A data decision method for smart park operation situation convergence

Abstract

This invention discloses a data-driven decision-making method for the convergence of operational status in smart parks, specifically relating to the field of data decision-making technology. The method includes collecting multi-source operational data from the smart park within the same operational cycle, performing time alignment and standardization processing on the multi-source operational data, and constructing a status node table in a digital twin, mapping the current operational status of the park to a set of status nodes containing status identifiers and resource occupancy fields. By explicitly constructing future operational paths under the assumption of parallel operation goals in the digital twin, and jointly modeling and layer-by-layer pruning resource occupancy conflicts, state transition conflicts, and their evolutionary uncertainties among future operational paths, the park's operational decision-making shifts from relying solely on the stability judgment of the current state to constraining and converging competitive future paths.

Description

Technical Field

[0001] This invention relates to the field of data decision-making technology, and more specifically, to a data-driven decision-making method for the convergence of operational status in smart parks. Background Technology

[0002] In current smart park operation and management technologies, operational status assessment is usually based on the stability judgment of real-time or short-term operational status. Through comprehensive calculation of multi-dimensional indicators such as energy consumption level, equipment load, safety indicators, and traffic efficiency, it is confirmed whether the park is in a controllable state within the current time window. Under this technological paradigm, park operation decisions often implicitly contain multiple parallel but unmodeled future assumptions. For example, different management cycles may focus on energy-saving optimization, operational safety assurance, or service efficiency improvement. These assumptions can coexist in the short term through local scheduling or parameter fine-tuning, so that the existing situation assessment model can continuously output stable operational judgment results. However, as the operation time goes on, the medium and long-term resource allocation paths, equipment evolution directions and management strategy objectives corresponding to various decision assumptions gradually develop irreconcilable differences. The park is actually being pulled by multiple mutually exclusive future evolution paths at the same time. Due to the lack of ability to characterize the number of future paths and their exclusive relationships, the existing technology still maintains the existing decision logic based on the stability of the current state. Ultimately, this causes the park to lose its ability to support multiple long-term goals at the same time during critical stages, resulting in the situation continuing to diverge at the macro level but failing to be identified in time. It is evident that existing smart park situation assessment and decision-making methods are unable to identify and constrain competing future paths that coexist, causing situation judgments to remain at a short-term stability level and failing to achieve true situation convergence for medium- and long-term operational goals. Summary of the Invention

[0003] To overcome the aforementioned deficiencies of the prior art, embodiments of the present invention provide a data-driven decision-making method for the convergence of smart park operation status. By explicitly constructing future operation paths under the assumption of parallel operation goals in a digital twin, and jointly modeling and layer-by-layer pruning resource occupation conflicts, state transition conflicts, and their evolutionary uncertainties among future operation paths, the park operation decision-making is transformed from relying solely on the stability judgment of the current state to constraining and converging competitive future paths.

[0004] To achieve the above objectives, the present invention provides the following technical solution: a data-driven decision-making method for the convergence of operational status in smart parks, comprising: S1. Collect multi-source operation data of the smart park within the same operation cycle, perform time alignment and standardization processing on the multi-source operation data, and build a status node table in the digital twin to map the current operation status of the park into a set of status nodes containing status identifiers and resource occupancy fields. S2. Based on the set of situation nodes, construct a future path graph in the digital twin. Perform state recursion calculation for each situation node according to different operational goal assumptions. Record the recursed state transition relationship and resource change relationship as directed edges to form a path graph structure containing multiple future operational paths. S3. Perform traversal calculations on the path graph structure according to the time hierarchy, identify resource occupation conflicts and state transition conflicts of different future running paths at the same level, and write the conflict results into the path exclusivity matrix to explicitly characterize the incompatible relationship between future running paths. S4. Based on the path exclusivity matrix, perform layer-by-layer pruning calculations on the path graph structure. Under the premise of satisfying the exclusivity constraint, calculate the compression effect of each future running path on the uncertainty of the subsequent situation, and remove the paths with decreasing compression effect layer by layer to form a situation convergence path subgraph. S5. Write the situation convergence path subgraph into the digital twin as an operational decision constraint, perform consistency solution on the reachable control actions in the situation convergence path subgraph, generate park operation control instructions and output them to the actual park for execution, and write back the execution results to update the situation node table.

[0005] In a preferred embodiment, S1 includes: S1-1. Within the same operating cycle, collect operating data from different business subdomains of the smart park, and attach corresponding collection time identifiers and data source identifiers to various types of operating data to form a set of raw operating data with time attributes and source attributes. S1-2. Based on the time identifier, perform time alignment processing on the original set of running data to map running data with different collection frequencies and different time granularities onto a unified time axis to obtain a time-consistent running data sequence. S1-3. Based on the data source identifier, perform unified processing on time-consistent operational data sequences, converting the measurement methods, statistical rules and expression forms under different business subdomains into a unified data expression standard to form a standardized operational dataset. S1-4. Input the operational data from the standardized operational dataset into the digital twin one by one in the order of data records; For each piece of input running data, read the data source identifier and business subdomain identifier carried in the corresponding running data, and determine the specific running object corresponding to the corresponding running data based on the data source identifier and business subdomain identifier; After determining the running object, read the unique object identifier and the current running cycle identifier of the running object, combine the unique object identifier and the running cycle identifier to generate the corresponding status identifier, and write the status identifier into the status identifier field corresponding to the running object; Resource usage data corresponding to the running object is selected from the standardized running dataset. The resource usage data is accumulated and calculated within the running cycle to obtain the resource usage value, and the resource usage value is written into the resource usage field corresponding to the running object. After completing the writing of the status identifier and resource usage value of the running object, the data record containing the status identifier field and resource usage field is written to the status node table. After the situation node table is constructed, situation nodes are generated one by one according to the data records in the situation node table, and the generated situation nodes are collected to form a situation node set, which is used to represent the current operating status of the park.

[0006] In a preferred embodiment, S2 includes: S2-1. Read the situation nodes one by one in the situation node set according to the node order. For each situation node, read the corresponding status identifier and resource occupation field, and write the status identifier, resource occupation field and initial simulation step number into a path seed record. Write the generated path seed records into the path seed queue in sequence. S2-2. Take out the path seed records one by one from the path seed queue in the queue order, and take out the running target hypothesis one by one from the running target hypothesis table in the hypothesis order. Combine the current path seed record with at least one running target hypothesis to generate a binding pair; write the binding pairs sequentially into the recursive task queue to output the recursive task queue; S2-3. Take out the recursive tasks one by one from the queue in the order of the queue, and read the status flag, resource occupation field and deduction step number of the corresponding binding pair of the recursive task. Based on the status identifier, determine the current state instance of the corresponding situation node in the digital twin, and read the state update rule corresponding to the running target assumption; According to the status update rules, status update calculations are performed on each status field corresponding to the status identifier. The resource occupancy field is used as the constraint input. The fields that are allowed to change are updated, while the fields that are not allowed to change remain unchanged. The updated state fields are combined to generate the next situation node, and the state identifier corresponding to the next situation node is written into the new path seed record to form the successor path seed record. At the same time, the deduction step number of the successor path seed record is updated to the next deduction step number.

[0007] In a preferred embodiment, S2 further includes: S2-4. When generating the subsequent path seed record, read the resource occupation field in the previous path seed record and the resource occupation field in the subsequent path seed record and perform difference calculation to obtain the resource change result. Read the state flags before and after recursion and perform a difference determination to obtain the state flag change results; Write the resource change results and the state identifier change results into the directed edge record, and take the previous state node as the starting point of the directed edge record and the next state node as the ending point of the directed edge record. S2-5. Perform gating judgment on each of the generated directed edge records, including reading the resource change results in the directed edge records and splitting the resource change results into resource change components according to the resource type. For each resource change component, read the incompatible conditions of the corresponding resource type in the exclusive constraint table one by one, and perform item-by-item comparison between the resource change component and the constraint threshold or conflict rule in the incompatible conditions; When comparing a certain resource change component, the predefined resource change determination method in the corresponding incompatibility condition is read; Among them, when the incompatibility condition is defined as a threshold condition, it is determined whether the change magnitude of the resource change component exceeds the corresponding constraint threshold. When the incompatibility condition is defined as a directional condition, determine whether the direction of change of the resource change component is consistent with the prohibited change direction specified in the incompatibility condition; When the incompatibility condition is defined as a combination condition, determine whether the combination result of the resource change component and other resource change components in the same directed edge record satisfies the conflict combination rules defined in the incompatibility condition. When any decision result is true, the directed edge record corresponding to the decision triggers an incompatible condition; When an incompatible condition is triggered, the corresponding directed edge record is blocked from being written into the future path graph and the corresponding successor path seed record is deleted. If no incompatible condition is triggered, the corresponding directed edge record is written to the future path graph, and the corresponding successor path seed record is written back to the recursive task queue. S2-6. Repeat S2-3 to S2-5 until the recursive task queue is empty or the deduction step number corresponding to the recursive task reaches the preset step size; traverse the future path graph, extract and output the path segments formed by connecting the beginning and end of continuous directed edges as future running paths, thereby forming a path graph structure containing multiple future running paths, and write a future running path identifier for each future running path.

[0008] In a preferred embodiment, S3 includes: S3-1. Read the directed edge records one by one from the path graph structure in the order of generation, extract the deduction step number corresponding to each directed edge record, and append the directed edge records with the same deduction step number to the same time level queue to form multiple time level queues distinguished by deduction step number. S3-2. For each time-level queue, retrieve the directed edge records one by one in the queue order, read the resource occupation field of the corresponding directed edge record endpoint state node, and split the corresponding resource occupation field into resource occupation components according to resource type. Write the resource usage component into the resource usage index table with the future running path identifier and resource type as the key, so that the resource usage index table forms a resource usage snapshot for each future running path in the corresponding time level; S3-3. Within the same time level, read the resource occupancy components in the resource occupancy index table one by one according to the resource type, and perform pairwise superposition calculations on the resource occupancy components belonging to different future running paths under the same resource type. When any superposition result exceeds the resource constraint threshold corresponding to the resource type, the future running path pair that triggers the threshold will be recorded as a resource conflict pair and written into the resource conflict registration table.

[0009] In a preferred embodiment, S3 further includes: S3-4. For each time level queue, take out the directed edge record one by one in the queue order, read the state identifier change result of the corresponding directed edge record, and write the state identifier change result into the state transition index table with the future running path identifier as the key, so that the state transition index table forms a state transition snapshot for each future running path in the corresponding time level. S3-5. Within the same time level, read the corresponding status identifier change results in the status transition index table one pair at a time according to the future running path. For each pair of future running paths, read the status identifier change results of the two future running paths within the corresponding time level, and compare the two sets of status identifier change results item by item according to the status field. When performing a step-by-step comparison of the status changes of two future running paths, for the same status field, the status change values ​​of the two future running paths in the corresponding status field are read respectively. When two future running paths have corresponding state change values ​​in the same state field that point to mutually exclusive set of state values, it is determined that there is a state value conflict in the corresponding state field. Alternatively, when the state change values ​​of two future running paths on the same state field point to different state transition directions, and the state transition directions are restricted in the preset state transition constraints to not being able to be met at the same time, it is determined that there is a state transition conflict in the corresponding state field. When a conflict in state value or state transition is determined to exist in any state field, it is determined that the corresponding future running path pair has a state transition conflict at the corresponding time level. When a state transition conflict is determined to exist, the corresponding future running path pair is recorded as a state conflict pair and written into the state conflict registration table. S3-6. Perform a merge and write process on the resource conflict registration table and the status conflict registration table according to the future running path pair. Write the resource conflict pair or status conflict pair of the future running path pair into the path exclusivity matrix, and write an incompatible mark in the path exclusivity matrix at the position corresponding to the future running path pair. S3-7. Repeat S3-2 to S3-6 for all time-level queues until all time-level queues have been processed, thereby outputting the filled path exclusivity matrix.

[0010] In a preferred embodiment, S4 includes: S4-1. Read the incompatible markers cell by cell according to the row and column indices of the path exclusivity matrix, extract the corresponding future running path pair for each incompatible marker, and write the corresponding future running path pair into the exclusivity registration table. S4-2. Extract future running paths one by one in the path diagram structure according to the future running path identifier. For each future running path, read the corresponding situation nodes layer by layer according to the simulation step number from small to large. Extract the status identifier and resource occupation field of each level of situation node, and write the status identifier and resource occupation field into the path sequence record corresponding to the future running path in hierarchical order. S4-3. Perform uncertainty decomposition calculation on the path sequence record corresponding to each future running path in hierarchical order, wherein the state identifier of the corresponding level is read at the current level and the difference count is performed with the state identifier of the previous level to obtain the state transition count. Simultaneously, the resource usage field of the corresponding level is read and the difference is calculated with the resource usage field of the previous level to obtain the resource fluctuation amount; Write the state transition count and resource fluctuation amount into the hierarchical uncertainty record.

[0011] In a preferred embodiment, S4 further includes: S4-4. For each future running path, perform compression calculations on the uncertainty records of each level in hierarchical order. At each level, read the uncertainty record of the current level and the uncertainty record of the previous level, and perform a subtraction operation on the two to obtain the level compression amount. Write the compression amounts of each level into the compression trajectory record in sequence. S4-5. Perform a descent judgment on the compression trajectory record layer by layer. In the current layer, read the compression amount of the corresponding future running path and compare it with the compression amount of the previous layer. When the compression amount of the current layer is less than the compression amount of the previous layer, write the corresponding future running path into the elimination candidate table and record the corresponding triggering layer. S4-6. Perform gating pruning based on the exclusive pair registration table and the elimination candidate table. Specifically, perform pruning operations on each future running path in the elimination candidate table: delete the situation node of the corresponding future running path at the trigger level and subsequent levels from the path graph structure, and delete the directed edge record connected to the deleted situation node; at the same time, when a future running path and another future running path in the exclusive pair registration table form a future running path pair, prevent the other future running path from being retained at the same trigger level and subsequent levels. S4-7. After completing the gating pruning, traverse the updated path graph structure to extract the future running paths that are still continuously reachable, and collect the future running paths as a situation convergence path subgraph to represent the set of future running paths that still maintain uncertainty compression after layer-by-layer pruning under exclusive constraints.

[0012] In a preferred embodiment, S5 includes: S5-1. Read the situation convergence path subgraph one by one in the order of directed edge records, and write the starting situation node identifier, ending situation node identifier and corresponding deduction step number of each directed edge record into the constraint edge ledger of the digital twin to form a reachability constraint relationship layered by deduction step number. S5-2. Generate reachable node index layer by layer based on constraint edge ledger. For each simulation step number, read the endpoint situation node identifier corresponding to the simulation step number one by one and write it into the reachable node index table after deduplication to obtain the set of situation nodes that are allowed to be reached under the corresponding simulation step number. S5-3. Read the executable control actions in the park one by one according to the action identifier, and generate an action trial calculation task for each control action. In this process, read the status identifier and resource occupation field associated with the corresponding control action in the current situation node table, and write the corresponding control action, status identifier and resource occupation field into the action trial calculation task record. S5-4. Perform action trial calculations for each action trial task record. For each action trial task record, call the state transition rule corresponding to the corresponding control action to perform an update calculation on the state identifier to generate a trial state identifier, and call the resource impact rule corresponding to the corresponding control action to perform an update calculation on the resource occupancy field to generate a trial resource occupancy field, thereby forming a trial result node.

[0013] In a preferred embodiment, S5 further includes: S5-5. Perform dual-gating judgment on each trial result node. First, read the simulation step number corresponding to the trial result node, and then compare the trial status identifier with the status node identifier corresponding to the simulation step number in the reachable node index table. If the trial status flag exists in the reachable node index table corresponding to the simulation step number, the reachability gate is passed; otherwise, the reachability gate is not passed. Under the premise of reaching the gate, read the trial resource usage field and split the trial resource usage field into resource usage components according to the resource type; Each resource occupancy component is compared with the preset resource constraint threshold of the corresponding resource type. When all resource occupancy components meet the corresponding resource constraint threshold, the resource gating is deemed to pass; otherwise, the resource gating is deemed to fail. The corresponding control action is written to the action table only when both reachability gating and resource gating are used simultaneously. S5-6. Perform consistency solution on the action table according to the simulation step number, wherein the control action record under the corresponding simulation step number is read one by one in the action table; Based on the situation convergence path subgraph, the situation nodes reachable by the corresponding future operation path under the corresponding simulation step number are read one by one according to the future operation path, and the control actions determined by dual gating at the situation nodes are selected to form the set of actionable actions corresponding to the future operation path. Under the same simulation step number, the intersection calculation is performed on the sets of possible actions corresponding to different future running paths one by one to obtain the set of consensus actions that can be held under all future running path constraints; When the consensus action set is empty, roll back to the previous derivation step number and re-execute the intersection calculation until a non-empty consensus action set is obtained; S5-7. Generate park operation control instructions one by one based on the consensus action set and output them to the actual park for execution. After execution, collect the execution result data and write it back to the status node table one by one. For each execution result data, read the unique object identifier of the corresponding running object and locate the corresponding status node record. Update the status identifier and resource occupation field of the status node record with the execution result data, thereby completing the synchronous update of the current operation status of the park.

[0014] The technical effects and advantages of this invention are as follows: This solution explicitly constructs a future operational path map under multiple operational objectives in a digital twin and further characterizes the resource and state exclusive relationships between paths. This allows the originally implicit and parallel competitive future evolution paths to be fully modeled and constrained, thereby breaking through the limitations of existing technologies that only judge the situation based on short-term stability and achieving real situation convergence for medium- and long-term operational objectives. By introducing a future path graph structure based on state recursion and resource change records, the park's operational status is expanded from static nodes to an evolvable set of paths. This enables the state transition and resource consumption processes under different operational goal assumptions to have a calculable and comparable structured expression, thereby enhancing the park's forward-looking analysis capabilities. By using a path traversal mechanism that is layered according to the deduction step number, resource occupation conflicts and state transition conflicts between future running paths are identified respectively, and the conflict results are uniformly written into the path exclusivity matrix. This mechanism realizes an explicit characterization of the incompatible relationship between future paths, avoiding the formation of implicit hedging at the resource and state levels by different long-term strategies. By combining the path exclusivity relationship with the uncertainty evolution characteristics of the path itself, a layer-by-layer pruning mechanism based on uncertainty compression effect is introduced. Under the premise of satisfying the exclusivity constraint, unstable or divergent future operation paths can be automatically eliminated, so that the situation gradually converges from multiple divergent paths to a small number of controllable paths. The situation convergence path subgraph is further transformed into operational decision constraints. By jointly solving the accessibility of control actions, resource constraints, and cross-path consistency, it is ensured that the final generated park operation control instructions are valid under all feasible future paths, thereby avoiding the risk of medium- and long-term operational failure caused by a single path assumption. Attached Figure Description

[0015] Figure 1 This is a flowchart outlining the method steps of the present invention. Detailed Implementation

[0016] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0017] Refer to the instruction manual appendix Figure 1 An embodiment of the present invention provides a data-driven decision-making method for the convergence of operational status in smart parks, comprising: S1. Collect multi-source operation data of the smart park within the same operation cycle, perform time alignment and standardization processing on the multi-source operation data, and build a status node table in the digital twin to map the current operation status of the park into a set of status nodes containing status identifiers and resource occupancy fields. S1 includes: In this embodiment, in order to accurately depict the overall operational status of the smart park within the same operational cycle in the actual operating environment of the park, and to provide unified and computable basic state data for subsequent operational status recursion, conflict analysis, and status convergence decision-making, it is first necessary to uniformly collect, time-align, standardize, and process the operational data from different business subdomains of the park into state-based models. Through this process, the originally scattered, heterogeneous, and inconsistent time-granular operational data is transformed into a set of status nodes that can be directly used in calculations within the digital twin. This implementation process includes the following steps: In step S1-1, within the same operating cycle, operational data from different business subdomains of the smart park are collected. These business subdomains are divided according to the park's operational functions, with different subdomains corresponding to different types of operational data sources. For example, the energy management subdomain corresponds to the energy consumption metering system, the equipment operation subdomain to the equipment monitoring system, the environmental monitoring subdomain to the environmental sensing system, the security management subdomain to the access control or alarm system, and the business service subdomain to the personnel access or business processing system. During the operational data collection process, when operational data is generated or received by the corresponding collection device or business system, the current system time of that collection device or business system is directly read and written into the collection time identifier field of the operational data to identify the actual collection time of the operational data within the operating cycle. Simultaneously, when operational data enters the collection channel, a unique data source identifier is written for the operational data based on the type of business subdomain to which it belongs and the specific collection device or business system, clearly distinguishing which business subdomain within the park the operational data originates from and its specific source location. Through the above methods, an original operational data set containing both the collection time identifier and the data source identifier is formed. In steps S1-2, time alignment processing is performed on the original running data set based on the acquisition time identifier carried in the original running data set. Specifically, firstly, a unified running time granularity is preset as the time interval of the unified time axis, such as a minute or five minutes as a time period. Then, according to the acquisition time identifier in each piece of original running data, the original running data is divided and mapped to the corresponding time period. For running data with an acquisition frequency higher than the time granularity, multiple pieces of running data in the same time period are merged according to their data type. For example, continuous numerical running data is accumulated or summed within the time period, instantaneous measurement running data is selected as the representative value closest to the end of the time period, and state running data is statistically analyzed for the duration of each state within the time period and the state with the longest duration is selected as the representative state. For running data with an acquisition frequency lower than the time granularity, its most recent valid acquisition value is extended and mapped to the subsequent corresponding time period. Through the above processing, running data from different business subdomains and different acquisition frequencies are all mapped to time periods on the unified time axis, thereby obtaining a time-consistent running data sequence. In steps S1-3, based on the data source identifier carried in the time-consistent operational data sequence, a unified data standard processing is performed on the operational data. Specifically, the business subdomain type to which each piece of operational data belongs is identified according to the data source identifier, and the standard conversion rule corresponding to that business subdomain is invoked. The standard conversion rule is used to convert the measurement units, statistical methods, and expression forms used in different business subdomains into a preset unified data expression standard. For example, the cumulative energy consumption is converted into the energy consumption change per unit time, the percentage form of equipment load data is converted into the corresponding absolute load value, and the different state codes used in different systems are uniformly mapped to standard state values. By executing the corresponding standard conversion rule on each time-consistent operational data sequence, the operational data of different business subdomains are kept consistent in terms of numerical meaning and expression, thereby forming a standardized operational dataset. In steps S1-4, the operational data from the standardized operational dataset is input into the digital twin one by one in the order of data records. For each input operational data, the data source identifier and business subdomain identifier carried in the operational data are first read, and based on the data source identifier and business subdomain identifier, the specific operational object corresponding to the operational data is determined. The operational object can be a specific device, energy consumption metering point, environmental monitoring point, or business processing unit. After determining the operational object, the unique object identifier and the current operational cycle identifier of the operational object are read, and the unique object identifier and the operational cycle identifier are combined to generate the corresponding status identifier, and the status identifier is written into the status identifier field corresponding to the operational object. Subsequently, from the standardized operation... The dataset centrally filters out resource usage data related to the running object, such as energy consumption data, load data, or resource call records. Cumulative calculations are then performed on the filtered resource usage data within the running cycle to obtain the resource occupancy value for the running object within the running cycle. This resource occupancy value is then written into the resource occupancy field corresponding to the running object. After writing the status identifier and resource occupancy value for the running object, the data record containing the status identifier and resource occupancy fields is written into the status node table. After the status node table is constructed, status nodes are generated one by one according to the data records in the status node table. The generated status nodes are then aggregated to form a status node set, which is used to characterize the overall operating status of the smart park within the current operating cycle. Through the above implementation process, operational data from different business subdomains within a smart park during the same operational cycle can be uniformly transformed into a set of status nodes with clear time attributes, source attributes, and state meanings in a digital twin. This provides consistent and computable basic status data for subsequent future operational path construction, conflict analysis, and operational status convergence decisions. In practical applications, for example, in a smart park encompassing multiple business subdomains such as energy management, equipment operation, and environmental monitoring, this implementation method can uniformly map various energy consumption metering data, equipment status data, and environmental monitoring data within the park into a set of status nodes. This accurately reflects the park's true operational status within the current operational cycle in the digital twin and provides a reliable data foundation for subsequent operational decisions and control based on multi-objective assumptions. S2. Based on the set of situation nodes, construct a future path graph in the digital twin. Perform state recursion calculation for each situation node according to different operational goal assumptions. Record the obtained state transition relationship and resource change relationship as directed edges to form a path graph structure containing multiple future operational paths. S2 includes: In this embodiment, in order to prospectively extrapolate the current operating state of the smart park in the digital twin and systematically characterize the multiple possible evolution paths of the park's operating status under different operational target assumptions, it is necessary to construct a path graph structure that reflects future operational evolution relationships, starting from the set of status nodes, through step-by-step state recursion and constraint determination. The core of this process is to transform the static current operating status into recursively possible path seeds, introduce different operational target assumptions into the state evolution calculation, continuously introduce resource constraints and exclusion conditions during the recursion process, and prune evolution paths that do not meet the operational constraints in real time, thereby obtaining a set of feasible future operating paths within a given extrapolation step size. This implementation process includes the following steps: In step S2-1, the set of situation nodes is read one by one in node order. For each situation node read, the status identifier and resource occupancy field corresponding to the situation node are extracted first. The status identifier is used to uniquely indicate the running status instance corresponding to the situation node in the current running cycle, and the resource occupancy field is used to reflect the resource usage level of the park running objects under the situation node. After reading the status identifier and resource occupancy field, the status identifier, resource occupancy field and the pre-set initial inference step number are combined to generate a path seed record, where the initial inference step number is used to identify the inference starting level of the path seed record. The generated path seed records are written into the path seed queue in the order of reading the situation nodes, so that each record in the path seed queue corresponds to a situation node that can be used as the starting point of future running evolution. In step S2-2, path seed records are retrieved one by one from the path seed queue in the queue order; for each retrieved path seed record, running target assumptions are retrieved one by one from the running target assumption table in a preset assumption order; the running target assumptions are used to describe the operational focus or control strategy orientation that the park may follow in the future operation simulation, such as minimizing energy consumption, load balancing, or service priority; the current path seed record is combined with at least one running target assumption to generate binding pairs, wherein each binding pair explicitly indicates that state recursion is performed on the path seed record under the constraints of a specific running target assumption; the generated binding pairs are sequentially written into the recursion task queue, thereby forming a recursion task queue containing multiple recursion tasks for subsequent execution of state recursion calculations; In steps S2-3, recursive tasks are retrieved one by one from the queue in order. For each recursive task, the status identifier, resource usage field, and simulation step number in the corresponding binding pair are read first. Based on the status identifier, the current state instance of the corresponding situation node is located and determined in the digital twin, thereby obtaining the complete state field values ​​under that state instance. Subsequently, the state update rule corresponding to the running target assumption in the recursive task is read. The state update rule is used to limit the variability and update method of each state field under the running target assumption. According to the state update rule, state update calculations are performed item by item for the state fields corresponding to the status identifier. The process uses the resource occupancy field as a constraint input. For fields in the state field that are marked as permissible to change by the state update rules, numerical adjustments or state transition calculations are performed based on the operational target assumptions. For fields marked as not permissible to change by the state update rules, their original values ​​are kept unchanged during the recursive process. After completing the update calculations for all state fields, the updated state fields are recombined to generate the next situation node, and a corresponding state identifier is generated for this next situation node. The state identifier of the next situation node is written into a new path seed record to form a successor path seed record, and the simulation step number in the successor path seed record is updated to the next step number of the current simulation step number. In steps S2-4, while generating subsequent path seed records, resource changes and state changes are recorded. Specifically, the resource occupancy field in the previous path seed record and the resource occupancy field in the subsequent path seed record are read, and the difference between the two is calculated to obtain the resource change result generated during this state recursion. This resource change result is used to reflect the resource consumption or release caused by state evolution under a specific operational goal assumption. At the same time, the state identifier before and after recursion are read, and the difference between the two is determined to determine whether the state has changed and the direction of the change during this recursion. The obtained resource change result and state identifier change result are written together into a directed edge record, and the previous state node is taken as the starting point of the directed edge record, and the next state node is taken as the ending point of the directed edge record, thereby establishing the evolution relationship corresponding to one state recursion in the path graph structure. In steps S2-5, gating is performed on each of the generated directed edge records. First, the resource change results in the directed edge records are read, and then the resource change results are split into multiple resource change components according to resource type, so that each resource change component corresponds to a specific resource type. For each resource change component, the incompatible conditions corresponding to that resource type in the exclusive constraint table are read one by one, and the resource change component is compared item by item with the constraint thresholds or conflict rules in the incompatible conditions. During the comparison process, the predefined resource change judgment method in the incompatible conditions is read first: when the incompatible condition is defined as a threshold condition, it is determined whether the change amplitude of the resource change component exceeds the corresponding resource constraint threshold; when the incompatible condition is defined as a directional condition... When the resource change component changes direction, it is determined whether the change direction is consistent with the prohibited change direction specified in the incompatibility condition. When the incompatibility condition is defined as a combination condition, it is determined whether the combination result of the resource change component and other resource change components in the same directed edge record satisfies the conflict combination rule specified in the incompatibility condition. When any of the above determination results are true, it is determined that the directed edge record triggers the incompatibility condition. For directed edge records that trigger the incompatibility condition, their writing to the future path graph is blocked, and the corresponding subsequent path seed record is deleted simultaneously. For directed edge records that do not trigger any incompatibility conditions, the directed edge record is written to the future path graph, and the corresponding subsequent path seed record is written back to the recursive task queue to continue participating in subsequent state recursion. In steps S2-6, steps S2-3 to S2-5 are executed repeatedly until the recursive task queue is empty or the corresponding deduction step number in the recursive task reaches the preset upper limit of the deduction step size. After the recursion is completed, the future path graph is traversed, and the path segments formed by connecting the beginning and end of continuous directed edges are extracted one by one. Each path segment is output as a future running path. For each output future running path, a unique future running path identifier is written so that different future running paths can be distinguished in the subsequent conflict analysis, path pruning and running decision process, thereby forming a complete path graph structure containing multiple future running paths. Through the above implementation process, a set of future operating paths that meet resource constraints and operational goal assumptions can be systematically generated in the digital twin. This transforms the future evolution of the park's operational status from implicit assumptions into explicit path structures, significantly improving the ability to analyze future operational risks and feasibility. In practical applications, for example, in a smart park simultaneously constrained by energy consumption and equipment load, this implementation method can be used to extrapolate the operational evolution paths of the park over several future time steps under multiple operational goal assumptions, and automatically eliminate evolution paths with severe resource conflicts or violations of operational constraints. This provides a reliable and computable path basis for subsequent operational status convergence analysis and control decisions. S3. Perform traversal calculations on the path graph structure according to the time hierarchy, identify resource occupation conflicts and state transition conflicts of different future running paths at the same level, and write the conflict results into the path exclusivity matrix. S3 includes: In this embodiment, to further identify potential resource conflicts and state transition conflicts that may arise at the same extrapolation level for different future paths based on the already constructed future path graph structure, and to solidify these conflict relationships in an explicit and computable manner, it is necessary to perform a hierarchical traversal of the future path graph according to the extrapolation step number. Through this process, the resource competition and state exclusion relationships originally implicit in the path evolution process can be transformed into a path exclusivity matrix, providing a clear constraint basis for subsequent path pruning and situation convergence analysis. This implementation process includes the following steps: In step S3-1, directed edge records are read one by one from the future path graph structure according to the generation order of the directed edge records. For each directed edge record read, the deduction step number information carried in the directed edge record is first extracted. The deduction step number is used to indicate the specific level at which the state recursion corresponding to the directed edge occurs in the future deduction process. Then, directed edge records with the same deduction step number are appended to the same time level queue, so that each time level queue corresponds to a specific deduction step number. Through the above processing, multiple time level queues distinguished by deduction step number are formed in the future path graph structure, thereby structurally layering the path graph according to the time evolution order. In step S3-2, each time-level queue is processed separately. Specifically, for the current time-level queue, directed edge records are retrieved one by one according to the queue order, and the resource occupancy field contained in the endpoint status node of the directed edge record is read. Since the resource occupancy field usually contains information on multiple resource types, such as energy resources, load resources, or other constrained resources, it is necessary to split the resource occupancy field according to the resource type to obtain multiple resource occupancy components. Each resource occupancy component corresponds to the occupancy value of a specific resource type under the status node. Subsequently, the resource occupancy components are written into the resource occupancy index table with the future running path identifier and resource type as a joint key, so that the resource occupancy index table forms a complete resource occupancy snapshot for each future running path in the current time level, which is used for subsequent resource conflict calculation. In step S3-3, within the same time level, resource conflict calculation is performed on the resource occupancy index table. Specifically, the resource occupancy components in the resource occupancy index table are read one by one according to the resource type. For resource occupancy components belonging to different future running paths under the same resource type, pairwise superposition calculation is performed for each future running path pair. Through superposition calculation, the comprehensive occupancy level of the same resource type when different future running paths exist simultaneously at this time level can be obtained. Subsequently, each set of superposition calculation results is compared with the resource constraint threshold corresponding to the resource type. When any superposition result exceeds the resource constraint threshold of the corresponding resource type, it is determined that there is a resource occupancy conflict for the future running path pair at this time level, and the future running path pair that triggers the threshold is recorded as a resource conflict pair and written into the resource conflict registration table to explicitly identify this type of resource conflict relationship. In steps S3-4, state transition analysis continues to be performed on the current time-level queue. Specifically, directed edge records in the time-level queue are retrieved one by one in the queue order, and the state identifier change results recorded in each directed edge record are read. The state identifier change results are used to reflect the changes in each state field during the state recursion process corresponding to the directed edge. Subsequently, the state identifier change results are written into the state transition index table with the future running path identifier as the key, so that the state transition index table forms a corresponding state transition snapshot for each future running path in the current time level, thereby providing a directly comparable data basis for subsequent state transition conflict determination. In steps S3-5, within the same time level, state transition conflict determination is performed on the state transition index table. Specifically, the corresponding state identifier change results in the state transition index table are read pair by pair according to future running paths. For each pair of future running paths, the state identifier change results of the two future running paths within the current time level are read, and the two sets of state identifier change results are compared item by item by state field. During the item-by-item comparison, for the same state field, the state change values ​​of the two future running paths in that state field are read. When the state change values ​​corresponding to the two future running paths in the same state field point to mutually exclusive state value sets, for example, one path points to the enabled state while the other path points to the disabled state, Furthermore, if these two states are predefined as not being able to be simultaneously true, then it is determined that there is a state value conflict in the state field; or, when the state change values ​​corresponding to two future running paths on the same state field point to different state transition directions, and the state transition directions are restricted to not being able to be simultaneously true in the preset state transition constraints, for example, one path requires the state to rise while the other path requires the state to fall, and this combination is restricted to not being able to coexist, then it is determined that there is a state transition conflict in the state field; when a state value conflict or a state transition conflict is determined to exist on any state field, it is determined that the corresponding future running path pair has a state transition conflict at the current time level, and the future running path pair is recorded as a state conflict pair and written into the state conflict registration table; In steps S3-6, the resource conflict registration table and the status conflict registration table are merged. Specifically, records in the resource conflict registration table and the status conflict registration table are read one by one according to the future running path, and resource conflict pairs or status conflict pairs under the same future running path are uniformly written into the path exclusivity matrix. In the path exclusivity matrix, the future running path is used as the row and column index, and an incompatible mark is written in the corresponding matrix position to explicitly indicate that the two future running paths have a relationship that cannot be established at the same time level. In this way, the scattered resource conflict information and status conflict information are uniformly mapped into a structured exclusivity relationship matrix. In step S3-7, steps S3-2 to S3-6 are repeated for all time-level queues until all time-level queues have been processed. After completing the traversal calculation of all time levels, the filled path exclusivity matrix is ​​output. The path exclusivity matrix fully records the resource occupation conflict relationship and state transition conflict relationship of different future running paths at each projection level, thus providing a clear and directly callable judgment basis for subsequent path pruning and situation convergence analysis based on exclusivity constraints. Through the above implementation process, the potential resource competition and state exclusion relationships in the future path map can be transformed into an explicit path exclusion matrix, so that the incompatible relationships between different future operation paths have a clear data expression form and computability. In practical applications, such as when there are multiple future operation paths in a smart park and each path has a competitive relationship in terms of energy consumption, equipment status or operation mode, this implementation method can be used to identify and record the conflict between paths layer by layer, thereby providing reliable data support for subsequent screening of coexisting or preferred operation path combinations. S4. Based on the path exclusivity matrix, perform layer-by-layer pruning calculations on the path graph structure. Under the premise of satisfying the exclusivity constraint, calculate the compression effect of each future running path on the uncertainty of the subsequent situation, and remove the paths with decreasing compression effect layer by layer to form a situation convergence path subgraph. S4 includes: In this embodiment, to further converge a large number of possible future operating paths into a small number of stable, controllable operating paths with practical decision-making value based on the obtained path exclusivity matrix and future path graph structure, a layer-by-layer pruning mechanism based on uncertainty compression effect needs to be introduced. This mechanism combines the exclusivity relationship between paths with the uncertainty evolution characteristics of the paths themselves, and performs layer-by-layer evaluation and dynamic elimination of future operating paths without violating exclusivity constraints, thereby gradually bringing the park's operating situation from divergent to convergent. The implementation process includes the following steps: In step S4-1, the path exclusivity matrix is ​​read cell by cell according to row and column indices. The path exclusivity matrix uses the future running path identifier as the row and column indices, and its matrix cells are used to record whether there is an incompatible relationship between two corresponding future running paths. During the cell-by-cell reading process, when an incompatible mark is read in any matrix cell, it is determined that the two future running paths indicated by the row and column indices corresponding to that matrix cell constitute a pair of future running paths that cannot be simultaneously true. For each identified pair of future running paths, the pair of future running paths is written into the exclusivity registration table, which is used as an important constraint basis for path gating in the subsequent pruning process. In step S4-2, future running paths are extracted one by one from the path graph structure according to their future running path identifiers. For each extracted future running path, the situation nodes corresponding to the future running path at each simulation level are read layer by layer in ascending order of the simulation step number. For each situation node at each level, the status identifier and resource occupancy field are extracted. The status identifier reflects the running status instance of the future running path at the corresponding level, and the resource occupancy field reflects the resource usage level of the status instance. Subsequently, the status identifiers and resource occupancy fields extracted from each level are written into the path sequence record corresponding to the future running path in the order of the simulation step number, so that the path sequence record fully reflects the state evolution process of the future running path as the simulation level changes. In step S4-3, uncertainty decomposition calculation is performed layer by layer on the path sequence record corresponding to each future running path in hierarchical order. Specifically, at the current simulation level, the state identifier corresponding to this level is first read and compared with the state identifier corresponding to the previous simulation level. By counting the number of differences between the two at the state field level, a state transition count is obtained, which is used to measure the magnitude of the state change of the future running path at the current level. At the same time, the resource occupancy field of the current level is read and the difference is calculated with the resource occupancy field of the previous level to obtain the resource fluctuation amount, which is used to reflect the changes in resource usage of the future running path. The state transition count and the resource fluctuation amount are written together into the hierarchical uncertainty record corresponding to this level, thereby forming structured data that characterizes the uncertainty changes layer by layer on the basis of the path sequence record. In step S4-4, for each future running path, compression is calculated based on its uncertainty records at each level. Specifically, the uncertainty records at each level are read layer by layer according to the order of the simulation step number. At each level, the uncertainty record of the current level is compared with the uncertainty record of the previous level, and a subtraction operation is performed between the two to obtain the level compression amount corresponding to that level. The level compression amount is used to indicate whether the uncertainty is compressed or amplified during the state evolution process. Subsequently, the level compression amounts calculated at each level are written into the compression trajectory record in the order of the simulation step number, so that the compression trajectory record can fully reflect the trend of uncertainty change of the future running path throughout the simulation process. In steps S4-5, a descent determination is performed layer by layer on the compression trajectory record. Specifically, at the current projection level, the layer compression amount corresponding to the future running path at that level is read and compared with the layer compression amount of the previous projection level. When the layer compression amount of the current level is less than the layer compression amount of the previous level, it is determined that the future running path has an uncertain compression effect decrease at that projection level. For future running paths that are determined to have a compression effect decrease, they are written into the elimination candidate table, and the specific projection level that triggered the determination is recorded at the same time, so as to accurately locate the clipping starting point in subsequent clipping operations. In steps S4-6, a gating pruning operation is performed based on the exclusive registration table and the elimination candidate table. Specifically, for each future running path in the elimination candidate table, pruning is performed, deleting the situation node corresponding to the future running path at the trigger level and its subsequent levels from the path graph structure, and simultaneously deleting the directed edge records connected to the deleted situation node. At the same time, when the pruned future running path and another future running path in the exclusive registration table form a future running path pair, it is necessary to further prevent the other future running path from being retained at the same trigger level and its subsequent levels to avoid forming logically conflicting path combinations under the exclusive constraint. Through the gating pruning process, only the path portions that are reasonable in both the uncertainty evolution and exclusive constraint aspects are retained in the path graph structure. In steps S4-7, after gating and pruning, the updated path graph structure is traversed, and the connectivity of the remaining paths at each inference level is checked one by one. For future running paths that can still be continuously reachable between the inference start point and the inference end point, they are extracted and aggregated for output, forming a situation convergence path subgraph. The situation convergence path subgraph is used to centrally represent the set of future running paths that still hold under the dual conditions of exclusive constraints and uncertainty compression, providing a direct basis for subsequent running decisions and control action solutions. Through the above implementation process, future operating paths can be screened and pruned layer by layer based on the characteristics of uncertainty evolution without violating the path exclusivity relationship. This allows a large number of possible future operating paths to gradually converge into a small number of stable and controllable path structures. In practical applications, such as when smart parks face multiple operating strategy choices and complex intertwined resource and state constraints, this implementation method can automatically eliminate unstable or high-risk operating paths, retaining only path subgraphs with continuously decreasing uncertainty and satisfying exclusivity constraints. This provides reliable, clear, and executable situational convergence results for park operation control and scheduling decisions. S5. Write the situation convergence path subgraph into the digital twin as an operational decision constraint, perform consistency solution on the reachable control actions in the situation convergence path subgraph, generate park operation control instructions and output them to the actual park for execution, and write back the execution results to update the situation node table. S5 includes: In this embodiment, after constructing the situation convergence path subgraph, it is necessary to further transform this subgraph into a decision result that directly constrains and guides the actual operation of the park. By comprehensively solving for the reachability, consistency, and resource constraints of control actions, operational control commands that can be executed by the actual park are generated. The core of this process is: using the situation convergence path subgraph as the feasible future boundary, the control actions are trial-and-error, filtered, and converged in the digital twin, and the final decision result is fed back to the park operation situation model for updating. Through this implementation process, it can be ensured that the park operation control commands are always within the situation convergence constraints, avoiding operational risks caused by local optima or single-path assumptions. This implementation process includes the following steps: In step S5-1, the situation convergence path subgraph is read one by one according to the order of directed edge records. For each directed edge record read, the starting situation node identifier, the ending situation node identifier, and the corresponding inference step number are extracted. Then, the information is written as a constraint relationship into the constraint edge ledger in the digital twin. The constraint edge ledger uses the inference step number as a hierarchical index to record which situation nodes have reachable evolutionary relationships at different inference levels. In this way, the reachability constraints implicit in the situation convergence path subgraph are explicitly transformed into structured ledger data, thereby providing a unified data foundation for subsequent reachability judgment. In step S5-2, a reachable node index is generated layer by layer based on the constraint edge ledger. Specifically, for each simulation step number, the endpoint situation node identifier corresponding to that simulation step number is read one by one from the constraint edge ledger, and deduplication is performed on the read endpoint situation node identifiers to prevent the same situation node from being recorded repeatedly. The deduplicated situation node identifiers are uniformly written into the reachable node index table, so that the reachable node index table forms a clear situation node set under each simulation step number. This set is used to represent all situation nodes that are allowed to be reached under the situation convergence path subgraph constraint conditions at the corresponding simulation level, thereby transforming path-level constraints into node-level constraints. In step S5-3, the currently executable control actions in the park are read one by one according to the action identifier; the control actions may include equipment start / stop commands, operating parameter adjustment commands, or resource scheduling commands, etc.; for each read control action, an action trial calculation task is generated; when generating the action trial calculation task, the status identifier and resource occupancy field associated with the control action are first read from the current situation node table. The association relationship is used to indicate which running objects and their current status will be affected by the control action; then, the control action is combined with the corresponding status identifier and resource occupancy field and written into the action trial calculation task record, so that each action trial calculation task record fully describes the trial calculation conditions for executing the control action under the current situation; In step S5-4, action trial calculations are performed on each action trial task record. Specifically, for each action trial task record, the state transition rule corresponding to the control action is first invoked to perform a state update calculation on the state identifier in the record, resulting in a trial state identifier. The state transition rule is used to clarify how the state fields of the relevant running objects should change after the control action is executed. Subsequently, the resource impact rule corresponding to the control action is invoked to perform an update calculation on the resource occupancy field, resulting in a trial resource occupancy field. The resource impact rule is used to describe the impact of the control action on energy consumption, load, or other resource occupancy. By simultaneously completing the state update and resource update, a complete trial result node is formed, which represents the possible running state after the control action is executed in the digital twin. In step S5-5, dual-gating judgment is performed on each generated trial result node. First, the simulation step number corresponding to the trial result node is read, and the trial status identifier is compared with the situation node identifier corresponding to the simulation step number in the reachable node index table. When the trial status identifier exists in the reachable node index table corresponding to the simulation step number, it is determined that the control action is reachable at the situation evolution path level, thus passing the reachability gating. If it does not exist, it is determined that the reachability gating is not passed. Under the premise of passing the reachability gating, the trial resource occupation field is further read, and the trial resource occupation field is split into multiple resource occupation components according to the resource type. Then, each resource occupation component is compared with the preset resource constraint threshold of the corresponding resource type. When all resource occupation components meet the corresponding resource constraint threshold, it is determined that the resource gating is passed; otherwise, it is determined that the resource gating is not passed. Only when a certain control action passes both the reachability gating and the resource gating is the control action written into the feasible action table as a feasible candidate action under the current simulation level. In steps S5-6, a consensus solution is performed on the action list according to the simulation step number. Specifically, the control action records under the corresponding simulation step number are read one by one in the action list. Then, based on the situation convergence path subgraph, the situation nodes reachable by the corresponding future running path under the simulation step number are read one by one according to the future running path. The control actions that have passed the dual-gating judgment are selected on these situation nodes to form the action set corresponding to each future running path. Under the same simulation step number, the intersection calculation is performed on the action sets corresponding to different future running paths one by one to obtain the consensus action set that can be held under all future running path constraints. When the consensus action set is empty, it means that there is no control action that simultaneously satisfies all path constraints under the current simulation level. At this time, the process is rolled back to the previous simulation step number and the intersection calculation process is re-executed until a non-empty consensus action set is obtained, thereby realizing the rollback consensus solution of control actions in the time dimension. In steps S5-7, based on the obtained consensus action set, park operation control instructions are generated one by one, and the operation control instructions are output to the actual park for execution. After the control instructions are executed, the corresponding execution result data is collected, such as equipment status change results, energy consumption change results, or operation parameter feedback information. Subsequently, the execution result data is written back to the status node table one by one. During the write-back process, for each execution result data, the unique object identifier of the corresponding running object is read, and the corresponding status node record in the status node table is located accordingly. The status identifier and resource occupancy fields in the status node record are updated using the execution result data, thereby completing the synchronous update of the current operation status of the park in the digital twin. Through the above implementation process, the situation convergence path subgraph can be effectively transformed into control decisions that directly constrain the actual operation of the park. After the control action is executed, the operational situation can be written back and updated in a timely manner, forming a complete process of deduction, decision-making, execution, and feedback. In practical applications, such as when energy consumption scheduling or equipment operation optimization is carried out in a smart park, this implementation method can be used to select consistent and feasible control actions under various future operational path constraints, and update the park's operational situation in real time after execution, thereby ensuring that the park's operation is always in a controllable, predictable, and risk-limited state.

[0018] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A data-driven decision-making method for the convergence of operational status in smart parks, characterized in that, include: S1. Collect multi-source operation data of the smart park within the same operation cycle, perform time alignment and standardization processing on the multi-source operation data, and build a status node table in the digital twin to map the current operation status of the park into a set of status nodes containing status identifiers and resource occupancy fields. S2. Based on the set of situation nodes, construct a future path graph in the digital twin. Perform state recursion calculation for each situation node according to different operational goal assumptions. Record the obtained state transition relationship and resource change relationship as directed edges to form a path graph structure containing multiple future operational paths. S3. Perform traversal calculations on the path graph structure according to the time hierarchy, identify resource occupation conflicts and state transition conflicts of different future running paths at the same level, and write the conflict results into the path exclusivity matrix. S4. Based on the path exclusivity matrix, perform layer-by-layer pruning calculations on the path graph structure. Under the premise of satisfying the exclusivity constraint, calculate the compression effect of each future running path on the uncertainty of the subsequent situation, and remove the paths with decreasing compression effect layer by layer to form a situation convergence path subgraph. S5. Write the situation convergence path subgraph into the digital twin as an operational decision constraint, perform consistency solution on the reachable control actions in the situation convergence path subgraph, generate park operation control instructions and output them to the actual park for execution, and write back the execution results to update the situation node table.

2. The data-driven decision-making method for smart park operation status convergence according to claim 1, characterized in that: S1 includes: S1-1. Within the same operating cycle, collect operating data from different business subdomains of the smart park, and attach corresponding collection time identifiers and data source identifiers to various types of operating data to form a set of raw operating data with time attributes and source attributes. S1-2. Based on the time identifier, perform time alignment processing on the original set of running data to map running data with different collection frequencies and different time granularities onto a unified time axis to obtain a time-consistent running data sequence. S1-3. Based on the data source identifier, perform unified processing on time-consistent operational data sequences, converting the measurement methods, statistical rules and expression forms under different business subdomains into a unified data expression standard to form a standardized operational dataset. S1-4. Input the operational data from the standardized operational dataset into the digital twin one by one in the order of data records; For each piece of input running data, read the data source identifier and business subdomain identifier carried in the corresponding running data, and determine the specific running object corresponding to the corresponding running data based on the data source identifier and business subdomain identifier; After determining the running object, read the unique object identifier and the current running cycle identifier of the running object, combine the unique object identifier and the running cycle identifier to generate the corresponding status identifier, and write the status identifier into the status identifier field corresponding to the running object; Resource usage data corresponding to the running object is selected from the standardized running dataset. The resource usage data is accumulated and calculated within the running cycle to obtain the resource usage value, and the resource usage value is written into the resource usage field corresponding to the running object. After completing the writing of the status identifier and resource usage value of the running object, the data record containing the status identifier field and resource usage field is written to the status node table. After the situation node table is constructed, situation nodes are generated one by one according to the data records in the situation node table, and the generated situation nodes are collected to form a situation node set.

3. The data-driven decision-making method for smart park operation status convergence according to claim 2, characterized in that: S2 includes: S2-1. Read the situation nodes one by one in the situation node set according to the node order. For each situation node, read the corresponding status identifier and resource occupation field, and write the status identifier, resource occupation field and initial simulation step number into a path seed record. Write the generated path seed records into the path seed queue in sequence. S2-2. Take out the path seed records one by one from the path seed queue in the queue order, and take out the running target hypothesis one by one from the running target hypothesis table in the hypothesis order. Combine the current path seed record with at least one running target hypothesis to generate a binding pair; write the binding pairs sequentially into the recursive task queue to output the recursive task queue; S2-3. Take out the recursive tasks one by one from the queue in the order of the queue, and read the status flag, resource occupation field and deduction step number of the corresponding binding pair of the recursive task. Based on the status identifier, determine the current status instance of the corresponding situation node in the digital twin, and read the status update rule corresponding to the running target assumption; According to the status update rules, status update calculations are performed on each status field corresponding to the status identifier. The resource occupancy field is used as a constraint input. The fields that are allowed to change are updated, while the fields that are not allowed to change remain unchanged. The updated state fields are combined to generate the next situation node, and the state identifier corresponding to the next situation node is written into the new path seed record to form the successor path seed record. At the same time, the deduction step number of the successor path seed record is updated to the next deduction step number.

4. The data-driven decision-making method for smart park operation status convergence according to claim 3, characterized in that: S2 also includes: S2-4. When generating the subsequent path seed record, read the resource occupation field in the previous path seed record and the resource occupation field in the subsequent path seed record and perform difference calculation to obtain the resource change result. Read the state flags before and after recursion and perform a difference determination to obtain the state flag change results; Write the resource change results and the state identifier change results into the directed edge record, and take the previous state node as the starting point of the directed edge record and the next state node as the ending point of the directed edge record. S2-5. Perform gating judgment on each of the generated directed edge records, including reading the resource change results in the directed edge records and splitting the resource change results into resource change components according to the resource type. For each resource change component, read the incompatible conditions of the corresponding resource type in the exclusive constraint table one by one, and perform item-by-item comparison between the resource change component and the constraint threshold or conflict rule in the incompatible conditions; When comparing a certain resource change component, the predefined resource change determination method in the corresponding incompatibility condition is read; Among them, when the incompatibility condition is defined as a threshold condition, it is determined whether the change magnitude of the resource change component exceeds the corresponding constraint threshold. When the incompatibility condition is defined as a directional condition, it is determined whether the direction of change of the resource change component is consistent with the prohibited change direction specified in the incompatibility condition. When the incompatibility condition is defined as a combination condition, determine whether the combination result of the resource change component and other resource change components in the same directed edge record satisfies the conflict combination rules defined in the incompatibility condition. When any decision result is true, the directed edge record corresponding to the decision triggers an incompatible condition; When an incompatible condition is triggered, the corresponding directed edge record is blocked from being written into the future path graph and the corresponding successor path seed record is deleted. When no incompatible condition is triggered, the corresponding directed edge record is written into the future path graph, and the corresponding successor path seed record is written back to the recursive task queue. S2-6. Repeat S2-3 to S2-5 until the recursive task queue is empty or the deduction step number corresponding to the recursive task reaches the preset step size; traverse the future path graph, extract and output the path segments formed by connecting the beginning and end of continuous directed edges as future running paths, thereby forming a path graph structure containing multiple future running paths, and write a future running path identifier for each future running path.

5. The data-driven decision-making method for smart park operation status convergence according to claim 4, characterized in that: S3 includes: S3-1. Read the directed edge records one by one from the path graph structure in the order of generation, extract the deduction step number corresponding to each directed edge record, and append the directed edge records with the same deduction step number to the same time level queue to form multiple time level queues distinguished by deduction step number. S3-2. For each time-level queue, retrieve the directed edge records one by one in the queue order, read the resource occupation field of the corresponding directed edge record endpoint state node, and split the corresponding resource occupation field into resource occupation components according to resource type. Write the resource usage component into the resource usage index table with the future running path identifier and resource type as the key, so that the resource usage index table forms a resource usage snapshot for each future running path in the corresponding time level; S3-3. Within the same time level, read the resource occupancy components in the resource occupancy index table one by one according to the resource type, and perform pairwise superposition calculations on the resource occupancy components belonging to different future running paths under the same resource type. When any superposition result exceeds the resource constraint threshold corresponding to the resource type, the future running path pair that triggers the threshold will be recorded as a resource conflict pair and written into the resource conflict registration table.

6. The data-driven decision-making method for smart park operation status convergence according to claim 5, characterized in that: S3 also includes: S3-4. For each time level queue, take out the directed edge record one by one in the queue order, read the state identifier change result of the corresponding directed edge record, and write the state identifier change result into the state transition index table with the future running path identifier as the key, so that the state transition index table forms a state transition snapshot for each future running path in the corresponding time level. S3-5. Within the same time level, read the corresponding status identifier change results in the status transition index table one pair at a time according to the future running path. For each pair of future running paths, read the status identifier change results of the two future running paths within the corresponding time level, and compare the two sets of status identifier change results item by item according to the status field. When performing a step-by-step comparison of the status changes of two future running paths, for the same status field, the status change values ​​of the two future running paths in the corresponding status field are read respectively. When two future running paths have corresponding state change values ​​in the same state field that point to mutually exclusive set of state values, it is determined that there is a state value conflict in the corresponding state field. Alternatively, when the state change values ​​of two future running paths on the same state field point to different state transition directions, and the state transition directions are restricted in the preset state transition constraints to not being able to be met at the same time, it is determined that there is a state transition conflict in the corresponding state field. When a conflict in state value or state transition is determined to exist in any state field, it is determined that the corresponding future running path pair has a state transition conflict at the corresponding time level. When a state transition conflict is determined to exist, the corresponding future running path pair is recorded as a state conflict pair and written into the state conflict registration table. S3-6. Perform a merge and write process on the resource conflict registration table and the status conflict registration table according to the future running path pair. Write the resource conflict pair or status conflict pair of the future running path pair into the path exclusivity matrix, and write an incompatible mark in the path exclusivity matrix at the position corresponding to the future running path pair. S3-7. Repeat S3-2 to S3-6 for all time-level queues until all time-level queues have been processed, thereby outputting the filled path exclusivity matrix.

7. A data-driven decision-making method for smart park operation status convergence according to claim 6, characterized in that: S4 includes: S4-1. Read the incompatible markers cell by cell according to the row and column indices of the path exclusivity matrix, extract the corresponding future running path pair for each incompatible marker, and write the corresponding future running path pair into the exclusivity registration table. S4-2. Extract future running paths one by one in the path diagram structure according to the future running path identifier. For each future running path, read the corresponding situation nodes layer by layer according to the simulation step number from small to large. Extract the status identifier and resource occupation field of each level of situation node, and write the status identifier and resource occupation field into the path sequence record corresponding to the future running path in hierarchical order. S4-3. Perform uncertainty decomposition calculation on the path sequence record corresponding to each future running path in hierarchical order, wherein the state identifier of the corresponding level is read at the current level and the difference count is performed with the state identifier of the previous level to obtain the state transition count. Simultaneously, the resource usage field of the corresponding level is read and the difference is calculated with the resource usage field of the previous level to obtain the resource fluctuation amount; Write the state transition count and resource fluctuation amount into the hierarchical uncertainty record.

8. A data-driven decision-making method for smart park operation status convergence according to claim 7, characterized in that: S4 also includes: S4-4. For each future running path, perform compression calculations on the uncertainty records of each level in hierarchical order. At each level, read the uncertainty record of the current level and the uncertainty record of the previous level, and perform a subtraction operation on the two to obtain the level compression amount. Write the compression amounts of each level into the compression trajectory record in sequence. S4-5. Perform a descent judgment on the compression trajectory record layer by layer. In the current layer, read the compression amount of the corresponding future running path and compare it with the compression amount of the previous layer. When the compression amount of the current layer is less than the compression amount of the previous layer, write the corresponding future running path into the elimination candidate table and record the corresponding triggering layer. S4-6. Perform gating pruning based on the exclusive pair registration table and the elimination candidate table. Specifically, perform pruning operations on each future running path in the elimination candidate table: delete the situation node of the corresponding future running path at the trigger level and subsequent levels from the path graph structure, and delete the directed edge record connected to the deleted situation node; at the same time, when a future running path and another future running path in the exclusive pair registration table form a future running path pair, prevent the other future running path from being retained at the same trigger level and subsequent levels. S4-7. After completing the gating pruning, traverse the updated path graph structure to extract the future running paths that are still continuously reachable, and collect and output the future running paths as a situation convergence path subgraph.

9. A data-driven decision-making method for smart park operation status convergence according to claim 8, characterized in that: S5 includes: S5-1. Read the situation convergence path subgraph one by one in the order of directed edge records, and write the starting situation node identifier, ending situation node identifier and corresponding deduction step number of each directed edge record into the constraint edge ledger of the digital twin to form a reachability constraint relationship layered by deduction step number. S5-2. Generate reachable node index layer by layer based on constraint edge ledger. For each simulation step number, read the endpoint situation node identifier corresponding to the simulation step number one by one and write it into the reachable node index table after deduplication to obtain the set of situation nodes that are allowed to be reached under the corresponding simulation step number. S5-3. Read the executable control actions in the park one by one according to the action identifier, and generate an action trial calculation task for each control action. In this process, read the status identifier and resource occupation field associated with the corresponding control action in the current situation node table, and write the corresponding control action, status identifier and resource occupation field into the action trial calculation task record. S5-4. Perform action trial calculations for each action trial task record. For each action trial task record, call the state transition rule corresponding to the corresponding control action to perform an update calculation on the state identifier to generate a trial state identifier, and call the resource impact rule corresponding to the corresponding control action to perform an update calculation on the resource occupancy field to generate a trial resource occupancy field, thereby forming a trial result node.

10. A data-driven decision-making method for smart park operation status convergence according to claim 9, characterized in that: S5 also includes: S5-5. Perform dual-gating judgment on each trial result node. First, read the simulation step number corresponding to the trial result node, and then compare the trial status identifier with the status node identifier corresponding to the simulation step number in the reachable node index table. If the trial status identifier exists in the reachable node index table corresponding to the simulation step number, the reachability gate is passed; otherwise, the reachability gate is not passed. Under the premise of reaching the gate, read the trial resource usage field and split the trial resource usage field into resource usage components according to the resource type; Each resource occupancy component is compared with the preset resource constraint threshold of the corresponding resource type. When all resource occupancy components meet the corresponding resource constraint threshold, the resource gating is deemed to pass; otherwise, the resource gating is deemed to fail. The corresponding control action is written to the action table only when both reachability gating and resource gating are used simultaneously. S5-6. Perform consistency solution on the action table according to the simulation step number, wherein the control action record under the corresponding simulation step number is read one by one in the action table; Based on the situation convergence path subgraph, the situation nodes reachable by the corresponding future operation path under the corresponding simulation step number are read one by one according to the future operation path, and the control actions determined by dual gating at the situation nodes are selected to form the set of actionable actions corresponding to the future operation path. Under the same simulation step number, the intersection calculation is performed on the sets of possible actions corresponding to different future running paths one by one to obtain the set of consensus actions that can be held under all future running path constraints; When the consensus action set is empty, roll back to the previous derivation step number and re-execute the intersection calculation until a non-empty consensus action set is obtained; S5-7. Generate park operation control instructions one by one based on the consensus action set and output them to the actual park for execution. After execution, collect the execution result data and write it back to the status node table one by one. For each execution result data, read the unique object identifier of the corresponding running object and locate the corresponding status node record. Update the status identifier and resource occupation field of the status node record with the execution result data, thereby completing the synchronous update of the current operation status of the park.

Images