A germplasm bank operation scene adaptive scheduling method based on a semantic network

Abstract

The application discloses a kind of germplasm bank operation scene self-adapting scheduling method based on semantic network, including the following steps: obtaining germplasm sample, operation task, resource state, environmental monitoring and abnormal event data, constructs germplasm bank operation scene semantic network;Fuzzy time compilation is carried out to semantic network, and fuzzy time scene description result is generated;Generation time compression, resource congestion and abnormal diffusion propagation edge;Input phase transition enhancement FT-SWRL semantic reasoning network, and output scene phase state, migration direction, scheduling stability and re-scheduling trigger result;Extract adjustable task, replaceable resource and correctable path and generate phase transition scheduling cone;Under the constraint of cone, gradually re-scheduling is generated, and self-adapting scheduling scheme is generated;Form closed loop by executing and feeding back update.The application uses semantic network and phase transition enhancement FT-SWRL reasoning method to realize germplasm bank self-adapting scheduling, with the advantages of timely response, controllable disturbance and high scheduling stability.

Description

Technical Field

[0001] This invention relates to the field of agricultural information technology, and in particular to an adaptive scheduling method for germplasm bank operation scenarios based on semantic networks. Background Technology

[0002] Germplasm banks undertake various tasks such as germplasm sample storage, retrieval, testing, verification, transportation, rewarming, re-storage, and inventory. In actual operation, they are simultaneously affected by sample preservation conditions, task time windows, resource occupancy status, environmental fluctuations, and abnormal events. Existing technologies often manage tasks, equipment resources, and environmental monitoring data separately, or use rule bases, scheduling tables, static priority ranking, and general knowledge graph association analysis to generate scheduling results. Some solutions also incorporate digital twins, reinforcement learning, or event-driven mechanisms to adjust the workflow, but their core remains focused on general scheduling optimization or single-event response.

[0003] However, existing technologies often struggle to integrate the accumulated time pressure, resource congestion propagation, anomaly impact diffusion, and high-sensitivity sample protection requirements in germplasm bank scenarios into a single semantic reasoning chain, resulting in insufficient ability to identify the scenario evolution process. At the scheduling execution level, existing methods often lack a constrained scheduling domain generation mechanism based on scenario phase and migration direction, as well as rescheduling rules that are progressively implemented by local reordering, resource substitution, path correction, and critical task preemption. They can often only perform coarse-grained adjustments, making it difficult to balance critical task protection, disturbance range control, and execution feedback closed-loop updates. Consequently, they are unable to meet the high-constraint, dynamic, and adaptive job scheduling requirements of germplasm banks.

[0004] Therefore, how to provide an adaptive scheduling method for germplasm bank operations based on semantic networks is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0005] One objective of this invention is to propose an adaptive scheduling method for germplasm bank operations based on semantic networks. This invention uses semantic networks and phase transition enhanced FT-SWRL inference to achieve adaptive scheduling of the germplasm bank, which has the advantages of timely response, controllable disturbances, and high scheduling stability.

[0006] An adaptive scheduling method for germplasm bank operation scenarios based on semantic networks according to an embodiment of the present invention includes the following steps: Acquire germplasm sample data, task data, resource status data, environmental monitoring data, and abnormal event data to construct a semantic network for germplasm bank operation scenarios; Fuzzy temporal compilation is performed on the semantic network of germplasm bank operation scenarios to generate fuzzy temporal scenario description results; Based on the semantic network of germplasm bank operation scenarios, propagation semantic relationship results are formed by generating time compression propagation edges, resource congestion propagation edges, and abnormal diffusion propagation edges. The fuzzy temporal scene description results and propagation semantic relationship results are input into the phase transition enhanced FT-SWRL semantic reasoning network to generate scene phase state results, phase state transition direction results, scheduling stability results, and rescheduling trigger results. Based on the scenario phase state results, phase transition direction results, and scheduling stability results, extract adjustable task subsets, alternative resource subsets, and correctable path subsets, and generate a phase change scheduling cone; Under the phase change scheduling cone constraint, an adaptive scheduling scheme is generated by progressively rescheduling the adjustable task subset, the alternative resource subset, and the correctable path subset based on the rescheduling triggering result. The adaptive scheduling scheme is executed and the execution feedback results are collected. The semantic network of the germplasm bank operation scenario and the phase transition enhanced FT-SWRL semantic reasoning network are updated to form an adaptive scheduling closed loop.

[0007] Optionally, the generation of the semantic network for the germplasm bank operation scenario specifically includes: Acquire germplasm sample data, extract fields and standardize formats to generate a set of germplasm sample entries; Acquire and organize the task data to generate a set of task entries; Acquire resource status data and encode it uniformly to generate a set of resource status entries; Acquire environmental monitoring data and abnormal event data, perform time alignment and spatial attribution mapping, and generate a set of environmental event entries; Based on the set of germplasm sample entries, the set of job task entries, the set of resource status entries, and the set of environmental event entries, perform primary key association and consistency verification to generate a set of cross-source association results. Initial results of the semantic network for the job scenario are constructed based on the cross-source association result set; The initial results of the job scenario semantic network are processed by writing node states, edge states, and masking of abnormal conflicts to generate the germplasm bank job scenario semantic network.

[0008] Optionally, the generation of the fuzzy temporal scene description result specifically includes: Based on the semantic network of germplasm bank operation scenarios, extract the status of task time-related nodes, resource status-related nodes, environment nodes, and abnormal event nodes to generate a fuzzy time compilation input set. Based on the fuzzy time compilation input set, the job task execution chain is parsed for time window, generating task time window results and remaining buffer duration results, and then written back to the corresponding job task node according to the task identifier; Based on the task time window results, remaining buffer duration results, and resource status related node status execution time pressure, a time compression result is generated; Based on the environmental node status, abnormal event node status, and germplasm sample node status, risk status compilation is performed to generate environmental deviation status results, resource busy status results, and event persistence status results. Based on the results of time compression, environmental deviation, resource busyness, event persistence, and the storage temperature zone, activity level, and current operation status of germplasm sample nodes, the scenario stage mapping is generated. The task time window result, remaining buffer duration result, time compression degree result, environment deviation state result, resource busy state result, event persistence state result, and scene stage result are uniformly encapsulated to generate fuzzy time scene description results.

[0009] Optionally, the scenario phase mapping divides the job chain state corresponding to each task object into phases based on the combined hit relationship of task time pressure, regional environmental deviation, resource load accumulation, event continuous impact and sample job sensitivity. These phases include stable phase, tense phase, disturbance phase and recovery phase. The scenario phase results are then jointly bound with task identifier, resource identifier, job area code and time grid index.

[0010] Optionally, the generation of the propagation semantic relationship results specifically includes: Based on the semantic network of germplasm bank operation scenario, extract each node set and sample-task association edge, task-resource requirement edge, resource-region affiliation edge, environment-region effect edge and abnormal event-object impact edge, generate propagation relationship to construct input set; Based on the propagation relationship, the task time window result, remaining buffer duration result, time compression degree result, and scene stage result in the input set and fuzzy time scene description result are constructed. Time compression candidate propagation pairs are constructed, and the initial result of time compression propagation edge is generated. Based on the propagation relationship, construct the resource busy status result, scenario stage result, and current load, number of bound tasks, and number of queued tasks corresponding to the resource node from the input set and fuzzy time scenario description result. Construct resource congestion candidate propagation pairs and generate initial results of resource congestion propagation edges. Based on the propagation relationship, construct the input set and the environmental deviation state result, event persistence state result, scene stage result, as well as the object of abnormal event, environmental node state and current operation state in the fuzzy time scene description result, construct the abnormal diffusion candidate propagation pair, and generate the initial result of abnormal diffusion propagation edge. Perform edge direction correction, edge validity screening, and edge conflict resolution on the initial results of the time compression propagation edge, the initial results of the resource congestion propagation edge, and the initial results of the abnormal diffusion propagation edge to generate time compression propagation edge, resource congestion propagation edge, and abnormal diffusion propagation edge; Perform unified association encapsulation on time-compression propagation edges, resource congestion propagation edges, and anomaly diffusion propagation edges to generate propagation semantic relationship results.

[0011] Optionally, the generation of the scenario phase state result, phase transition direction result, scheduling stability result, and rescheduling trigger result specifically includes: The fuzzy temporal scene description results and propagated semantic relationship results are input into the fact compilation unit of the phase transition enhanced FT-SWRL semantic reasoning network to perform fact compilation and generate a set of reasoning facts. The set of reasoning facts is input into the rule mapping unit to generate a set of rule-acting objects and a set of rule-hitting candidates; In the phase transition candidate inference unit, phase transition candidate inference is performed based on the rule action object set and the rule hit candidate set to generate scene phase candidate results and basic results for transition determination. The scene phase candidate results and the basic results of the migration decision are input into the migration adjudication unit to generate the scene phase results and the phase migration direction results; Based on the scenario phase state results, phase state transition direction results, and propagation fact set, a stability assessment is performed, and scheduling stability results are generated. Based on the trigger decision unit, trigger decisions are executed on the scene phase state result, phase transition direction result, scheduling stability result and event fact set to generate rescheduling trigger result.

[0012] Optionally, the phase transition enhanced FT-SWRL semantic reasoning network includes a fact compilation unit, a rule mapping unit, a phase transition candidate reasoning unit, a transition adjudication unit, and a trigger adjudication unit.

[0013] Optionally, the generation of the phase change scheduling cone specifically includes: Based on the scene phase state results, phase transition direction results, scheduling stability results, and rescheduling trigger results, perform cone generation input alignment to generate a phase change scheduling cone generation input set; Based on the phase change scheduling cone, the input set is generated to perform scheduling domain partitioning, and task scheduling domain results, resource scheduling domain results, path scheduling domain results, and critical task protection results are generated. Based on the results of task scheduling domain, resource scheduling domain, path scheduling domain, critical task protection, and semantic network execution object selection for germplasm bank operation scenarios, adjustable task subsets, alternative resource subsets, and correctable path subsets are generated. Based on the scene phase state results, phase transition direction results, scheduling stability results, and critical task protection results, cone boundary adjudication is performed to generate cone boundary results and cone stability results; Based on the results of task scheduling domain, resource scheduling domain, path scheduling domain, critical task protection, cone boundary, and cone stability, a unified encapsulation and consistency verification are performed to generate a phase change scheduling cone.

[0014] Optionally, the generation of the adaptive scheduling scheme specifically includes: Based on the phase change scheduling cone, rescheduling trigger results, adjustable task subsets, alternative resource subsets, and correctable path subsets, perform rescheduling input alignment to generate a rescheduling input set; Based on the rescheduled input set, local reordering is performed on the adjustable task subset under the constraints of the task scheduling domain results to generate local reordering candidate results; When the rule target corresponding to the local reorder candidate result still has a rescheduling trigger flag, resource substitution is performed on the substitutable resource subset under the resource scheduling domain result constraint to generate resource substitution candidate results; When the rule object corresponding to the resource substitution candidate result still has a rescheduling trigger flag, path correction is performed on the correctable path subset under the constraint of the path scheduling domain result to generate path correction candidate results. When the rule object corresponding to the path correction candidate result still has a rescheduling trigger flag, and the scene phase result is in the disturbance stage or the phase migration direction result is forward migration, perform critical task preemption on the adjustable task subset and the alternative resource subset, and generate preemption candidate results. A progressive decision-making process is performed on each candidate result in the order of local reordering, resource substitution, path correction, and priority of critical tasks to generate an adaptive scheduling scheme.

[0015] Optionally, the generation of the adaptive scheduling closed loop specifically includes: Based on an adaptive scheduling scheme, task assignment and status feedback collection are performed to generate an execution feedback input set. Based on the execution feedback input set, execution feedback regularization is performed to generate execution feedback results. Based on the execution feedback results, update the node and edge states in the semantic network of the germplasm bank operation scenario to generate semantic network update results; Based on the execution feedback results and semantic network update results, the propagation feedback write-back is performed to generate propagation relationship correction results; The phase transition enhanced FT-SWRL semantic inference network is updated based on the execution feedback results, semantic network update results, and propagation relation correction results, and the inference network update results are generated. Based on the semantic network update results, propagation relationship correction results, and inference network update results, a closed-loop consistency check is performed to form an adaptive scheduling closed loop.

[0016] The beneficial effects of this invention are: This invention constructs a semantic network for germplasm bank operation scenarios, unifying germplasm sample data, operation task data, resource status data, environmental monitoring data, and abnormal event data into a single semantic association framework. It then combines this with fuzzy time compilation to generate task time window results, remaining buffer duration results, time compression degree results, environmental deviation status results, resource busy status results, and event persistence status results. This allows the previously scattered time constraints, environmental disturbances, resource loads, and abnormal impacts in the germplasm bank scenario to be uniformly expressed and continuously tracked, thereby improving the completeness of the operation scenario representation and the accuracy of status identification. It enables earlier identification of operational instability trends caused by time compression, resource congestion, and abnormal spread.

[0017] This invention further forms propagation semantic relationship results through time compression propagation edges, resource congestion propagation edges, and anomaly diffusion propagation edges. It also uses the phase transition enhanced FT-SWRL semantic reasoning network to output scene phase results, phase transition direction results, scheduling stability results, and rescheduling trigger results. This enables the scheduling process to no longer rely solely on a single event or static rule for response, but to implement dynamic adjudication based on the migration relationship between the stable phase, stress phase, disturbance phase, and recovery phase of the scene. This improves the reasoning depth and triggering accuracy for complex germplasm bank operation scenarios and avoids the amplification of local anomalies into global scheduling disorder.

[0018] Meanwhile, this invention extracts adjustable task subsets, alternative resource subsets, and correctable path subsets based on scene phase state results, phase state transition direction results, and scheduling stability results, and generates a phase transition scheduling cone. Under the constraints of the phase transition scheduling cone, progressive event rescheduling is implemented in the order of local reordering, resource substitution, path correction, and critical task preemption. This ensures that the rescheduling scope is always limited to the controlled scheduling domain, which can not only eliminate the disturbance source corresponding to the rescheduling trigger result in a timely manner, but also reduce unnecessary interference to critical tasks, tasks corresponding to highly sensitive samples, and tasks associated with critical resources. Therefore, it is beneficial to reduce the number of invalid reorderings, narrow the scope of scheduling disturbances, and improve the continuous execution capability of critical operation chains.

[0019] Furthermore, after executing the adaptive scheduling scheme, this invention synchronously updates the semantic network of the germplasm bank operation scenario, the propagation of semantic relationship results, and the phase transition enhanced FT-SWRL semantic inference network through the execution feedback results. It also forms an adaptive scheduling closed loop through closed-loop consistency verification, enabling subsequent scheduling to directly inherit the actual state, propagation correction state, and inference correction state after the previous round of execution. This enhances the scheme's adaptability to dynamic environments and continuous disturbances. Therefore, this invention can achieve higher scheduling stability, resource utilization coordination, and timely anomaly response in the high-constraint, time-sensitive, and multi-disturbance concurrent operation environment of the germplasm bank. Attached Figure Description

[0020] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is a flowchart of an adaptive scheduling method for germplasm bank operation scenarios based on semantic networks proposed in this invention; Figure 2 This is a flowchart illustrating the fuzzy temporal compilation and propagation relationship construction of an adaptive scheduling method for germplasm bank operations based on semantic networks proposed in this invention. Figure 3 This is a flowchart illustrating the phase transition reasoning and progressive scheduling process of an adaptive scheduling method for germplasm bank operations based on semantic networks proposed in this invention. Detailed Implementation

[0021] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0022] refer to Figures 1-3 An adaptive scheduling method for germplasm bank job scenarios based on semantic networks includes the following steps: Acquire germplasm sample data, task data, resource status data, environmental monitoring data, and abnormal event data to construct a semantic network for germplasm bank operation scenarios; Fuzzy temporal compilation is performed on the semantic network of germplasm bank operation scenarios to generate fuzzy temporal scenario description results; Based on the semantic network of germplasm bank operation scenarios, propagation semantic relationship results are formed by generating time compression propagation edges, resource congestion propagation edges, and abnormal diffusion propagation edges. The fuzzy temporal scene description results and propagation semantic relationship results are input into the phase transition enhanced FT-SWRL semantic reasoning network to generate scene phase state results, phase state transition direction results, scheduling stability results, and rescheduling trigger results. Based on the scenario phase state results, phase transition direction results, and scheduling stability results, extract adjustable task subsets, alternative resource subsets, and correctable path subsets, and generate a phase change scheduling cone; Under the phase change scheduling cone constraint, an adaptive scheduling scheme is generated by progressively rescheduling the adjustable task subset, the alternative resource subset, and the correctable path subset based on the rescheduling triggering result. The adaptive scheduling scheme is executed and the execution feedback results are collected. The semantic network of the germplasm bank operation scenario and the phase transition enhanced FT-SWRL semantic reasoning network are updated to form an adaptive scheduling closed loop.

[0023] In this embodiment, the generation of the semantic network for germplasm bank operation scenarios specifically includes: Acquire germplasm sample data, extract fields and standardize formats to generate a set of germplasm sample entries; The field extraction process uses a pre-set field mapping table to convert the sample attribute fields in the collection source, including sample number, germplasm category, storage temperature zone, activity level, batch number, current location, current operation status and timestamp, into one-to-one correspondences. Missing fields are marked with null values ​​and illegal fields are marked with exceptions. The unified sample number is used as the unique index in the germplasm sample item set. Acquire and organize the task data to generate a set of task entries; The task data includes task type, task target, task source, starting work area, target work area, preceding task identifier, task priority level, planned start time, planned end time, and task execution status. The task types include inbound tasks, outbound tasks, inspection tasks, verification tasks, transfer tasks, reheating tasks, re-entry into storage tasks, and inventory tasks. The task target is identified by combining the sample number and batch number. Acquire resource status data and encode it uniformly to generate a set of resource status entries; The resource status data includes resource identifiers, associated work areas, occupancy status, availability status, service capacity, current load, and status timestamps for cold storage locations, temporary storage locations, testing stations, transfer devices, robotic arms, manual workstations, and passageway sections. The service capacity is written according to a pre-defined correspondence table between resource types and executable task types. The current load is generated by combining the duration of resource occupancy, the number of queued tasks, and the number of bound tasks. Acquire environmental monitoring data and abnormal event data, perform time alignment and spatial attribution mapping, and generate a set of environmental event entries; The environmental monitoring data includes temperature, humidity, vibration, access control status, area congestion status, and equipment alarm status. The abnormal event data includes the type of abnormal event, the object affected by the abnormal event, the time of occurrence of the abnormal event, and the duration of the abnormal event. The time alignment uses a unified time grid to resample the data from each sampling source. The spatial attribution mapping binds the environmental monitoring data and abnormal event data to the corresponding work area based on the work area code, storage location code, and channel segment code. Based on the set of germplasm sample entries, the set of job task entries, the set of resource status entries, and the set of environmental event entries, perform primary key association and consistency verification to generate a set of cross-source association results. The primary key association uses sample number, task identifier, resource identifier, job area code and time grid index for joint matching. The consistency check determines the state conflict, position conflict and time reversal conflict of the same object in different data sources and outputs the conflict flag. Initial results of the semantic network for the job scenario are constructed based on the cross-source association result set; The initial results of the semantic network for the work scenario include a set of germplasm sample nodes, a set of work task nodes, a set of resource nodes, a set of environment nodes, and a set of abnormal event nodes, as well as sample-task association edges, task-resource requirement edges, resource-region affiliation edges, environment-region effect edges, and abnormal event-object impact edges. Among them, the sample-task association edge is generated by matching the task object identifier with the sample number, the task-resource requirement edge is generated by matching the task type with the service capability, and the abnormal event-object impact edge is generated by matching the abnormal event object with the sample number, task identifier, or resource identifier. Perform node state writing, edge state writing, and exception conflict masking on the initial results of the job scenario semantic network to generate the germplasm bank job scenario semantic network; The node status writing includes writing the current job status, task execution status, resource usage status, environmental monitoring status, and abnormal event persistence status to the corresponding nodes. The edge status writing includes writing the association validity marker, time consistency marker, and spatial consistency marker to the corresponding edges. The abnormal conflict masking includes attaching a conflict masking identifier to the nodes or edges carrying conflict markers and retaining the original index records.

[0024] In this embodiment, the generation of the fuzzy time scene description result specifically includes: Based on the semantic network of germplasm bank operation scenarios, extract the status of task time-related nodes, resource status-related nodes, environment nodes, and abnormal event nodes to generate a fuzzy time compilation input set. The task time-related node status includes the time sequence status corresponding to the planned start time, planned end time, task execution status, and preceding task identifier. The resource status-related node status includes the time sequence status corresponding to the resource occupancy status, current load, and status timestamp. The environment node status includes the continuous monitoring status corresponding to the environmental monitoring status. The abnormal event node status includes the event time sequence status corresponding to the abnormal event occurrence time and abnormal event duration. The extraction process encapsulates each node status according to the time grid index to obtain the fuzzy time compilation input set. Based on the fuzzy time compilation input set, the job task execution chain is parsed for time window, generating task time window results and remaining buffer duration results, and then written back to the corresponding job task node according to the task identifier; The time window parsing process reads the planned start time, planned end time, completion status of previous tasks, and current execution status for each job task, determines the task start time interval, the task completion time interval, and the current waiting time interval, and generates the remaining buffer duration result based on the difference between the current time grid index and the task completion time interval. Based on the task time window results, remaining buffer duration results, and resource status related node status execution time pressure, a time compression result is generated; The time pressure merging process reads the remaining buffer time, the waiting status of the preceding tasks, the current resource occupancy status, and the current load for each task. It then performs merging in the same direction according to the direction of buffer time tightening, the direction of preceding task blocking, and the direction of resource waiting accumulation. The merging results are then mapped to a level to obtain the time compression degree of the corresponding task. Based on the environmental node status, abnormal event node status, and germplasm sample node status, risk status compilation is performed to generate environmental deviation status results, resource busy status results, and event persistence status results. The risk status compilation reads temperature, humidity, vibration, access control status, area congestion status, and equipment alarm status for each work area, and generates environmental deviation status results based on the preset area operation baseline. For each resource node, it reads resource occupancy status, current load, number of bound tasks, and number of queued tasks, and generates resource busy status results based on the preset resource load classification table. For each abnormal event node, it reads the time of occurrence of the abnormal event, the duration of the abnormal event, and the object affected by the abnormal event, and generates event duration status results based on the event duration interval and event impact continuation rules. Based on the results of time compression, environmental deviation, resource busyness, event persistence, and the storage temperature zone, activity level, and current operation status of germplasm sample nodes, the scenario stage mapping is generated. The task time window result, remaining buffer duration result, time compression degree result, environmental deviation state result, resource busy state result, event persistence state result, and scene stage result are uniformly encapsulated to generate fuzzy time scene description results; The unified encapsulation process consistently organizes the results corresponding to the same task identifier according to the time grid index, task target identifier, resource identifier, and job area code. It adds a masking inheritance flag to the node results with an anomaly conflict masking flag and adds a valid compilation flag to the node results without an anomaly conflict masking flag.

[0025] In this embodiment, the scenario phase mapping divides the job chain state corresponding to each task object into phases based on the combined hit relationship of task time pressure, regional environment deviation, resource load accumulation, event continuous impact and sample job sensitivity. These phases include stable phase, tense phase, disturbance phase and recovery phase. The scenario phase results are then jointly bound with task identifier, resource identifier, job area code and time grid index. Specifically, when the time pressure is low, the environmental deviation is low, the resource load is low or medium, and the event's continuous impact is no continuous impact or low continuous impact, it is judged as a stable phase. When any of the four states is high, and the corresponding target is a critical task, critical resource, or highly sensitive sample, or when the work chain corresponding to the highly sensitive sample simultaneously hits the time compression result increase, the environmental deviation result increase, and the event's continuous state result continue to exist, it is judged as a disturbance phase. When any of the four states is high, and the disturbance phase conditions are not met, or when the time pressure is medium, the resource load is medium or high, the event's continuous impact is medium, or the remaining buffer time result corresponding to the highly sensitive sample enters the warning interval, it is judged as a tense phase. When the current time grid index corresponds to a tense phase or a disturbance phase, and at least two of the current four states are decreasing, the disturbance phase conditions are not met, and the highly sensitive sample is not in a high-risk continuous state, it is judged as a recovery phase. When the same work chain simultaneously meets multiple phase conditions, priority is given to the disturbance phase, recovery phase, tense phase, and stable phase.

[0026] In this embodiment, the generation of propagation semantic relationship results specifically includes: Based on the semantic network of germplasm bank operation scenario, extract each node set and sample-task association edge, task-resource requirement edge, resource-region affiliation edge, environment-region effect edge and abnormal event-object impact edge, generate propagation relationship to construct input set; The input set for constructing propagation relationships is aligned and organized according to the task identifier, task target identifier, resource identifier, work area code, and time grid index. Nodes or edges with attached abnormal conflict shielding identifiers are written with propagation shielding markers, while nodes or edges without attached abnormal conflict shielding identifiers are written with propagation valid markers. Based on the propagation relationship, the task time window result, remaining buffer duration result, time compression degree result, and scene stage result in the input set and fuzzy time scene description result are constructed. Time compression candidate propagation pairs are constructed, and the initial result of time compression propagation edge is generated. The time compression candidate propagation pairs are formed by combining the preceding and succeeding task nodes corresponding to the same task object identifier, parallel task nodes sharing the same resource node, and task nodes located in the same work area with overlapping time windows. A directed propagation relationship is established from front to back, from busy to waiting, and from tight to slow for task nodes whose remaining buffer time is shortened, whose time compression degree is increased, or whose scenario stage is in a tense or disturbed stage, to obtain the initial result of the time compression propagation edge. Based on the propagation relationship, construct the resource busy status result, scenario stage result, and current load, number of bound tasks, and number of queued tasks corresponding to the resource node from the input set and fuzzy time scenario description result. Construct resource congestion candidate propagation pairs and generate initial results of resource congestion propagation edges. The resource congestion candidate propagation pair is formed by combining task nodes that share the same resource node, resource nodes that belong to the same job area code or adjacent job area codes, and task nodes that are bound to the same key resource through the task-resource demand edge. A directed propagation relationship is established from the congestion source node to the affected node for nodes whose resource busy status results increase, current load accumulates, or the number of queued tasks increases, so as to obtain the initial result of the resource congestion propagation edge. Based on the propagation relationship, construct the input set and the environmental deviation state result, event persistence state result, scene stage result, as well as the object of abnormal event, environmental node state and current operation state in the fuzzy time scene description result, construct the abnormal diffusion candidate propagation pair, and generate the initial result of abnormal diffusion propagation edge. The anomaly propagation candidate pair is formed by combining the anomaly event node with its corresponding object node, the environment node with the resource node within its assigned work area code, the environment node with the work task node within the same work area code, and the anomaly event node with the work chain node corresponding to the object identifier of the same task. A directed propagation relationship is established from the anomaly source node to the affected node for nodes where the event's persistent state result persists, the environment's deviation state result increases, or the scenario stage result is in a disturbance stage, thus obtaining the initial result of the anomaly propagation edge. Perform edge direction correction, edge validity screening, and edge conflict resolution on the initial results of the time compression propagation edge, the initial results of the resource congestion propagation edge, and the initial results of the abnormal diffusion propagation edge to generate time compression propagation edge, resource congestion propagation edge, and abnormal diffusion propagation edge; The edge direction correction is based on the order of tasks, the direction of resource load accumulation, the direction of impact of the work area, and the direction of action of abnormal events to determine the propagation direction. The edge validity screening is based on the propagation validity mark, time consistency mark, and spatial consistency mark to remove invalid edges. The edge conflict resolution retains candidate edges with the same source node and target node but different propagation types in the order of abnormal diffusion propagation edge, resource congestion propagation edge, and time compression propagation edge. Perform unified association encapsulation on time-compression propagation edges, resource congestion propagation edges, and anomaly diffusion propagation edges to generate propagation semantic relationship results.

[0027] In this embodiment, the generation of scene phase state results, phase transition direction results, scheduling stability results, and rescheduling trigger results specifically includes: The fuzzy temporal scene description results and propagated semantic relationship results are input into the fact compilation unit of the phase transition enhanced FT-SWRL semantic reasoning network to perform fact compilation and generate a set of reasoning facts. The fact compilation unit performs unified alignment of task time window results, remaining buffer duration results, time compression degree results, environmental deviation state results, resource busy state results, event persistence state results, scenario stage results, as well as time compression propagation edge, resource congestion propagation edge, and anomaly propagation propagation edge according to task identifier, task target identifier, resource identifier, job area code, and time grid index. It compiles and generates time fact set, environmental fact set, resource fact set, event fact set, stage fact set, and propagation fact set respectively, and encapsulates them into a unified reasoning fact set. The set of reasoning facts is input into the rule mapping unit to generate a set of rule-acting objects and a set of rule-hitting candidates; The rule mapping unit performs item-by-item matching with the task node set, resource node set, environmental node set, event fact set, phase fact set, and propagation fact set, as well as the task chain node set corresponding to the same task object identifier. It binds the hit time compression judgment rule, resource congestion judgment rule, anomaly propagation judgment rule, scenario phase candidate rule, and trigger adjudication rule to the task node, resource node, work area node, and work chain node, respectively. Based on the time compression propagation edge coverage, resource congestion propagation edge coverage, and anomaly propagation edge coverage, it merges the hit relationships of each rule object to generate a rule hit candidate set. In the phase transition candidate inference unit, phase transition candidate inference is performed based on the rule action object set and the rule hit candidate set to generate scene phase candidate results and basic results for transition determination. The phase transition candidate reasoning unit reads the corresponding time fact set, environmental fact set, resource fact set, event fact set, stage fact set, and propagation fact set for each rule application object, generates scene phase candidate results according to the hit relationship of stable stage, tense stage, disturbance stage, and recovery stage, and generates migration judgment basis results based on the increase and decrease relationship of each stage candidate under the current time grid index and the propagation direction change relationship in the propagation fact set. The scene phase candidate results and the basic results of the migration decision are input into the migration adjudication unit to generate the scene phase results and the phase migration direction results; The migration adjudication unit prioritizes the candidate phase results of the scene corresponding to the same rule object in the order of disturbance phase, recovery phase, tension phase, and stable phase to obtain the scene phase result under the current time grid index. It combines the scene phase result corresponding to the previous time grid index and the migration judgment basis result to determine the phase migration direction result, including stable to tension, tension to disturbance, disturbance to recovery, recovery to stable, and phase maintenance. Based on the scenario phase state results, phase state transition direction results, and propagation fact set, a stability assessment is performed, and scheduling stability results are generated. The stability assessment involves reading the scenario phase results and task execution status according to the time grid index for each rule-applied job chain, counting the task segments that maintain the original task order without interruption, rollback, or cross-chain switching to obtain the task continuous execution maintenance result, reading the resource occupancy status, current load, number of bound tasks, and number of queued tasks of critical resources under adjacent time grid indices, counting the number of resource status changes and duration segments to obtain the critical resource fluctuation result, sequentially scanning the task nodes and resource nodes covered by time compression propagation edges, resource congestion propagation edges, and abnormal diffusion propagation edges, counting the number of hits, consecutive hit lengths, and path offsets on the main execution path to obtain the critical path disturbance result, reading the environmental deviation status results and event duration results of high-sensitivity samples during the warm-up task, detection task, and re-entry storage task, counting the number of hits and duration segments of the three types of propagation edges to obtain the high-sensitivity sample impact result, and jointly adjudicating the task continuous execution maintenance result, critical resource fluctuation result, critical path disturbance result, and high-sensitivity sample impact result to obtain the scheduling stability result; The joint decision is as follows: when the task is continuously executed and the result is maintained and the other three items do not reach the high disturbance state, it is judged as high stability; when the task is continuously executed and the result decreases and any two of the other three items reach the medium disturbance state, it is judged as medium stability; when the task is continuously executed and the result is interrupted or any one of the other three items reaches the high disturbance state and persists, it is judged as low stability. Based on the trigger decision unit, trigger decision is executed on the scene phase state result, phase state transition direction result, scheduling stability result and event fact set to generate rescheduling trigger result; The triggering adjudication unit writes a rescheduling trigger flag to the rule application object that satisfies the scenario phase state result of being in the tense phase or the disturbance phase, the phase state migration direction result of being in the forward migration or the phase maintenance, and the scheduling stability result of being in the medium or low stability. It writes a suppression trigger flag to the rule application object that satisfies the scenario phase state result of being in the recovery phase and the scheduling stability result of being in the continuous recovery phase.

[0028] In this embodiment, the phase transition enhanced FT-SWRL semantic reasoning network includes a fact compilation unit, a rule mapping unit, a phase transition candidate reasoning unit, a transition adjudication unit, and a trigger adjudication unit.

[0029] In this embodiment, the generation of the phase change scheduling cone specifically includes: Based on the scene phase state results, phase transition direction results, scheduling stability results, and rescheduling trigger results, perform cone generation input alignment to generate a phase change scheduling cone generation input set; The cone generation input alignment is performed according to the task identifier, task target identifier, resource identifier, job area code and time grid index. The scene phase state result, phase transition direction result, scheduling stability result and rescheduling trigger result are uniformly aligned. The job task nodes, resource nodes, job area nodes and job chain nodes corresponding to the same rule target are merged and organized, and the trigger type, trigger level and current scene phase state mark are written to obtain the phase change scheduling cone generation input set. Based on the phase change scheduling cone, the input set is generated to perform scheduling domain partitioning, and task scheduling domain results, resource scheduling domain results, path scheduling domain results, and critical task protection results are generated. The scheduling domain division reads the scene phase state result, phase state transition direction result, and rescheduling trigger result for each rule-applied object. For rule-applied objects in the stable phase, only the task scheduling domain result corresponding to the local reordering is opened. For rule-applied objects in the tense phase, the task scheduling domain result and resource scheduling domain result are opened. For rule-applied objects in the disturbance phase, the task scheduling domain result, resource scheduling domain result, and path scheduling domain result are opened. For rule-applied objects in the recovery phase, the restricted task scheduling domain result is retained and path expansion is suppressed. At the same time, a prohibition on disturbance flag is written for critical tasks, tasks corresponding to highly sensitive samples, and tasks associated with critical resources to obtain critical task protection results. Based on the results of task scheduling domain, resource scheduling domain, path scheduling domain, critical task protection, and semantic network execution object selection for germplasm bank operation scenarios, adjustable task subsets, alternative resource subsets, and correctable path subsets are generated. The object screening process involves screening job task nodes within the task scheduling domain coverage area based on task execution status, task priority, task target identifier, and preceding task identifier. Job task nodes carrying a prohibition on disturbance, a completed status, or an uninterruptible status are removed to obtain an adjustable task subset. Resource nodes within the resource scheduling domain coverage area are screened based on service capacity, availability, assigned work area, and current load. Resource nodes in an unavailable state, those with persistent critical resource occupancy status, and carrying a prohibition on disturbance are removed to obtain a substitute resource subset. The path scheduling domain coverage area involves screening channel segments, work area connection relationships, and the impact range of abnormal diffusion propagation edges. Path nodes and edges with propagation shielding markers, invalid spatial consistency markers, or persistent abnormal diffusion are removed to obtain a correctable path subset. Based on the scene phase state results, phase transition direction results, scheduling stability results, and critical task protection results, cone boundary adjudication is performed to generate cone boundary results and cone stability results; The cone boundary adjudication process involves statistically analyzing the number of tasks in the adjustable task subset, the number of resources in the alternative resource subset, and the number of path nodes in the correctable path subset for each rule-applied object. It then combines this with the number of open scheduling domains corresponding to the scenario phase result, the migration intensity corresponding to the phase migration direction result, and the stability level corresponding to the scheduling stability result to determine the task boundary, resource boundary, and path boundary of the current rule-applied object. When the prohibition of disturbance marker covers the task corresponding to a critical task or a highly sensitive sample, the task boundary and path boundary are contracted to obtain the cone boundary result. Based on the boundary change magnitude and the boundary duration segment, the cone stability result is generated. Based on the results of task scheduling domain, resource scheduling domain, path scheduling domain, critical task protection, cone boundary, and cone stability, a unified encapsulation and consistency check are performed to generate a phase change scheduling cone; The consistency check examines the results of the task scheduling domain, resource scheduling domain, path scheduling domain, and critical task protection item by item. When the adjustable task subset, alternative resource subset, and correctable path subset are all within the corresponding boundary range and have not exceeded the critical task protection result corresponding to the prohibition of disturbance, the cone constraint valid mark is written. When any object exceeds the corresponding boundary range or conflicts with the critical task protection result, boundary rollback is performed and re-encapsulation is performed.

[0030] In this embodiment, the generation of the adaptive scheduling scheme specifically includes: Based on the phase change scheduling cone, rescheduling trigger results, adjustable task subsets, alternative resource subsets, and correctable path subsets, perform rescheduling input alignment to generate a rescheduling input set; The rescheduling input alignment is based on the task identifier, task target identifier, resource identifier, job area code, and time grid index. The trigger type, trigger level, and rule target in the rescheduling trigger result are uniformly matched with the task scheduling domain result, resource scheduling domain result, path scheduling domain result, critical task protection result, cone boundary result, and cone constraint validity mark in the phase change scheduling cone to obtain the rescheduling input set. Based on the rescheduled input set, local reordering is performed on the adjustable task subset under the constraints of the task scheduling domain results to generate local reordering candidate results; The local reordering process rearranges the task nodes in the adjustable task subset according to the trigger level, task priority level, remaining buffer time, preceding task identifier, and scene phase result. It keeps the prohibition of disturbance flags corresponding to the critical task protection results unchanged, keeps the order corresponding to the preceding task identifiers intact, and writes the task sequence after the reordering into the local reordering candidate results. When the rule target corresponding to the local reorder candidate result still has a rescheduling trigger flag, resource substitution is performed on the substitutable resource subset under the resource scheduling domain result constraint to generate resource substitution candidate results; Resource substitution is based on service capacity, availability, work area, current load, and critical resource occupancy status. For work task nodes in the local reorder candidate results that have not eliminated the trigger mark, resource nodes in the subset of replaceable resources are reallocated. Resource nodes within the same work area code are retained first, and then resource nodes within adjacent work area codes that do not carry the prohibition of disturbance mark are selected to obtain the resource substitution candidate results. When the rule object corresponding to the resource substitution candidate result still has a rescheduling trigger flag, path correction is performed on the correctable path subset under the constraint of the path scheduling domain result to generate path correction candidate results. The path correction process involves replacing task paths affected by congestion or continuous abnormal diffusion in the resource substitution candidate results according to the connection order of channel segments, the connection relationship of work areas, the impact range of abnormal diffusion propagation, and spatial consistency. Priority is given to retaining passable paths within the original work area, and then alternative paths between adjacent work areas that are not affected by continuous abnormal diffusion are selected to obtain the path correction candidate results. When the rule object corresponding to the path correction candidate result still has a rescheduling trigger flag, and the scene phase result is in the disturbance stage or the phase migration direction result is forward migration, perform critical task preemption on the adjustable task subset and the alternative resource subset, and generate preemption candidate results. The critical task preemption process involves pausing, shifting, or releasing resources for job task nodes that do not carry a prohibition on disturbance marker and whose task priority is lower than that of the triggering task. The released resource nodes are then redistributed to critical tasks, tasks corresponding to highly sensitive samples, or job task nodes with higher trigger levels. The task boundaries, resource boundaries, and path boundaries corresponding to the critical task protection results are not breached, thus obtaining preemption candidate results. A progressive decision is made on each candidate result in the order of local reordering, resource substitution, path correction, and priority of critical tasks to generate an adaptive scheduling scheme. In the progressive decision-making process, after each level of processing, the corresponding candidate results are checked for effective cone constraint marking. When the current level candidate result has eliminated the rescheduling trigger mark of the corresponding rule object and has not broken the cone boundary result, the subsequent level processing is stopped and the current level candidate result is written as an adaptive scheduling scheme. When the current level candidate result has not eliminated the rescheduling trigger mark or conflicts with the critical task protection result, the process enters the next processing level to continue rescheduling until an adaptive scheduling scheme that satisfies the phase change scheduling cone constraint is generated.

[0031] In this embodiment, the generation of the adaptive scheduling closed loop specifically includes: Based on an adaptive scheduling scheme, task assignment and status feedback collection are performed to generate an execution feedback input set. The task assignment and status feedback collection process writes the local reordering results, resource substitution results, path correction results, and critical task preemption results in the adaptive scheduling scheme into the corresponding job task nodes, resource nodes, and job area nodes, respectively. It also collects feedback on the task execution status, resource occupancy status, channel segment passage status, environmental monitoring status, and abnormal event persistence status after execution, and obtains the execution feedback input set. Based on the execution feedback input set, execution feedback regularization is performed to generate execution feedback results. The feedback normalization process unifies the task completion status, task delay status, task interruption status, resource release status, resource replacement status, path switching status, environment recovery status, and anomaly mitigation status in the execution feedback input set. It continuously aligns the job chain nodes corresponding to the same task object identifier according to the time grid index. It writes feedback anomaly flags to nodes with execution failure, status rollback, or missing receipts, and writes feedback valid flags to the remaining nodes to obtain the execution feedback result. Based on the execution feedback results, update the node and edge states in the semantic network of the germplasm bank operation scenario to generate semantic network update results; Based on the execution feedback results and semantic network update results, the propagation feedback write-back is performed to generate propagation relationship correction results; The propagation feedback writeback process involves rereading the remaining buffer duration, task execution status, and pre-task waiting status of the source and target nodes corresponding to the time compression propagation edge; rereading the resource occupancy status, current load, number of bound tasks, and number of queued tasks of the source and target nodes corresponding to the resource congestion propagation edge; rereading the environmental deviation status, event persistence status, and current job status of the source and target nodes corresponding to the abnormal diffusion propagation edge; writing a propagation failure flag for propagation edges that no longer meet the propagation conditions; retaining a propagation validity flag for propagation edges that continuously meet the propagation conditions; and rewriting the propagation direction for propagation edges whose propagation direction has reversed. The phase transition enhanced FT-SWRL semantic inference network is updated based on the execution feedback results, semantic network update results, and propagation relation correction results, and the inference network update results are generated. The update phase transition enhanced FT-SWRL semantic reasoning network includes recompiling the time fact set, environment fact set, resource fact set, event fact set, stage fact set, and propagation fact set in the fact compilation unit; rematching the rule action object set and rule hit candidate set in the rule mapping unit; recalculating the scene phase state candidate results and migration decision base results in the phase transition candidate reasoning unit; re-deciding the scene phase state results and phase migration direction results in the migration decision unit; and rewriting the rescheduling trigger results in the trigger decision unit. Based on the semantic network update results, propagation relationship correction results, and inference network update results, a closed-loop consistency check is performed to form an adaptive scheduling closed loop. The closed-loop consistency check compares the node and edge states in the semantic network of the germplasm bank operation scenario, the propagation validity and failure markers in the propagation relationship correction results, and the scenario phase state results, phase transition direction results, scheduling stability results, and rescheduling trigger results in the phase transition enhanced FT-SWRL semantic reasoning network. When the task execution state, resource occupancy state, propagation relationship state, and rescheduling trigger results are consistent under the same time grid index, a closed-loop validity marker is written. When there are state conflicts, propagation conflicts, or trigger conflicts, a closed-loop exception marker is written, and the corresponding task identifier, resource identifier, and job area code are returned to the next round of adaptive scheduling processing.

[0032] Example 1: To verify the feasibility of this invention in practice, it was applied to a combined operation environment of a national-level germplasm resource conservation site, encompassing a low-temperature storage area, a buffer storage area, a testing area, and a transport channel. The application occurred during the spring period of concentrated germplasm sample entry and verification, at a germplasm resource management site in East China responsible for long-term preservation, periodic testing, and anomaly handling. In actual operation, this site simultaneously handles sample entry, temperature recovery testing, verification and storage, and inter-regional transport, which easily leads to overlapping task time windows, fluctuating resource usage, channel congestion, and the superposition of environmental anomalies. Especially after highly sensitive samples enter the testing chain, traditional methods relying on manual experience and fixed schedules struggle to promptly identify disruptions to the operational chain, often resulting in localized delays propagating to subsequent stages, critical resources being occupied by low-priority tasks, and the escalating impact of abnormal events.

[0033] In the application process, germplasm sample data, task data, resource status data, environmental monitoring data, and abnormal event data are first collected to construct a semantic network for the germplasm bank's operational scenarios. Then, fuzzy temporal compilation is performed on the semantic network to generate task time window results, remaining buffer duration results, environmental deviation status results, resource busy status results, and event persistence status results. Subsequently, time compression propagation edges, resource congestion propagation edges, and abnormal diffusion propagation edges are generated based on the semantic network. The fuzzy temporal scenario description results and propagation semantic relationship results are input into a phase transition enhanced FT-SWRL semantic inference network to obtain scenario phase state results, phase transition direction results, scheduling stability results, and rescheduling trigger results. Afterwards, depending on whether the current scenario is in a stable, tense, disturbed, or recovery phase, adjustable task subsets, alternative resource subsets, and correctable path subsets are extracted to generate a phase transition scheduling cone. Within constraints, local reordering, resource substitution, path correction, and critical task preemption are executed sequentially to output an adaptive scheduling scheme. After execution, the system continues to collect data on task execution status, resource usage status, environment recovery status, and anomaly mitigation status, and updates the semantic network, propagation relationships, and inference network synchronously to form a continuously rolling adaptive scheduling closed loop.

[0034] The operation logs, environmental records, channel receipts, and scheduling receipts accumulated during continuous operation at the site demonstrate that, with this invention, time-pressured work chains can be identified earlier. Resource congestion and anomaly propagation are no longer passively exposed only in subsequent executions but are pre-constrained during the propagation phase. The continuous execution capability of tasks corresponding to key samples is significantly enhanced, additional disturbances caused by cross-regional transfers are suppressed, and the impact range of continuous environmental deviations on the detection and replication chains is significantly reduced. Scheduling adjustments are more concentrated in the affected areas rather than spreading globally. Combining semantic network records, fuzzy time compilation records, propagation relationship records, phase adjudication records, and execution feedback records formed under application time and location, it can be proven that this invention can effectively solve the scheduling instability problem caused by the superposition of time compression, resource congestion, and anomaly propagation in germplasm bank scenarios, possessing the beneficial effects of timely response, controllable disturbances, and strong closed-loop update capabilities.

[0035] Table 1. Comparison of the overall effects of adaptive scheduling in germplasm bank operation scenarios.

[0036] As shown in Table 1, under the conditions of multi-task parallelism in the germplasm bank, dynamic resource fluctuations, and continuous disturbances from abnormal events, the method of this invention exhibits better overall performance across multiple key indicators. Taking the on-time completion rate of high-priority tasks as an example, the traditional fixed-rule scheduling method achieves 88.6%, the scheduling method based on ordinary knowledge graphs improves to 92.4%, and this invention further improves it to 96.3%. This change indicates that while relying solely on static rules or general knowledge graph associations can improve some task ranking issues, it still falls short in continuously perceiving task time pressure. This invention, through fuzzy temporal compilation, propagation of semantic relationship results, and a phase-change enhanced FT-SWRL semantic reasoning network, can identify time compression trends before tasks actually time out, thereby implementing scheduling intervention earlier. Therefore, it has a stronger ability to guarantee the timeliness of high-priority tasks.

[0037] The average task waiting time decreased significantly from 27.8 minutes to 14.9 minutes. This is because the invention does not blindly reorder tasks globally. Instead, it first extracts adjustable task subsets, alternative resource subsets, and correctable path subsets based on scenario phase state results, phase transition direction results, and scheduling stability results. Then, it generates a phase transition scheduling cone to constrain the scheduling feasible region. This ensures that affected tasks are prioritized for processing within the correctable range, reducing the repeated adjustments of irrelevant tasks and lowering the probability of local congestion propagating globally. Therefore, the task waiting chain is significantly shortened.

[0038] In terms of the number of resource conflicts and the duration of channel congestion, the method of this invention reduces the number of times per day and the duration of channel congestion to 6 minutes per day, respectively, both by nearly half compared to traditional methods. This demonstrates that the present invention is more targeted in handling resource competition and path interference. Traditional fixed-rule methods often can only passively adjust after a conflict occurs. Although scheduling methods based on ordinary knowledge graphs can identify some relationships, their granularity in describing congestion propagation and anomaly diffusion is still insufficient. The present invention generates time-compressed propagation edges, resource congestion propagation edges, and anomaly diffusion propagation edges, so that resource conflicts and path disturbances are no longer single-point information, but propagation semantics that can be continuously transmitted along the work chain and identified in advance. Therefore, scheduling actions can intervene in the congestion formation process earlier, thereby significantly shortening the duration of congestion.

[0039] The number of tasks affected by the spread of abnormal events decreased from 11 per day to 4 per day, the number of critical operation chain interruptions decreased from 9 per week to 2 per week, and the number of timeouts for highly sensitive sample operations decreased from 13 per month to 3 per month. These data collectively demonstrate that the present invention significantly enhances the disturbance suppression capability in highly constrained operational environments. This is because the present invention incorporates the persistent state of abnormal events, the environmental deviation state, and the disrupted state of the operation chain into the phase transition enhanced FT-SWRL semantic reasoning network. It not only determines whether an anomaly has occurred but also further determines whether the anomaly is spreading from a local area to the critical link, and determines the processing intensity based on whether the scenario is in a stable, tense, disturbed, or recovery phase. Particularly for tasks related to highly sensitive samples, the present invention avoids ordinary tasks crowding out the critical link through critical task protection results and boundary constraints of the phase transition scheduling cone. Therefore, the continuity of the critical chain and the timeliness of samples are more stably guaranteed.

[0040] The rescheduling response time was reduced from 46.5 seconds to 19.7 seconds, demonstrating the invention's high real-time response capability in dynamic scenarios. Traditional methods often rely on manual intervention or fixed threshold triggers for task adjustments, limiting processing speed. While ordinary knowledge graph methods can assist in judgment, they are prone to problems such as excessively large filtering ranges and too many scheduling candidates when multiple states overlap. This invention introduces a progressive event rescheduling mechanism, processing layer by layer in the order of local reordering, resource substitution, path correction, and critical task preemption. After each level is completed, the effective marking of the cone constraint is reviewed. Only when the current level is insufficient to eliminate the rescheduling trigger mark will the next level be entered. Therefore, scheduling correction can be completed quickly with a small computational cost, demonstrating good response efficiency.

[0041] The number of invalid cross-regional transfers decreased from 21 times / week to 7 times / week, and the proportion of high scheduling stability increased from 61.8% to 88.1%, further demonstrating that this invention not only improves scheduling speed but also significantly improves scheduling quality. The reduction in invalid transfers indicates more accurate path correction, and the increased proportion of high scheduling stability indicates that the scheduling results are not based on short-term effects obtained through frequent rescheduling, but rather maintain better continuous stability throughout the overall execution process. The fundamental reason is that this invention uses scene phase results, phase transition direction results, and scheduling stability results as prerequisites for scheduling domain generation and progressive event rescheduling, enabling the scheduling process to balance local agile adjustments with global stability constraints, avoiding over-scheduling and redundant scheduling.

[0042] As can be seen from all the data in Table 1, this invention can achieve early identification of the work chain status, effective suppression of propagation disturbances, and limited and gradual adjustment of scheduling actions under conditions of concurrent multi-source dynamic information in germplasm banks, compressed time windows, enhanced resource competition, and the continuous existence of abnormal events. This results in a more balanced improvement in terms of task timeliness, resource coordination, path accessibility, anomaly control, and overall stability, demonstrating comprehensive advantages such as timely response, controllable disturbances, strong critical link protection capabilities, and high closed-loop adaptive level.

[0043] The above are merely preferred embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. An adaptive scheduling method for germplasm bank operation scenarios based on semantic networks, characterized in that, Includes the following steps: Acquire germplasm sample data, task data, resource status data, environmental monitoring data, and abnormal event data to construct a semantic network for germplasm bank operation scenarios; Fuzzy temporal compilation is performed on the semantic network of germplasm bank operation scenarios to generate fuzzy temporal scenario description results; Based on the semantic network of germplasm bank operation scenarios, propagation semantic relationship results are formed by generating time compression propagation edges, resource congestion propagation edges, and abnormal diffusion propagation edges. The fuzzy temporal scene description results and propagation semantic relationship results are input into the phase transition enhanced FT-SWRL semantic reasoning network to generate scene phase state results, phase state transition direction results, scheduling stability results, and rescheduling trigger results. Based on the scenario phase state results, phase transition direction results, and scheduling stability results, extract adjustable task subsets, alternative resource subsets, and correctable path subsets, and generate a phase change scheduling cone; Under the phase change scheduling cone constraint, an adaptive scheduling scheme is generated by progressively rescheduling the adjustable task subset, the alternative resource subset, and the correctable path subset based on the rescheduling triggering result. The adaptive scheduling scheme is executed and the execution feedback results are collected. The semantic network of germplasm bank operation scenario and phase transition enhanced FT-SWRL semantic reasoning network are updated to form an adaptive scheduling closed loop. The phase transition enhanced FT-SWRL semantic reasoning network includes a fact compilation unit, a rule mapping unit, a phase transition candidate reasoning unit, a transition adjudication unit, and a trigger adjudication unit; The generation of the phase change scheduling cone specifically includes: Based on the scene phase state results, phase transition direction results, scheduling stability results, and rescheduling trigger results, perform cone generation input alignment to generate a phase change scheduling cone generation input set; Based on the phase change scheduling cone, the input set is generated to perform scheduling domain partitioning, and task scheduling domain results, resource scheduling domain results, path scheduling domain results, and critical task protection results are generated. Based on the results of task scheduling domain, resource scheduling domain, path scheduling domain, critical task protection, and semantic network execution object selection for germplasm bank operation scenarios, adjustable task subsets, alternative resource subsets, and correctable path subsets are generated. Based on the scene phase state results, phase transition direction results, scheduling stability results, and critical task protection results, cone boundary adjudication is performed to generate cone boundary results and cone stability results; Based on the results of task scheduling domain, resource scheduling domain, path scheduling domain, critical task protection, cone boundary, and cone stability, a unified encapsulation and consistency verification are performed to generate a phase change scheduling cone.

2. The adaptive scheduling method for germplasm bank operation scenarios based on semantic networks according to claim 1, characterized in that, The generation of the semantic network for the germplasm bank operation scenario specifically includes: Acquire germplasm sample data, extract fields and standardize formats to generate a set of germplasm sample entries; Acquire and organize the task data to generate a set of task entries; Acquire resource status data and encode it uniformly to generate a set of resource status entries; Acquire environmental monitoring data and abnormal event data, perform time alignment and spatial attribution mapping, and generate a set of environmental event entries; Based on the set of germplasm sample entries, the set of job task entries, the set of resource status entries, and the set of environmental event entries, perform primary key association and consistency verification to generate a set of cross-source association results. Initial results of the semantic network for the job scenario are constructed based on the cross-source association result set; The initial results of the job scenario semantic network are processed by writing node states, edge states, and masking of abnormal conflicts to generate the germplasm bank job scenario semantic network.

3. The adaptive scheduling method for germplasm bank operation scenarios based on semantic networks according to claim 1, characterized in that, The generation of the fuzzy temporal scene description result specifically includes: Based on the semantic network of germplasm bank operation scenarios, extract the status of task time-related nodes, resource status-related nodes, environment nodes, and abnormal event nodes to generate a fuzzy time compilation input set. Based on the fuzzy time compilation input set, the job task execution chain is parsed for time window, generating task time window results and remaining buffer duration results, and then written back to the corresponding job task node according to the task identifier; Based on the task time window results, remaining buffer duration results, and resource status related node status execution time pressure, a time compression result is generated; Based on the environmental node status, abnormal event node status, and germplasm sample node status, risk status compilation is performed to generate environmental deviation status results, resource busy status results, and event persistence status results. Based on the results of time compression, environmental deviation, resource busyness, event persistence, and the storage temperature zone, activity level, and current operation status of germplasm sample nodes, the scenario stage mapping is generated. The task time window result, remaining buffer duration result, time compression degree result, environment deviation state result, resource busy state result, event persistence state result, and scene stage result are uniformly encapsulated to generate fuzzy time scene description results.

4. The adaptive scheduling method for germplasm bank operation scenarios based on semantic networks according to claim 3, characterized in that, The scenario phase mapping divides the job chain state corresponding to each task object into phases based on the combined hit relationship of task time pressure, regional environmental deviation, resource load accumulation, event continuous impact and sample job sensitivity. These phases include stable phase, tense phase, disturbance phase and recovery phase. The scenario phase results are then jointly bound with task identifier, resource identifier, job area code and time grid index.

5. The adaptive scheduling method for germplasm bank operation scenarios based on semantic networks according to claim 1, characterized in that, The generation of the propagation semantic relationship results specifically includes: Based on the semantic network of germplasm bank operation scenario, extract each node set and sample-task association edge, task-resource requirement edge, resource-region affiliation edge, environment-region effect edge and abnormal event-object impact edge, generate propagation relationship to construct input set; Based on the propagation relationship, the task time window result, remaining buffer duration result, time compression degree result, and scene stage result in the input set and fuzzy time scene description result are constructed. Time compression candidate propagation pairs are constructed, and the initial result of time compression propagation edge is generated. Based on the propagation relationship, construct the resource busy status result, scenario stage result, and current load, number of bound tasks, and number of queued tasks corresponding to the resource node from the input set and fuzzy time scenario description results. Construct resource congestion candidate propagation pairs and generate initial results of resource congestion propagation edges. Based on the propagation relationship, construct the input set and the environmental deviation state result, event persistence state result, scene stage result, as well as the object of abnormal event, environmental node state and current operation state in the fuzzy time scene description result, construct the abnormal diffusion candidate propagation pair, and generate the initial result of abnormal diffusion propagation edge. Perform edge direction correction, edge validity screening, and edge conflict resolution on the initial results of the time compression propagation edge, the initial results of the resource congestion propagation edge, and the initial results of the abnormal diffusion propagation edge to generate time compression propagation edge, resource congestion propagation edge, and abnormal diffusion propagation edge; Perform unified association encapsulation on time-compression propagation edges, resource congestion propagation edges, and anomaly diffusion propagation edges to generate propagation semantic relationship results.

6. The adaptive scheduling method for germplasm bank operation scenarios based on semantic networks according to claim 1, characterized in that, The generation of the scenario phase state result, phase transition direction result, scheduling stability result, and rescheduling trigger result specifically includes: The fuzzy temporal scene description results and propagated semantic relationship results are input into the fact compilation unit of the phase transition enhanced FT-SWRL semantic reasoning network to perform fact compilation and generate a set of reasoning facts. The set of reasoning facts is input into the rule mapping unit to generate a set of rule-acting objects and a set of rule-hitting candidates; In the phase transition candidate inference unit, phase transition candidate inference is performed based on the rule action object set and the rule hit candidate set to generate scene phase candidate results and basic results for transition determination. The scene phase candidate results and the basic results of the migration decision are input into the migration adjudication unit to generate the scene phase results and the phase migration direction results; Based on the scenario phase state results, phase state transition direction results, and propagation fact set, a stability assessment is performed, and scheduling stability results are generated. Based on the trigger decision unit, trigger decisions are executed on the scene phase state result, phase state transition direction result, scheduling stability result and event fact set to generate rescheduling trigger result.

7. The adaptive scheduling method for germplasm bank operation scenarios based on semantic networks according to claim 1, characterized in that, The generation of the adaptive scheduling scheme specifically includes: Based on the phase change scheduling cone, rescheduling trigger results, adjustable task subsets, alternative resource subsets, and correctable path subsets, perform rescheduling input alignment to generate a rescheduling input set; Based on the rescheduled input set, local reordering is performed on the adjustable task subset under the constraints of the task scheduling domain results to generate local reordering candidate results; When the rule object corresponding to the local reorder candidate result still has a rescheduling trigger flag, resource substitution is performed on the substitutable resource subset under the resource scheduling domain result constraint to generate resource substitution candidate results; When the rule target corresponding to the resource substitution candidate result still has a rescheduling trigger flag, path correction is performed on the correctable path subset under the constraint of the path scheduling domain result to generate path correction candidate results. When the rule object corresponding to the path correction candidate result still has a rescheduling trigger flag, and the scenario phase result is in the disturbance stage or the phase migration direction result is forward migration, perform critical task preemption on the adjustable task subset and the alternative resource subset, and generate preemption candidate results. A progressive decision-making process is performed on each candidate result in the order of local reordering, resource substitution, path correction, and priority of critical tasks to generate an adaptive scheduling scheme.

8. The adaptive scheduling method for germplasm bank operation scenarios based on semantic networks according to claim 1, characterized in that, The generation of the adaptive scheduling closed loop specifically includes: Based on an adaptive scheduling scheme, task assignment and status feedback collection are performed to generate an execution feedback input set. Based on the execution feedback input set, execution feedback regularization is performed to generate execution feedback results. Based on the execution feedback results, update the node and edge states in the semantic network of the germplasm bank operation scenario to generate semantic network update results; Based on the execution feedback results and semantic network update results, the propagation feedback write-back is performed to generate propagation relationship correction results; The phase transition enhanced FT-SWRL semantic inference network is updated based on the execution feedback results, semantic network update results, and propagation relation correction results, and the inference network update results are generated. Based on the semantic network update results, propagation relationship correction results, and inference network update results, a closed-loop consistency check is performed to form an adaptive scheduling closed loop.

Images