Redundant transmission system of remote instructions of water conservancy dispatching center

Abstract

The present application relates to the field of water conservancy remote control transmission technology, specifically to a redundant transmission system of remote instructions of a water conservancy dispatching center, comprising: an instruction analysis module, an instruction processing and weighting module, a path calculation module and a decision generation module. The instruction analysis module receives remote control instructions and extracts relevant information to generate an initial instruction data set. The instruction processing and weighting module performs structured analysis on instruction semantics to form an instruction logical dependency graph, performs credibility weighting mapping on priority identification to obtain an initial transmission weight, and completes the binding of the two. The path calculation module determines the primary and backup path node sequences, calculates the path evaluation value and the redundancy value in combination with the instruction historical execution record. The decision generation module inputs multiple types of parameters into a path-instruction joint decision model to generate a transmission scheduling scheme. The present scheme realizes the joint scheduling of instruction characteristics and transmission paths, and optimizes the redundant transmission effect of remote instructions of water conservancy dispatching.

Description

Technical Field

[0001] This invention relates to the field of water conservancy remote control transmission technology, and in particular to a redundant transmission system for remote commands from a water conservancy dispatch center. Background Technology

[0002] The existing remote command transmission system of the water conservancy dispatch center mainly completes the reception, basic parsing and routine primary and backup path transmission operations of remote control commands. The system only extracts the basic identification information of the commands, generates simplified command data, and only performs basic parsing processing on the commands. Path planning relies on the preset water conservancy network topology and adopts a fixed primary path and backup path allocation mode. Transmission scheduling is only executed based on the status of the basic path. The system does not perform correlation analysis between the attributes of the commands themselves and the path scheduling. Command processing and path planning are independent operational links.

[0003] The existing transmission scheme does not perform semantic structured parsing of the instruction semantic description, and cannot form corresponding instruction logical dependencies. The instruction priority identifier is only read directly and used without performing credibility weighted mapping. The instruction logical structure and transmission weight cannot be associated and bound. The determination of the path node sequence does not combine the instruction historical execution record, and it is impossible to calculate the current path evaluation value of the main path node sequence and the path redundancy value of the alternative path node sequence. The transmission decision is generated based on a single parameter and does not form a joint decision-making mode that combines instructions and paths.

[0004] This invention requires semantic structured parsing and priority-reliability weighted binding of remote control commands, and calculation of specific parameters for primary and backup paths by combining command history execution records. Command-related parameters and path-related parameters are input into a dedicated model to achieve joint decision-making, solving the problems of command processing and path scheduling being disconnected and insufficient adaptability of redundant transmission scheduling in conventional systems, and adapting to the actual operational needs of redundant transmission of remote commands in water conservancy dispatch centers. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of existing technologies and propose a redundant transmission system for remote commands of water conservancy dispatch centers.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: a redundant transmission system for remote commands of a water conservancy dispatch center, comprising: The instruction parsing module receives remote control instructions from the water conservancy dispatch center and extracts the instruction semantic description, instruction priority identifier, instruction receiver identifier, and instruction historical execution record of the remote control instructions to generate an initial instruction dataset. The instruction processing weighting module performs semantic structure parsing on the semantic description of the instruction in the initial instruction dataset to form an instruction logical dependency graph, performs credibility weighting mapping on the instruction priority identifier to form an initial transmission weight, and associates and binds the instruction logical dependency graph with the initial transmission weight. The path calculation module determines the main path node sequence and the alternative path node sequence from the pre-constructed water conservancy network topology according to the instruction receiving end identifier, and calculates the current path evaluation value of the main path node sequence and the path redundancy value of the alternative path node sequence based on the instruction historical execution record. The decision generation module inputs the instruction logic dependency graph, the initial transmission weight, the current path evaluation value, and the path redundancy value into the path-instruction joint decision model to generate a transmission task scheduling scheme.

[0007] As a further aspect of the present invention, the semantic semantic description of the initial instruction dataset is subjected to semantic structure parsing processing to form an instruction logical dependency graph, including: Identify the instruction verbs, controlled objects, control parameters, and constraints in the instruction semantic description, and generate instruction element tuples; Based on a pre-set knowledge base of water conservancy control operation logic, logical relationships are deduced from the instruction element tuples to determine the control dependency, temporal dependency, and data dependency relationships among the instruction element tuples. Using the instruction element tuple as nodes and the control dependency, timing dependency, and data dependency as directed edges, an instruction logic dependency graph representing the internal execution logic of an instruction is constructed.

[0008] As a further aspect of the present invention, the instruction priority identifier is subjected to a reliability-weighted mapping process to form an initial transmission weight, including: Query the preset instruction priority-weight mapping table to obtain the baseline transmission weight corresponding to the instruction priority identifier; The success rate and timeliness of executing the same or similar remote control commands are retrieved from the command history execution records. Based on the execution success rate and the timeliness achievement rate, the baseline transmission weight is dynamically adjusted. The dynamic adjustment process is as follows: the weighted average of the execution success rate and the timeliness achievement rate is used as a adjustment factor, and the adjustment factor is multiplied by the baseline transmission weight to obtain the initial transmission weight.

[0009] As a further aspect of the present invention, based on the instruction execution history records, the current path evaluation value of the main path node sequence and the path redundancy value of the alternative path node sequences are calculated, including: Extract a set of historical transmission records for the instruction receiver identifier from the instruction execution history records; Based on the historical transmission record set, the historical transmission success frequency, average transmission delay, and delay jitter variance of the main path node sequence are statistically analyzed. The historical transmission success frequency, average transmission delay, and delay jitter variance are input into a preset path evaluation function to calculate the current path evaluation value. Analyze the physical node overlap between the candidate path node sequence and the main path node sequence, and calculate the independent historical transmission success rate of the candidate path node sequence itself. Input the physical node overlap and the independent historical transmission success rate into a preset redundancy value function to calculate the path redundancy value.

[0010] As a further aspect of the present invention, the instruction logic dependency graph, the initial transmission weight, the current path evaluation value, and the path redundancy value are input into the path-instruction joint decision model to generate a transmission task scheduling scheme, including: The transmission task scheduling scheme includes: main transmission task decomposition, redundant transmission task decomposition, transmission path specification for each sub-task, and transmission timing planning. In the path-instruction joint decision-making model, the instruction logical dependency graph is cut according to the logical dependency edges and decomposed into multiple atomic instruction subtasks with logical order. Assign a subtask transmission weight proportionally from the initial transmission weight to each of the atomic instruction subtasks; Based on the subtask transmission weight, the logical order of the atomic instruction subtasks, the current path evaluation value, and the path redundancy value, path selection and task scheduling are solved in the path-instruction joint decision model. The task scheduling solution process is as follows: calculate the primary and backup path collaborative transmission strategy for subtasks with weights higher than the threshold in the atomic instruction subtasks, calculate the primary path transmission strategy for subtasks with weights lower than the threshold, and plan the transmission start time window of all subtasks according to the logical order of the atomic instruction subtasks, and finally output the transmission task scheduling scheme.

[0011] As a further aspect of the present invention, each of the atomic instruction subtasks is assigned a subtask transmission weight proportionally allocated from the initial transmission weight, including: Traverse the instruction logic dependency graph to identify the in-degree and out-degree of each atomic instruction subtask node; Calculate the centrality measure of a node in the overall instruction logic based on the in-degree and out-degree of each atomic instruction subtask node; The initial transmission weights are divided according to the proportion of the centrality metric of each atomic instruction subtask node in the sum of the centrality metrics of all nodes, and the divided weight values ​​are used as the subtask transmission weights of the corresponding atomic instruction subtasks.

[0012] As a further aspect of the present invention, a primary / backup path collaborative transmission strategy is calculated for subtasks with weights higher than a threshold in the atomic instruction subtasks, including: For atomic instruction subtasks with weights higher than the threshold, they are identified as key subtasks. Generate a primary path transport copy and at least one alternative path transport copy for the critical subtask, wherein the primary path transport copy is specified to be transported through the primary path node sequence, and the alternative path transport copy is specified to be transported through the alternative path node sequence with the highest evaluation value.

[0013] As a further aspect of the present invention, the system further includes: The message delivery execution module generates a transmission control message for a specified transmission path node according to the transmission task scheduling scheme, and sends the transmission control message to the corresponding transmission path node to perform redundant transmission of remote instructions. Based on the aforementioned transmission task scheduling scheme, a transmission control message is generated for the specified transmission path node, specifically including: The transmission task scheduling scheme is analyzed to extract the subtask description, transmission path indication, and timing requirements that need to be sent to specific transmission path nodes; The subtask description is encapsulated into a payload data unit; Add a message header containing the target node address, subtask sequence number, path identifier and timestamp to the payload data unit to form the original control message; Using a key shared with the target transmission path node, the original control message is encrypted and integrity-signed to generate the final securely delivered transmission control message.

[0014] As a further aspect of the present invention, the system further includes: The result arbitration module receives a transmission status feedback message from the transmission path node. The transmission status feedback message contains a subtask sequence number, a transmission result status code, and an actual arrival timestamp. Based on the subtask sequence number, match the corresponding original transmission task scheduling scheme entry from the local record; Based on the transmission result status code and the actual arrival timestamp, the execution result of the original transmission task scheduling scheme entry is verified and arbitrated. The arbitration process is as follows: when multiple feedbacks are received from the main path and alternative paths, a valid transmission result is selected based on the transmission result status code and the actual arrival timestamp, and duplicate or delayed transmission results are discarded. Based on the valid transmission result after arbitration, update the instruction history execution record related to the instruction receiver identifier and transmission path node sequence, specifically including: Extract the success identifier, actual transmission delay, and path identifier of the transmission path node sequence contained in the valid transmission result; In the instruction history execution record database, search for a historical record entry that matches the current instruction receiver identifier and the path identifier; Using the success identifier and the actual transmission delay, the historical transmission success frequency, average transmission delay and delay jitter variance statistics in the historical record entry are incrementally updated to obtain new statistical values. The new statistical value replaces the original statistical value in the historical record entry, thus completing the update of the instruction execution history record.

[0015] As a further aspect of the present invention, the construction steps of the path-instruction joint decision-making model include: Based on historical transmission task scheduling schemes and their corresponding transmission result feedback, a training sample dataset is constructed. The training samples include input feature vectors and expected output labels. The input feature vectors are composed of graph embedding vectors of instruction logic dependency graphs, normalized initial transmission weights, normalized current path evaluation values, and normalized path redundancy values. The expected output labels are the decomposition structure and scheduling time sequence of the corresponding historical transmission task scheduling scheme. A joint decision neural network model is constructed, comprising a graph convolutional layer, a fully connected layer, and a sequence output layer. The graph convolutional layer is used to extract features from the input instruction logic dependency graph. The fully connected layer is used to fuse the initial transmission weights, the current path evaluation value, and the path redundancy value. The sequence output layer is used to generate the decomposition and scheduling sequence of atomic instruction subtasks. Using the training sample dataset, the joint decision neural network model is iteratively trained using the gradient descent algorithm, and the model parameters are adjusted until the loss function value between the model output and the expected output label is lower than a preset threshold, thus completing the construction of the path-instruction joint decision model.

[0016] Compared with the prior art, the advantages and positive effects of the present invention are as follows: The semantic descriptions of remote control commands from the water conservancy dispatch center are processed using semantic structure parsing to form a command logical dependency graph. The semantic information of the commands is decomposed and corresponding logical relationships are constructed, clearly outlining the execution logic and dependencies within each command. A credibility-weighted mapping process is applied to command priority identifiers to form initial transmission weights. The configuration of command priorities corresponds to their actual credibility levels, and the values ​​of the transmission weights are consistent with the credibility attributes of the command priorities. The command logical dependency graph is then linked and bound to the initial transmission weights, establishing a corresponding adaptation relationship between the command's logical structure and the transmission weights. The weight allocation for command transmission aligns with the command's own semantic logic and priority credibility characteristics. The semantic parsing results and weight configuration results in the command processing stage are synchronously correlated, and the pre-configuration of command transmission remains consistent with the attribute characteristics of the command itself.

[0017] Based on the identifier of the instruction receiver, the main path node sequence and alternative path node sequence are determined from the pre-constructed water conservancy network topology, ensuring a precise correspondence between the selected path nodes and the instruction receiving target. Based on the instruction's historical execution records, the current path evaluation value of the main path node sequence and the path redundancy value of the alternative path node sequences are calculated. The operational status of the main path is reflected through quantitative values, while the redundancy effect of the alternative paths is characterized by specific value parameters. The instruction logical dependency graph, initial transmission weights, current path evaluation values, and path redundancy values ​​are input into the path-instruction joint decision-making model. Various parameters undergo collaborative computation within the model, and the transmission task scheduling scheme is generated based on the comprehensive parameters of the instruction and path. The decision-making basis for transmission scheduling is expanded from single path parameters to a combination of instruction attributes and path status parameters. The path scheduling for redundant transmission is adapted to the actual needs of instruction transmission. Instruction-related features and path operational status are integrated and matched in the scheduling decision, optimizing the targeting and fit of transmission scheduling. Attached Figure Description

[0018] Figure 1 This is a timing diagram of the redundant transmission system for remote commands of the water conservancy dispatch center as described in this invention. Figure 2 A flowchart for forming an instruction logic dependency graph; Figure 3 A flowchart for calculating the current path evaluation value and path redundancy value. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0020] In the description of this invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, in the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0021] See Figure 1 This invention provides a redundant transmission system for remote commands from a water conservancy dispatch center, the specific method of which includes: The instruction parsing module receives remote control instructions from the water conservancy dispatch center and parses them, extracting the instruction semantic description, instruction priority identifier, instruction receiver identifier, and historical execution records associated with the instruction or related instructions. These extracted elements together constitute the initial instruction dataset. The instruction processing weighting module performs deep processing on the initial instruction dataset. On the one hand, it performs semantic structure parsing on the instruction semantic descriptions in the dataset, transforming them into a formalized instruction logical dependency graph to characterize the internal execution logic relationships of the instructions. On the other hand, it performs credibility-weighted mapping on the instruction priority identifiers in the dataset, generating a quantified initial transmission weight, and associating this initial transmission weight with the aforementioned instruction logical dependency graph. The path calculation module locates the destination of the instruction in the pre-constructed water conservancy network topology based on the instruction receiver identifier, and determines the main path node sequence and several alternative path node sequences accordingly. This module further calls data from the instruction historical execution records to calculate the current path evaluation value of the main path node sequence and the path redundancy value of each alternative path node sequence. The decision generation module takes the instruction logic dependency graph, initial transmission weight, current path evaluation value and path redundancy value generated in the previous steps as input and feeds them into a pre-trained path-instruction joint decision model. This model performs comprehensive calculation and reasoning on the input information and outputs a detailed transmission task scheduling scheme, thereby completing the entire process from instruction reception to transmission planning.

[0022] In one embodiment of the present invention, the instruction processing weighting module performs semantic structure parsing on the instruction semantic description of the initial instruction dataset to form an instruction logical dependency graph. The process is as follows: [Refer to...] Figure 2The core components of the semantic description of the instruction are identified, including the instruction verb, the controlled object, the control parameters, and the constraints. These components are combined to form structured instruction element tuples. Based on a pre-set knowledge base of water conservancy control operation logic, the logical relationships of the identified instruction element tuples are deduced to determine the control dependencies, temporal dependencies, and data dependencies between the tuples. Using these instruction element tuples as nodes and the deduced control dependencies, temporal dependencies, and data dependencies as directed edges, an instruction logic dependency graph that can represent the internal execution logic of the instruction is constructed. The process of performing credibility-weighted mapping on instruction priority identifiers to form initial transmission weights involves querying a pre-defined instruction priority-weight mapping table, finding and obtaining the corresponding baseline transmission weight based on the instruction priority identifier, retrieving historical execution records of the same or similar remote control instructions from the instruction history execution records, and calculating the historical execution success rate and timeliness compliance rate from these records. Based on the retrieved execution success rate and timeliness compliance rate, the baseline transmission weight is dynamically corrected. This dynamic correction process involves weighting the execution success rate and timeliness compliance rate, using the result as a correction factor, and multiplying this correction factor by the baseline transmission weight to obtain the final initial transmission weight used for subsequent processing.

[0023] In practical implementation, semantic structured parsing begins with receiving remote control commands from the water conservancy dispatch center. For example, a remote control command might read, "Open the floodgate G1 of Reservoir A to a height of 2 meters within 30 minutes, and simultaneously monitor the data from the downstream water level sensor S1." The command parsing module extracts the command semantic description, command priority identifier, command receiver identifier, and command execution history of this remote control command. The command semantic description includes the natural language text "Open the floodgate G1 of Reservoir A to a height of 2 meters within 30 minutes, and simultaneously monitor the data from the downstream water level sensor S1." The command verbs identified in the command semantic description include "open" and "monitor," the controlled objects include "floodgate G1" and "water level sensor S1," the control parameters include "height of 2 meters," and the constraints include "within 30 minutes," generating a command element tuple. Based on a pre-defined knowledge base of water conservancy control operation logic, logical relationships are deduced from instruction element tuples. This knowledge base stores domain rules such as "opening the floodgate must take precedence over water level monitoring to ensure safety" and "water level monitoring data depends on the gate status." This determines the control dependency between instruction element tuples: the "open" instruction element tuple controls the "monitor" instruction element tuple; the temporal dependency: the "open" instruction element tuple must be executed before the "monitor" instruction element tuple; and the data dependency: the "monitor" instruction element tuple requires the result state data of the "open" instruction element tuple. Using instruction element tuples as nodes and control, temporal, and data dependencies as directed edges, an instruction logic dependency graph is constructed to represent the internal execution logic of the instructions. Nodes in the graph represent instruction element tuples, and directed edges represent the direction of dependencies.

[0024] In some embodiments, a reliability-weighted mapping process is performed on the instruction priority identifier, which may be of the "urgent" level. A preset instruction priority-weight mapping table is queried, containing entries, to obtain a baseline transmission weight of 0.9 corresponding to the "urgent" instruction priority identifier. The execution success rate and timeliness compliance rate of the same or similar remote control instructions are retrieved from the instruction history execution records. The historical transmission records show that the historical execution success rate for similar instructions targeting the same instruction receiver identifier is 95%, and the historical timeliness compliance rate is 90%. The baseline transmission weight is dynamically adjusted based on the execution success rate and timeliness compliance rate. The dynamic adjustment process is as follows: the weighted average of the execution success rate and timeliness compliance rate is used as the adjustment factor, setting the execution success rate weight to 0.6 and the timeliness compliance rate weight to 0.4. The adjustment factor is calculated as 0.95 * 0.6 + 0.90 * 0.4 = 0.93. The adjustment factor is multiplied by the baseline transmission weight to obtain an initial transmission weight of 0.9. 0.93 = 0.837. This is understandable; the data comparison reflects how different command priority markers and historical records lead to different initial transmission weights. For example, when another command priority marker is "normal," the baseline transmission weight is 0.6. Under the same historical execution success rate and timeliness compliance rate, the correction factor is still 0.93, and the initial transmission weight is 0.6. 0.93 = 0.558, thus reflecting the combined impact of priority and historical performance on the weight. Optionally, the formula for calculating the correction factor using the weighted average is expressed as:

[0025] in: Indicates the correction factor. Indicates the success rate of execution. Indicates the timeliness achievement rate. and These are the weighting coefficients for execution success rate and timeliness achievement rate, respectively, and they satisfy... Initial transmission weights Transmit weights from the baseline With correction factor Multiplying them together yields the result, i.e. In some embodiments, the execution success rate and timeliness compliance rate can be derived from the instruction history execution records by statistically analyzing the transmission results of several recent similar instructions. For example, by retrieving the past 10 remote control instructions for the same controlled object, the proportion of successfully transmitted instructions is calculated as the execution success rate, and the proportion of instructions that meet the timeliness requirements is calculated as the timeliness compliance rate. It is understood that the instruction priority-weight mapping table can be preset through system configuration, and the mapping relationship is set based on the needs of water conservancy scheduling operations. For example, emergency instructions correspond to a higher baseline transmission weight to ensure the allocation of transmission resources.

[0026] In one embodiment of the present invention, the process by which the path calculation module calculates the current path evaluation value of the main path node sequence and the path redundancy value of the alternative path node sequences based on the instruction history execution record is as follows: (See [link to relevant documentation]). Figure 3From the instruction execution history, all historical transmission records targeting the current instruction receiver are selected to form a historical transmission record set. For the main path node sequence, the historical transmission success frequency, average transmission delay, and delay jitter variance of the sequence in past transmissions are statistically calculated based on the historical transmission record set. These three statistics are input into a preset path evaluation function for calculation, and the output value of the function is the current path evaluation value of the main path node sequence. For the candidate path node sequence, the calculation of its path redundancy value requires analysis of two indicators: first, the physical node overlap between the candidate path node sequence and the main path node sequence; and second, the independent historical transmission success rate of the candidate path node sequence itself. The calculated physical node overlap and independent historical transmission success rate are input into a preset redundancy value function, and the output value of the function is the path redundancy value of the candidate path node sequence.

[0027] In practical implementation, the path calculation module extracts a set of historical transmission records for the instruction receiving end identifier from the instruction history execution record. The instruction receiving end identifier is "Gate Control Unit G-07". The preset main path node sequence in the water conservancy network topology is "Communication Relay Station R1 → Fiber Optic Node F5 → Field Control Cabinet C-07" and the preset alternative path node sequence is "Communication Relay Station R2 → Microwave Tower M3 → Field Control Cabinet C-07". The module retrieves all transmission records in the past 100 times with "Gate Control Unit G-07" as the instruction receiving end identifier from the instruction history execution record database, thus forming a set of historical transmission records. Based on the historical transmission record set, the historical transmission success frequency, average transmission delay, and delay jitter variance of the main path node sequence "Communication Relay Station R1 → Fiber Optic Node F5 → Field Control Cabinet C-07" are statistically analyzed. In 100 historical records, 95 transmissions were successful via this path, resulting in a historical transmission success frequency of 95. The average delay of all successful transmission instances is calculated to be 120 milliseconds. The variance of the delay of all successful transmission instances relative to the average is calculated to be 15 milliseconds². The historical transmission success frequency, average transmission delay, and delay jitter variance are input into a preset path evaluation function to calculate the current path evaluation value. The path evaluation function can be designed as a mathematical expression that comprehensively considers reliability, delay, and stability. In some embodiments, a specific calculation form of the path evaluation function is as follows:

[0028] in: This represents the current path evaluation value of the calculated main path node sequence. Indicates the frequency of successful historical transmissions. This represents the total number of times a transmission was attempted via this path in the historical transmission record set. Indicates the average transmission delay. This represents the variance of latency jitter. , , The preset positive weighting coefficients are used to balance the influence of different indicators. It can be understood that the data comparison reflects the fact that different paths have different historical statistical values, leading to different current path evaluation values. For example, the statistical values ​​of the main path node sequence are (95, 120, 15), while the candidate path node sequence "Communication Relay Station R2 → Microwave Tower M3 → Field Control Cabinet C-07" has a historical transmission success frequency of 80, an average transmission delay of 200 milliseconds, and a delay jitter variance of 50 milliseconds². Under the same weighting coefficients, the current path evaluation value of the main path node sequence will be significantly higher than that of the candidate path node sequence.

[0029] Optionally, the physical node overlap between the candidate path node sequence and the main path node sequence is analyzed. Physical node overlap is defined as the ratio of the number of identical physical nodes in the two path node sequences to the total number of nodes in the main path node sequence. The main path node sequence includes nodes "Communication Relay Station R1", "Fiber Optic Node F5", and "Field Control Cabinet C-07", while the candidate path node sequence includes nodes "Communication Relay Station R2", "Microwave Tower M3", and "Field Control Cabinet C-07". The two paths share only one node, "Field Control Cabinet C-07", resulting in a physical node overlap of 1 / 3. The independent historical transmission success rate of the candidate path node sequence itself is also calculated. The independent historical transmission success rate, without considering the main path, is the ratio of the number of successful historical transmissions to the total number of attempts for that candidate path node sequence. Based on the historical transmission record set, the independent historical transmission success rate of the candidate path node sequence "Communication Relay Station R2 → Microwave Tower M3 → Field Control Cabinet C-07" is 80%. The calculated physical node overlap and independent historical transmission success rate are input into a preset redundancy value function to calculate the path redundancy value. This redundancy value function can be designed to encourage paths with low overlap and high independent success rates. In some embodiments, a specific calculation form of the redundancy value function is:

[0030] in: This represents the path redundancy value of the calculated sequence of candidate path nodes. Indicates the degree of overlap between physical nodes. Indicates the success rate of independent historical transmissions. and These are preset weighting coefficients. It can be understood that the data comparison reflects the different path redundancy values ​​resulting from the differences in physical node overlap and independent historical transmission success rates among different alternative paths. For example, another alternative path node sequence, "Communication Relay Station R1 → Satellite Terminal S4 → Field Control Cabinet C-07," shares the two nodes "Communication Relay Station R1" and "Field Control Cabinet C-07" with the main path node sequence, with a physical node overlap of 2 / 3 and an independent historical transmission success rate of 70%. Under the same weighting coefficients, its path redundancy value will be lower than the aforementioned alternative path with an overlap of 1 / 3 and an independent success rate of 80%. The calculated current path evaluation value and path redundancy value will be passed as key input parameters to the subsequent decision generation module.

[0031] In one embodiment of the present invention, the decision generation module inputs the instruction logical dependency graph, initial transmission weight, current path evaluation value, and path redundancy value into the path-instruction joint decision model to generate a transmission task scheduling scheme. This scheme specifically includes main transmission task decomposition, redundant transmission task decomposition, assigning transmission paths to each subtask, and planning transmission timing. In the path-instruction joint decision model, the input instruction logical dependency graph is first cut according to its internal logical dependency edges, decomposing it into multiple atomic instruction subtasks with logical order. Then, each decomposed atomic instruction subtask is further divided... A subtask transmission weight is assigned, which is derived from the initial transmission weight according to a specific ratio. Then, based on the subtask transmission weight of each atomic instruction subtask, its logical order, and the evaluation information of the primary and backup paths, the model performs path selection and task scheduling solutions. The task scheduling solution process calculates the primary and backup path collaborative transmission strategy for subtasks with weights higher than a certain preset threshold, calculates the primary path transmission strategy for subtasks with weights lower than the threshold, and plans the transmission start time window for each subtask according to the logical order among all atomic instruction subtasks. Finally, the model outputs a complete transmission task scheduling scheme. The construction steps of the path-instruction joint decision-making model include: constructing a training sample dataset for model training based on historically accumulated transmission task scheduling schemes and their corresponding actual transmission result feedback; the training samples consist of input feature vectors and expected output labels; the input feature vectors are concatenated from the graph embedding vector of the instruction logical dependency graph, the normalized initial transmission weights, the normalized current path evaluation value, and the normalized path redundancy value; the expected output labels are the decomposition structure and scheduling time sequence of the corresponding historical transmission task scheduling schemes; the model structure is constructed as a joint decision-making neural network model containing graph convolutional layers, fully connected layers, and sequence output layers; the graph convolutional layers are responsible for extracting graph structure features from the input instruction logical dependency graph; the fully connected layers are used to fuse numerical features such as the initial transmission weights, the current path evaluation value, and the path redundancy value; and the sequence output layers are used to generate the decomposition sequence and scheduling time sequence of atomic instruction subtasks; using the constructed training sample dataset, the joint decision-making neural network model is iteratively trained using the gradient descent algorithm, continuously adjusting the internal parameters of the model until the loss function value between the model output and the expected output label is lower than a preset threshold, at which point the construction of the path-instruction joint decision-making model is completed.

[0032] In implementation, the decision generation module receives input data from the preceding module. The input instruction logic dependency graph contains three atomic instruction subtask nodes: node A (opens the floodgate G1 to a height of 1 meter), node B (wait 5 minutes), and node C (monitors water level sensor S1 data). There is a time-series dependency edge from node A to node B, and a data-series dependency edge from node B to node C. The initial transmission weight is 0.8, the current path evaluation value of the main path node sequence is 0.92, and the path redundancy values ​​of the two alternative path node sequences are 0.75 and 0.60, respectively. The internal processing flow of the path-instruction joint decision model is as follows: First, the instruction logic dependency graph is cut according to its internal logical dependency edges, and after cutting, it is decomposed into three atomic instruction subtasks with a logical order, namely A→B→C. A subtask transmission weight is assigned to each atomic instruction subtask. The allocation method is to divide the initial transmission weight of 0.8 according to a specific ratio. For example, one ratio is that node A receives 0.4, node B receives 0.2, and node C receives 0.2. Based on the subtask transmission weights, the logical order of atomic instruction subtasks, the current path evaluation value, and the path redundancy value, path selection and task scheduling are solved in the path-instruction joint decision model. A weight threshold is set during the task scheduling process, for example, a threshold of 0.3. The subtask transmission weights of atomic instruction subtasks are compared with the threshold. Atomic instruction subtask A, with a weight of 0.4, is higher than the threshold of 0.3, and is therefore identified as a critical subtask. A primary / backup path collaborative transmission strategy is calculated for critical subtask A. The weights of atomic instruction subtasks B and C are not higher than the threshold, so a primary path transmission strategy is calculated for them. According to the logical order of atomic instruction subtasks A, B, and C (A→B→C), the transmission start time windows for all subtasks are planned. For example, atomic instruction subtask A starts at time T0, atomic instruction subtask B starts at T0+Δt1, and atomic instruction subtask C starts at T0+Δt2. Finally, the model outputs a complete transmission task scheduling scheme. In some embodiments, the transmission task scheduling scheme is presented in the form of a structured data table, see Table 1: Table 1: Transmission Task Scheduling Scheme

[0033] It is understandable that data comparison shows that different inputs lead to different scheduling schemes. For example, if the subtask transmission weight of atomic instruction subtask A is adjusted to 0.25 (below the threshold of 0.3), the transmission strategy of atomic instruction subtask A will change from "primary-backup collaboration" to "primary path," and the redundant transmission part in the entire scheduling scheme will be reduced. The construction steps of the path-instruction joint decision model are based on historical data. Based on the historical transmission task scheduling schemes and their corresponding transmission result feedback, a training sample dataset is constructed. A training sample includes an input feature vector and a desired output label. The input feature vector is composed of the graph embedding vector of the instruction logical dependency graph, the normalized initial transmission weight, the normalized current path evaluation value, and the normalized path redundancy value. The desired output label is the decomposition structure and scheduling time sequence of the corresponding historical transmission task scheduling scheme. A joint decision neural network model is constructed, comprising graph convolutional layers, fully connected layers, and a sequence output layer. The graph convolutional layers extract graph structure features from the input instruction logic dependency graph, extracting features from nodes and edges. The fully connected layers fuse graph features from the graph convolutional layers with numerical features such as initial transmission weights, current path evaluation values, and path redundancy values. The sequence output layer generates the decomposition and scheduling sequence of atomic instruction subtasks, and can employ a recurrent neural network or a Transformer decoder structure. Using a training sample dataset, the joint decision neural network model is iteratively trained using the gradient descent algorithm to adjust the model parameters. The goal of model training is to minimize the difference between the predicted scheduling scheme output by the model and the desired output label. Optionally, the loss function used during model training can be designed as follows:

[0034] in: This represents the value of the loss function. This indicates the number of samples in a training batch. Indicates the first The expected output label vector for each training sample. The joint decision neural network model represents the first... The model outputs a predicted vector for each sample. Iterative training continues until the loss function value between the model's predictions and the expected output labels on the validation set is found to be correct. If the value is below a preset threshold, such as 0.01, the construction of the path-instruction joint decision-making model is complete. This data comparison demonstrates that path-instruction joint decision-making models trained using different historical datasets or different preset thresholds have different internal parameters. For the same input, they may produce slightly different transmission task scheduling schemes, reflecting the model's ability to learn scheduling strategies from data.

[0035] In one embodiment of the present invention, the specific process of assigning subtask transmission weights proportionally divided from the initial transmission weights to each atomic instruction subtask is as follows: traversing the instruction logic dependency graph, identifying the in-degree and out-degree of each atomic instruction subtask node in the graph, calculating the centrality metric of each node in the entire instruction logic dependency graph based on the in-degree and out-degree of each node, and then dividing the initial transmission weights according to the proportion of the centrality metric of each atomic instruction subtask node to the sum of the centrality metrics of all nodes, and using the weight values ​​obtained from the division as the subtask transmission weights of the corresponding atomic instruction subtasks. The process of calculating the primary and backup path collaborative transmission strategy for subtasks with weights higher than a threshold in the atomic instruction subtasks is as follows: firstly, atomic instruction subtasks with weights higher than a preset threshold are identified as key subtasks; then, a primary path transmission copy and at least one backup path transmission copy are generated for each key subtask, wherein the primary path transmission copy is specified to be transmitted through the primary path node sequence, and the backup path transmission copy is specified to be transmitted through the backup path node sequence with the highest path redundancy value assessment value.

[0036] In practical implementation, the process of assigning subtask transmission weights to atomic instruction subtasks begins with traversing the instruction logic dependency graph. The instruction logic dependency graph contains four atomic instruction subtask nodes: node A (instruction: start the water pump), node B (instruction: adjust the valve opening to 50%), node C (instruction: monitor pressure value), and node D (instruction: generate an operation report). The graph contains temporal dependency edges from node A to node B, data dependency edges from node B to node C, and control dependency edges from node C to node D. The in-degree and out-degree of each atomic instruction subtask node are identified: node A has an in-degree of 0 and an out-degree of 1; node B has an in-degree of 1 and an out-degree of 1; node C has an in-degree of 1 and an out-degree of 1; and node D has an in-degree of 1 and an out-degree of 0. Based on the in-degree and out-degree of each atomic instruction subtask node, the centrality metric of the node in the overall instruction logic is calculated. One method for calculating the centrality metric is to consider the activity level of node connections. In some embodiments, the centrality metric... The calculation formula is:

[0037] in: This represents the centrality measure of the atomic instruction subtask node v. This represents the in-degree of the atomic instruction subtask node v. This represents the out-degree of the atomic instruction subtask node v. Calculate the centrality metric of each node; the centrality metric of node A is... The centrality metric of node B is The centrality metric of node C is The centrality metric of node D is The sum of the centrality measures of all nodes is The initial transmission weights are divided according to the proportion of the centrality metric of each atomic instruction subtask node to the total centrality metric of all nodes. The initial transmission weight is 1.2, and node A is assigned a weight of [missing value]. Node B is assigned a weight of Node C is assigned a weight of The weight assigned to node D is The weights after segmentation are used as the subtask transmission weights for the corresponding atomic instruction subtasks. It's understandable that data comparisons show that different centrality metric calculation methods lead to different weight allocation ratios. For example, if the centrality metric only calculates the sum of in-degree and out-degree, then the centrality metric for nodes B and C is 2, and the centrality metric for nodes A and D is 1. The weight allocation result will be: node A receives 0.24, node B receives 0.48, node C receives 0.48, and node D receives 0.24. Refer to Table 2 for the allocation relationship between atomic instruction subtask nodes and subtask transmission weights: Table 2: Centrality Measurement and Weight Allocation of Atomic Instruction Subtask Nodes

[0038] In practical implementation, a primary and backup path collaborative transmission strategy is calculated for subtasks with weights higher than a threshold in the atomic instruction subtasks. A weight threshold is set, for example, 0.25. The transmission weights of the atomic instruction subtasks are compared with the threshold. Subtask B (weight 0.4) is higher than the threshold 0.25, as is subtask C (weight 0.4). Subtask A (weight 0.2) and subtask D (weight 0.2) are not higher than the threshold 0.25. Subtasks B and C are then identified as critical subtasks. A primary path transmission copy and at least one backup path transmission copy are generated for each critical subtask. For critical subtask B, one primary path transmission copy and one backup path transmission copy are generated. The primary path transmission copy is specified to be transmitted via the primary path node sequence "Communication Relay Station R1 -> Fiber Optic Node F2 -> Field Controller B". The backup path transmission copy is specified to be transmitted via the backup path node sequence "Communication Relay Station R3 -> Microwave Link M1 -> Field Controller B" with the highest path redundancy value assessment value. It is understandable that data comparison shows that different threshold settings will change the identification of key subtasks. For example, if the weight threshold is set to 0.35, the weights of atomic instruction subtasks B and C (0.4) are still higher than the threshold, and they are still identified as key subtasks. However, if the threshold is set to 0.45, the weights of atomic instruction subtasks B and C (0.4) are lower than the threshold, and atomic instruction subtasks B and C will not be identified as key subtasks. The system will not calculate the primary and backup path collaborative transmission strategy for atomic instruction subtasks B and C, but will only calculate a single primary path transmission strategy for them.

[0039] In one embodiment of the present invention, the message delivery execution module generates transmission control messages for designated transmission path nodes based on the transmission task scheduling scheme output by the decision generation module, and delivers these messages to the corresponding transmission path nodes to perform redundant transmission of remote commands. The specific process of generating transmission control messages is as follows: parsing the transmission task scheduling scheme, extracting the subtask description, transmission path indication, and timing requirements that need to be delivered to specific transmission path nodes, encapsulating the subtask description into a payload data unit, adding a message header to the payload data unit, the message header containing the target node address, subtask sequence number, path identifier, and timestamp information, thereby forming the original control message, and finally using a key pre-shared with the target transmission path node to encrypt and perform integrity signing on the original control message to generate the final securely deliverable transmission control message. The system also includes a result arbitration module. This module receives transmission status feedback messages from transmission path nodes. These messages contain subtask sequence numbers, transmission result status codes, and actual arrival timestamps. Based on the subtask sequence numbers in the feedback messages, it matches the corresponding original transmission task scheduling scheme entry from local records. Based on the transmission result status code and actual arrival timestamp, it verifies and arbitrates the execution result of that entry. The arbitration process involves receiving multiple feedbacks from the main path and alternative paths, selecting the best valid transmission result based on the transmission result status code and actual arrival timestamp of each feedback, and discarding duplicate or delayed transmission results. The final valid transmission result is then determined based on the arbitration result. The process involves updating the instruction history execution records associated with the instruction receiver identifier and the transmission path node sequence. This update process extracts the success identifier, actual transmission delay, and path identifier of the transmission path node sequence used from the valid transmission results. In the instruction history execution record database, it searches for historical records that match the current instruction receiver identifier and path identifier. Using the success identifier and actual transmission delay, it incrementally updates the historical transmission success frequency, average transmission delay, and delay jitter variance statistics stored in the historical record entry to obtain new statistical values. These new statistical values ​​then replace the original statistical values ​​in the historical record entry, thus completing the update of the instruction history execution records.

[0040] In practical implementation, the message delivery execution module generates a transmission control message for a specified transmission path node based on the transmission task scheduling scheme. The transmission task scheduling scheme includes a subtask described as "setting the valve V101 opening to 30%", the transmission path indicator as "main path PA", and the timing requirement as "starting within 10 milliseconds after timestamp T=1620000000". The module parses the transmission task scheduling scheme to extract this information and encapsulates the subtask description "setting the valve V101 opening to 30%" into a payload data unit. The payload data unit can use binary or a specific encoding format. A message header is added to the payload data unit, containing the target node address. The original control message is formed by combining the subtask sequence number "TSK_2023052_001", the path identifier "PA", and the timestamp "1620000000100". The original control message format includes header and payload fields. Using the key K_node7a shared with the target transmission path node Node_7A, the original control message is encrypted and signed for integrity. The encryption algorithm can be AES, and the signing algorithm can be HMAC-SHA256. This generates a final, securely deliverable transmission control message, which is then sent to the corresponding transmission path node Node_7A to perform redundant transmission of remote commands. In some embodiments, the encryption and signing processes can be combined into applying an authentication encryption algorithm, such as AES-GCM mode, to the original control message to generate a transmission control message containing ciphertext and an authentication tag. It is understandable that the data comparison is reflected in the use of different shared keys at different transmission path nodes. For example, the transmission control message sent to node Node_8B needs to be encrypted and signed using key K_node8b, which ensures the target and security of the message.

[0041] The arbitration module receives transmission status feedback messages from transmission path nodes. One message from the primary path node Node_7A contains the subtask sequence number "TSK_2023052_001", transmission result status code "200OK", and actual arrival timestamp "1620000000200". Another message from the alternative path node Node_9C contains the same subtask sequence number and the same transmission result status code. Based on the subtask sequence number "TSK_2023052_001", the module matches the corresponding original transmission task scheduling scheme entry in the local records. The matched entry shows that the subtask "TSK_2023052_001" is scheduled to be transmitted via the primary path PA and the alternative path PB. Based on the transmission result status code and the actual arrival timestamp, the execution results of the original transmission task scheduling scheme entries are verified and arbitrated. The arbitration process is that when multiple feedbacks are received from the main path and alternative paths, a valid transmission result is selected based on the transmission result status code and the actual arrival timestamp. The transmission result status code "200 OK" indicates success, and the actual arrival timestamp "1620000000200" is earlier than "1620000000250". The selection strategy is defined as selecting the first successful feedback to arrive. Therefore, the feedback from the main path node Node_7A is determined to be a valid transmission result, and duplicate or delayed transmission results from the alternative path node Node_9C are discarded. Optionally, in another scenario, the result arbitration module may receive two feedbacks: a status code "200OK" (timestamp T1) and a status code "500Error" (timestamp T2). Based on the transmission result status code, the feedback with the status code "200OK" will be selected as the valid transmission result, while the feedback with the status code "500Error" will be ignored, regardless of whether its timestamp is earlier or later.

[0042] Based on the valid transmission result after arbitration, update the instruction history execution records related to the instruction receiver identifier and transmission path node sequence. Extract the success identifier "Success", actual transmission delay "100 milliseconds", and path identifier "PA" from the valid transmission result. In the instruction history execution record database, search for historical records that match the current instruction receiver identifier "valve controller VC_101" and path identifier "PA". The statistical items stored in the found historical records include historical transmission success frequency "45 times", average transmission delay "105 milliseconds", and delay jitter variance "20 milliseconds²". Using the success identifier "Success" and actual transmission delay "100 milliseconds", perform incremental update calculations on the historical transmission success frequency, average transmission delay, and delay jitter variance statistical items in the historical records. Update the historical transmission success frequency to 45 + 1 = 46 times, and update the average transmission delay to (105) / (10 ... (45+100) / 46≈104.78 milliseconds. The delay jitter variance update requires calculating the new delay sequence variance. In some embodiments, a simplified calculation formula for the incremental update of delay jitter variance is:

[0043] in: This represents the updated latency jitter variance. This indicates the updated historical transmission success frequency. This represents the variance of latency jitter before the update. This indicates the actual transmission delay. This represents the average transmission latency before the update. This represents the updated average transmission latency. The new statistical value replaces the original statistical value in the historical record entry, completing the update of the instruction execution history. It's understandable that data comparisons show that different transmission results lead to different update calculations. If a successful transmission result is marked as "failure," the frequency of successful historical transmissions does not increase, and the average transmission latency and latency jitter variance may not be updated or may be updated using different algorithms that include failure latency, thus affecting the calculation of the current path evaluation value for that path.

[0044] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A redundant transmission system for remote commands of a water conservancy dispatch center, characterized in that, The system includes: The instruction parsing module receives remote control instructions from the water conservancy dispatch center and extracts the instruction semantic description, instruction priority identifier, instruction receiver identifier, and instruction historical execution record of the remote control instructions to generate an initial instruction dataset. The instruction processing weighting module performs semantic structure parsing on the semantic description of the instruction in the initial instruction dataset to form an instruction logical dependency graph, performs credibility weighting mapping on the instruction priority identifier to form an initial transmission weight, and associates and binds the instruction logical dependency graph with the initial transmission weight. The path calculation module determines the main path node sequence and the alternative path node sequence from the pre-constructed water conservancy network topology according to the instruction receiving end identifier, and calculates the current path evaluation value of the main path node sequence and the path redundancy value of the alternative path node sequence based on the instruction historical execution record. The decision generation module inputs the instruction logic dependency graph, the initial transmission weight, the current path evaluation value, and the path redundancy value into the path-instruction joint decision model to generate a transmission task scheduling scheme.

2. The redundant transmission system for remote commands of a water conservancy dispatch center according to claim 1, characterized in that, The semantic semantic descriptions of the initial instruction dataset are subjected to semantic structure parsing to form an instruction logical dependency graph, including: Identify the instruction verbs, controlled objects, control parameters, and constraints in the instruction semantic description, and generate instruction element tuples; Based on a pre-set knowledge base of water conservancy control operation logic, logical relationships are deduced from the instruction element tuples to determine the control dependency, temporal dependency, and data dependency relationships among the instruction element tuples. Using the instruction element tuple as nodes and the control dependency, timing dependency, and data dependency as directed edges, an instruction logic dependency graph representing the internal execution logic of an instruction is constructed.

3. The redundant transmission system for remote commands of the water conservancy dispatch center according to claim 1, characterized in that, The instruction priority identifier is subjected to a trustworthiness-weighted mapping process to form an initial transmission weight, including: Query the preset instruction priority-weight mapping table to obtain the baseline transmission weight corresponding to the instruction priority identifier; The success rate and timeliness of executing the same or similar remote control commands are retrieved from the command history execution records. Based on the execution success rate and the timeliness achievement rate, the baseline transmission weight is dynamically adjusted. The dynamic adjustment process is as follows: the weighted average of the execution success rate and the timeliness achievement rate is used as a adjustment factor, and the adjustment factor is multiplied by the baseline transmission weight to obtain the initial transmission weight.

4. The redundant transmission system for remote commands of the water conservancy dispatch center according to claim 1, characterized in that, Based on the instruction execution history, the current path evaluation value of the primary path node sequence and the path redundancy value of the alternative path node sequences are calculated, including: Extract a set of historical transmission records for the instruction receiver identifier from the instruction execution history records; Based on the historical transmission record set, the historical transmission success frequency, average transmission delay, and delay jitter variance of the main path node sequence are statistically analyzed. The historical transmission success frequency, average transmission delay, and delay jitter variance are input into a preset path evaluation function to calculate the current path evaluation value. Analyze the physical node overlap between the candidate path node sequence and the main path node sequence, and calculate the independent historical transmission success rate of the candidate path node sequence itself. Input the physical node overlap and the independent historical transmission success rate into a preset redundancy value function to calculate the path redundancy value.

5. The redundant transmission system for remote commands of a water conservancy dispatch center according to claim 1, characterized in that, The instruction logical dependency graph, the initial transmission weight, the current path evaluation value, and the path redundancy value are input into the path-instruction joint decision model to generate a transmission task scheduling scheme, including: The transmission task scheduling scheme includes: main transmission task decomposition, redundant transmission task decomposition, transmission path specification for each sub-task, and transmission timing planning. In the path-instruction joint decision-making model, the instruction logical dependency graph is cut according to the logical dependency edges and decomposed into multiple atomic instruction subtasks with logical order. Assign a subtask transmission weight proportionally from the initial transmission weight to each of the atomic instruction subtasks; Based on the subtask transmission weight, the logical order of the atomic instruction subtasks, the current path evaluation value, and the path redundancy value, path selection and task scheduling are solved in the path-instruction joint decision model. The task scheduling solution process is as follows: calculate the primary and backup path collaborative transmission strategy for subtasks with weights higher than the threshold in the atomic instruction subtasks, calculate the primary path transmission strategy for subtasks with weights lower than the threshold, and plan the transmission start time window of all subtasks according to the logical order of the atomic instruction subtasks, and finally output the transmission task scheduling scheme.

6. The redundant transmission system for remote commands of a water conservancy dispatch center according to claim 5, characterized in that, Assigning a subtask transmission weight proportionally from the initial transmission weight to each of the atomic instruction subtasks, including: Traverse the instruction logic dependency graph to identify the in-degree and out-degree of each atomic instruction subtask node; Calculate the centrality measure of a node in the overall instruction logic based on the in-degree and out-degree of each atomic instruction subtask node; The initial transmission weights are divided according to the proportion of the centrality metric of each atomic instruction subtask node in the sum of the centrality metrics of all nodes, and the divided weight values ​​are used as the subtask transmission weights of the corresponding atomic instruction subtasks.

7. The redundant transmission system for remote commands of a water conservancy dispatch center according to claim 5, characterized in that, Calculate a primary / backup path cooperative transmission strategy for subtasks with weights higher than a threshold in the atomic instruction subtasks, including: For atomic instruction subtasks with weights higher than the threshold, they are identified as key subtasks. Generate a primary path transport copy and at least one alternative path transport copy for the critical subtask, wherein the primary path transport copy is specified to be transported through the primary path node sequence, and the alternative path transport copy is specified to be transported through the alternative path node sequence with the highest evaluation value.

8. The redundant transmission system for remote commands of a water conservancy dispatch center according to claim 1, characterized in that, The system also includes: The message delivery execution module generates a transmission control message for a specified transmission path node according to the transmission task scheduling scheme, and sends the transmission control message to the corresponding transmission path node to perform redundant transmission of remote instructions. Based on the aforementioned transmission task scheduling scheme, a transmission control message is generated for the specified transmission path node, specifically including: The transmission task scheduling scheme is analyzed to extract the subtask description, transmission path indication, and timing requirements that need to be sent to specific transmission path nodes; The subtask description is encapsulated into a payload data unit; Add a message header containing the target node address, subtask sequence number, path identifier and timestamp to the payload data unit to form the original control message; Using a key shared with the target transmission path node, the original control message is encrypted and integrity-signed to generate the final securely delivered transmission control message.

9. The redundant transmission system for remote commands of a water conservancy dispatch center according to claim 1, characterized in that, The system also includes: The result arbitration module receives a transmission status feedback message from the transmission path node. The transmission status feedback message contains a subtask sequence number, a transmission result status code, and an actual arrival timestamp. Based on the subtask sequence number, match the corresponding original transmission task scheduling scheme entry from the local record; Based on the transmission result status code and the actual arrival timestamp, the execution result of the original transmission task scheduling scheme entry is verified and arbitrated. The arbitration process is as follows: when multiple feedbacks are received from the main path and alternative paths, a valid transmission result is selected based on the transmission result status code and the actual arrival timestamp, and duplicate or delayed transmission results are discarded. Based on the valid transmission result after arbitration, update the instruction history execution record related to the instruction receiver identifier and transmission path node sequence, specifically including: Extract the success identifier, actual transmission delay, and path identifier of the transmission path node sequence contained in the valid transmission result; In the instruction history execution record database, search for a historical record entry that matches the current instruction receiver identifier and the path identifier; Using the success identifier and the actual transmission delay, the historical transmission success frequency, average transmission delay and delay jitter variance statistics in the historical record entry are incrementally updated to obtain new statistical values. The new statistical value replaces the original statistical value in the historical record entry, thus completing the update of the instruction execution history record.

10. The redundant transmission system for remote commands of a water conservancy dispatch center as described in claim 1, characterized in that, The construction steps of the path-instruction joint decision-making model include: Based on historical transmission task scheduling schemes and their corresponding transmission result feedback, a training sample dataset is constructed. The training samples include input feature vectors and expected output labels. The input feature vectors are composed of graph embedding vectors of instruction logic dependency graphs, normalized initial transmission weights, normalized current path evaluation values, and normalized path redundancy values. The expected output labels are the decomposition structure and scheduling time sequence of the corresponding historical transmission task scheduling scheme. A joint decision neural network model is constructed, comprising a graph convolutional layer, a fully connected layer, and a sequence output layer. The graph convolutional layer is used to extract features from the input instruction logic dependency graph. The fully connected layer is used to fuse the initial transmission weights, the current path evaluation value, and the path redundancy value. The sequence output layer is used to generate the decomposition and scheduling sequence of atomic instruction subtasks. Using the training sample dataset, the joint decision neural network model is iteratively trained using the gradient descent algorithm, and the model parameters are adjusted until the loss function value between the model output and the expected output label is lower than a preset threshold, thus completing the construction of the path-instruction joint decision model.

Images