distribution network in a weak network environment based on network-data fusion as described in claim 2, characterized in that, The network location record field is a fixed-length field, which includes a path identifier field, a location sequence number field, a path delay field, a link jitter field, a queue occupancy field, a link packet loss statistics status field, and a path hop number field. During the forwarding of data slices along candidate paths, nodes along the route update the path delay field, link jitter field, queue occupancy field, link packet loss statistics status field, and path hop number field based on the link status monitored by the node, while the total length of the network location record field remains unchanged. The receiving end aggregates and analyzes the network location record information carried by multiple data slices corresponding to the same service message to construct the path weak network state characteristics corresponding to each candidate path and calculate the location consistency score.

4. The reliable data delivery method for a distribution network in a weak network environment based on network-data fusion as described in claim 3, characterized in that, The predicted average loss rate for the candidate paths is calculated using the following formula: ； In the formula, Representing a path In scheduling time slots The average loss probability within the next round-trip prediction window; Representing a path In scheduling time slots The number of discrete time slots included in the corresponding round-trip prediction window, and It is a positive integer; This represents a path loss probability prediction model. This represents the set of parameters for the path loss probability prediction model. Representing a path In scheduling time slots The path weak network state feature vector is composed of historical confirmation feedback features, historical retransmission feedback features, and location-related state features extracted from network location record information. This indicates that the path loss probability prediction model predicts the probability of the path loss in the future. The real-valued prediction of each prediction slot output; This represents a probability mapping function that maps real-valued predictions to the probability of loss within the interval [0,1]. This represents the index of the relative prediction step size within the next round-trip prediction window, and .

5. A reliable data delivery method for a distribution network in a weak network environment based on network-data fusion, as described in claim 4, is characterized in that... The location consistency score is calculated according to the following formula: ； In the formula, Representing a path In scheduling time slots Positional consistency score; Representing a path In scheduling time slots The latency matching degree; Representing a path In scheduling time slots Path hop count matching degree; Representing a path In scheduling time slots The packet loss statistics status matching degree; Representing a path In scheduling time slots The queue occupies matching degree; , , , The weight coefficients are non-negative and satisfy the following conditions: ； The latency matching degree, path hop count matching degree, packet loss statistics status matching degree, and queue occupancy matching degree are all normalized quantities within the interval [0,1].

6. The reliable data delivery method for a distribution network in a weak network environment based on network-data fusion as described in claim 5, characterized in that, For business messages The A data slice in the path The in-order delivery cost is calculated using the following formula, and the path with the lowest cost is determined as the sending path: ； In the formula, Indicates business message The A data slice via path The cost of normalized in-order delivery delay during transmission; Represents data slices via path Predicted arrival time slot number after transmission; This indicates the predicted delivery slot number when the previous data slice satisfies the in-order delivery constraint. Indicates the reference time slot number, and It is a positive integer; Representing a path In scheduling time slots The average loss probability within the next round-trip prediction window; Representing a path In scheduling time slots Normalized queue occupancy; Representing a path In scheduling time slots Positional consistency score; , , and Weighting coefficients that are greater than zero; Indicates scheduling time slot The set of candidate paths; when hour, Take the current scheduling slot .

7. A reliable data delivery method for a distribution network in a weak network environment based on network-data fusion, as described in claim 6, is characterized in that... The slice length, number of slices, and encoding window width of the service message are determined based on the total length of the service message, the average loss probability of the current bearer path, path latency, link jitter, and service priority. The number of coded redundant slices to be sent is determined based on the coding window width, the average loss probability of the current bearer path, and the number of currently unrecovered linear independent degrees of freedom reported by the receiver.

8. A reliable data delivery method for a distribution network in a weak network environment based on network-data fusion, as described in claim 7, is characterized in that... The coded redundant slice is generated over a preset finite field by linearly combining unconfirmed data slices within the current coding window according to preset or randomly selected non-zero coding coefficients. The coded redundancy slice carries coded coefficient identification information or seed information for recovering coded coefficients, so that the receiving end can perform decoding and recovery. When the number of unconfirmed data slices in the current encoding window is 1, the encoding redundant slice is generated from the unconfirmed data slice; When there are at least two unconfirmed data slices in the current encoding window, the redundant encoding slice is generated by a linear combination of multiple unconfirmed data slices.

9. A reliable data delivery method for a distribution network in a weak network environment based on network-data fusion, as described in claim 1, is characterized in that... The transmitter updates the number of encoded redundant slices to be sent for the current encoding window based on the number of linear independent degrees of freedom that have not yet been recovered and the average loss probability of the current bearer path, as reported by the receiver. When the number of linear independent degrees of freedom that have not yet been recovered is zero, stop sending coded redundant slices for the current coding window; The receiving end performs in-order delivery only when the current business message meets the following conditions: the number of linear independent degrees of freedom that have not yet been restored corresponding to the current business message is zero, or the current business message can be reassembled without decoding; and the integrity check passes and the position consistency score is not lower than the preset threshold.

10. A reliable data delivery system for distribution networks in weak network environments based on network-data integration, characterized in that, The system includes a sender, a receiver, and one or more intermediate nodes located between the sender and receiver, with the sender and receiver communicating through the intermediate nodes. The sender includes: an identifier slice module, used to generate a network-data fusion identifier corresponding to a service message, slice the service message into data slices, and set a network location record field for each data slice; a path state construction module, used to construct a candidate path set based on available links between the sender and receiver, construct path weak network state characteristics corresponding to each candidate path based on historical confirmation feedback, historical retransmission feedback, and path state information carried by the network location record field, and calculate the average loss probability and location consistency score of each candidate path in the next round-trip prediction window based on the path weak network state characteristics; and a path scheduling module, used to schedule data slices to be sent based on the predicted arrival time slots after the data slices are sent through each candidate path, the on-order delivery constraints of preceding data slices, and so on. The system calculates the in-order delivery cost of each candidate path based on the current scheduling time slot, average loss probability, path queue occupancy, and location consistency score. It selects the candidate path with the lowest in-order delivery cost as the bearer path and sends data slices through it. The redundant coding scheduling module determines the slice length, number of slices, and coding window width based on the average loss probability, path latency, link jitter, and service priority of the current bearer path. It generates coded redundant slices for data slices that have not received acknowledgment feedback and determines the number of coded redundant slices to send based on the average loss probability of the bearer path and the number of currently unrecovered linear independent degrees of freedom reported by the receiver. The path update module updates the candidate path set and re-determines the number of bearer paths and coded redundant slices to send when no acknowledgment feedback is received for a preset number of consecutive transmission cycles on the bearer path, or when the average loss probability of the bearer path exceeds a preset threshold. The intermediate node includes: a location record update module, used to update the path status information carried in the network location record field according to the link status monitored by this node during the forwarding of data slices along the candidate path; The receiving end includes: an association recovery module, used to perform association, decoding recovery, reassembly, and integrity verification on the received data slices based on the network data fusion identifier and network location record field; a delivery determination module, used to perform in-order delivery when the preceding data slices to be delivered in sequence have been delivered, the current business message has been successfully decoded and recovered or can be reassembled without decoding, the integrity verification has passed, and the location consistency score is not lower than a preset threshold; and a degree of freedom feedback module, used to feed back to the sending end the number of linear independent degrees of freedom that have not yet been recovered.

Images