A water conservancy multi-professional model pluggable coupling scheduling method and system

Abstract

The application provides a water conservancy multi-professional model pluggable coupling scheduling method and system, relating to the technical field of water conservancy digital twin. The method encapsulates models through unified interface specifications and parameter mapping rules, constructs model linkage topologies according to simulation tasks, and selects coupling modes according to accuracy requirements, response delay requirements and computing power occupation; calls models according to basic working condition data, standardizes and adapts the pre-sequenced output to meet the input format, unit and space-time reference requirements of subsequent models. The loose coupling mode determines the period according to the simulation step and the synchronization step number, collects the output in the period and adapts it to batch input, and after the period is synchronized, the model whose input is ready is called according to the dependence; the tight coupling mode adapts the new output to incremental input, updates the input of the subsequent model and calculates, forms the output used for subsequent interaction, and obtains the linkage simulation result. Through mode selection and data adaptation, model execution is matched, the batch linkage coordination overhead is reduced, and the incremental linkage information lag is reduced.

Description

Technical Field

[0001] This application relates to the field of water conservancy digital twin technology, and in particular to a pluggable coupling scheduling method and system for multi-disciplinary water conservancy models. Background Technology

[0002] With the development of water conservancy informatization and digital twin technology, water conservancy digital twin platforms are gradually being applied to scenarios such as basin flood forecasting, flood control scheduling, water resource allocation, and water conservancy project operation management. These platforms analyze basin runoff generation and flow evolution, as well as project operation processes, by organizing hydrological models, hydrodynamic models, and water conservancy project scheduling models for joint simulation, providing a basis for hydrological projection and scheduling plan formulation. Since different professional models describe different processes within the water conservancy system, the models need to transfer calculation results, connect boundary conditions, and operate collaboratively according to business requirements. Therefore, the effective integration and coordinated calculation of multiple professional models is a crucial foundation for water conservancy digital twin platforms to conduct comprehensive simulations.

[0003] In existing technologies, the integration of multi-disciplinary water conservancy models typically involves adapting the calling interfaces, input / output parameters, and data transmission processes of each model separately based on specific projects and business scenarios, and then conducting linked calculations according to pre-defined model combinations and execution methods. However, different professional models differ in development environments, calling methods, and parameter organization. When integrating a new model or replacing an existing one, it is often necessary to rewrite the interface adaptation program, and in some cases, modify the underlying model code, resulting in a large workload for model integration and making cross-project reuse and expansion difficult. Furthermore, the formats, units, time steps, and spatial ranges of the input and output data of different models are inconsistent. The calculation results of preceding models cannot be directly used as input for subsequent models. Manual data processing, unit conversion, and alignment of time and spatial information are usually required, increasing the complexity of data integration and easily introducing conversion errors or causing data update delays. Furthermore, different simulation tasks have different requirements for computational accuracy, response time, and computing resources. When a relatively fixed model linkage method is adopted, frequent data exchange will increase computational coordination and resource overhead. Long intervals of data exchange may also affect the timeliness of information updates between models, making it difficult to adapt to the needs of different tasks such as long-term multi-scheme simulation and emergency scheduling.

[0004] Therefore, in the joint simulation of multi-disciplinary models in the water conservancy digital twin platform, the large workload of heterogeneous model access and replacement adaptation, the complexity of cross-model data connection, and the difficulty of adapting the model linkage method to the accuracy and timeliness requirements of different tasks have become urgent problems to be solved. Summary of the Invention

[0005] This application provides a pluggable coupling scheduling method and system for multi-disciplinary models in water conservancy, aiming to solve the problems in the existing technology of multi-disciplinary model joint simulation in water conservancy digital twin platform, such as large workload of heterogeneous model access and replacement adaptation, complex cross-model data connection, and model linkage mode that is difficult to adapt to the accuracy and timeliness requirements of different tasks.

[0006] Firstly, a pluggable coupled scheduling method for a multi-disciplinary water conservancy model is provided, the method comprising: Based on the unified interface specification and parameter mapping rules, multiple water conservancy professional models are encapsulated to obtain model components that can be called through the unified interface. The parameter mapping rules are used to establish the correspondence between the input and output parameters of each water conservancy professional model and the platform's standardized parameters. Acquire water conservancy simulation task and basic working condition data, determine the execution dependency and data transmission relationship between the model components based on the water conservancy simulation task, and obtain the model linkage topology; Based on the accuracy requirements, response delay requirements, and current computing power usage of the water conservancy simulation task, a target coupling mode is determined according to the preset coupling mode matching rules. The target coupling mode is either a loosely coupled mode that synchronizes data in batches according to a preset simulation cycle, or a tightly coupled mode that transmits incremental data and updates the model input in real time. Based on the basic working condition data, according to the model linkage topology and the target coupling mode, the model components are called through the unified interface. When data is transferred between model components, the output data of the preceding model components is converted into standardized data according to the parameter mapping rules, and then adapted into input data that meets the format, unit and spatiotemporal reference requirements of the subsequent model components, driving the subsequent model components to run and obtaining the linkage simulation results. When the target coupling mode is a loose coupling mode, the execution process includes: Based on the preset simulation step size and batch synchronization steps, the simulation time range between two adjacent batch synchronizations is determined, and the batch synchronization cycle is obtained. According to the batch synchronization cycle, each model component runs independently and asynchronously under the condition that its own input data is available, so as to obtain the model output data within the cycle. Based on the model linkage topology and the parameter mapping rules, the model output data of the preceding model components within the cycle are collected, standardized, and adapted for input to obtain the batch input data of the corresponding subsequent model components. According to the batch synchronization cycle, the batch input data is synchronized to the corresponding subsequent model components, and the subsequent model components with ready input data are called according to the execution dependency to perform calculations and obtain the staged linkage simulation results. When the target coupling mode is a tightly coupled mode, the execution process includes: Obtain the new output data generated by the preceding model component at the current data interaction moment to obtain the incremental output data; Based on the model linkage topology and the parameter mapping rules, the incremental output data is standardized and adapted for input to subsequent model components to obtain incremental input data. Based on the incremental input data, update the input parameters corresponding to the subsequent model components to obtain the updated model input; Based on the updated model input, the subsequent model component is invoked to perform calculations, obtaining model output data for subsequent incremental interactions, thereby advancing the calculation of the water conservancy simulation task.

[0007] Optionally, in the above scheme, the encapsulation of multiple water conservancy professional models according to a unified interface specification and parameter mapping rules to obtain model components that can be called through a unified interface includes: Based on the calling protocol and input / output parameters of each water conservancy professional model, the calling method and parameter configuration are determined according to the unified interface specification to obtain the model encapsulation configuration; Based on the model encapsulation configuration, the calling process of each water conservancy professional model is encapsulated through the model adapter, and the parameter mapping rules are configured to obtain the model components to be connected; According to the parameter mapping rules, the output parameter set of the model component to be connected is standardized and transformed, and verified according to the platform's preset field, unit, time series and spatial reference specifications to obtain the parameter verification result; Based on the parameter verification results, the verified model components to be connected are connected to the WIM water conservancy digital twin platform to obtain model components that can be called through a unified interface.

[0008] Optionally, in the above scheme, the step of acquiring the water conservancy simulation task and basic working condition data, determining the execution dependencies and data transfer relationships between the model components based on the water conservancy simulation task, and obtaining the model linkage topology includes: Obtain the water conservancy simulation task and basic working condition data, and determine the model components participating in the linkage calculation and the input and output parameters of each model component based on the water conservancy business processes involved in the water conservancy simulation task, so as to obtain the model task configuration. Based on the model task configuration, establish the correlation between the output parameters of the preceding model components and the input parameters of the following model components to obtain the data transfer relationship between models; Based on the data transfer relationships between the models, the execution priority and serial or parallel execution relationship of each model component are determined, and the execution dependency relationship is obtained. Based on the execution dependencies and the data transfer relationships between models, the model components in the model task configuration are organized to obtain the model linkage topology.

[0009] Optionally, in the above scheme, determining the target coupling mode according to the accuracy requirements, response delay requirements, and current computing power usage of the hydraulic simulation task, and in accordance with preset coupling mode matching rules, includes: Based on the accuracy requirements, response latency requirements, and current computing power utilization, the accuracy requirement index value, allowable response latency, and current computing power utilization rate are obtained to obtain the mode determination parameters. Based on the mode determination parameters, the results are compared with preset accuracy thresholds, delay thresholds, and computing power utilization thresholds to obtain the comparison results for each threshold. The coupling pattern matching conditions are obtained by combining the results of the comparison of various thresholds. Based on the coupling mode matching conditions, a matching rule is determined from the preset coupling mode matching rules, and the loose coupling mode or tight coupling mode corresponding to the rule is determined as the target coupling mode.

[0010] Optionally, in the above scheme, the step of converting the output data of the preceding model component into standardized data according to the parameter mapping rule, and then adapting it into input data that conforms to the format, unit, and spatiotemporal reference requirements of the subsequent model component, includes: Based on the parameter mapping rules, determine the field correspondence and unit conversion relationship between the output parameters of the preceding model component and the platform standardized parameters, and obtain the output conversion configuration; According to the output transformation configuration, the output data of the preceding model component is mapped to fields and transformed to units, and aligned according to the time and space references specified by the platform to obtain the standardized data; Based on the correspondence between the input parameters of the subsequent model components and the platform standardization parameters, the data fields, units of measurement, time steps and spatial range required by the subsequent model components are determined to obtain the input adaptation configuration; Based on the input adaptation configuration, the standardized data is subjected to format conversion, unit conversion, temporal alignment, and spatial matching to obtain input data that meets the input requirements of the subsequent model components.

[0011] Optionally, in the above scheme, before performing field mapping and unit conversion on the output data of the preceding model component, the following method is further included: Based on preset data integrity rules and anomaly detection rules, the output data of the preceding model component is detected to obtain the location of missing data and the location of abnormal data. Based on the location of the abnormal data, the output data of the preceding model component is filtered to obtain the filtered output data; Based on the location of the missing data, the filtered output data is padded with missing data to obtain the data to be converted for the field mapping and unit conversion.

[0012] Optionally, in the above scheme, the model task configuration includes a hydrological model component, a hydrodynamic model component, and a water conservancy project scheduling model component connected sequentially according to the data transmission direction. The process of performing linked calculations according to the model linkage topology includes: Based on the basic working condition data, the hydrological model component is invoked to perform runoff generation and runoff calculation, and the runoff generation and runoff calculation results are obtained. Based on the data transfer relationship between the models and the parameter mapping rules, the runoff generation and confluence calculation results are standardized and adapted for the input of the hydrodynamic model components to obtain hydrodynamic input data. Based on the hydrodynamic input data and the basic working condition data, the hydrodynamic model component is invoked to perform hydrodynamic evolution calculations and obtain hydrodynamic evolution results. Based on the data transfer relationship between the models and the parameter mapping rules, the hydrodynamic evolution results are standardized and adapted to the input of the water conservancy project scheduling model components to obtain the engineering scheduling input data. Based on the engineering scheduling input data, the water conservancy engineering scheduling model component is invoked to perform engineering scheduling calculations and obtain engineering scheduling simulation results.

[0013] Secondly, a pluggable coupled scheduling system for multi-disciplinary water conservancy models, the system comprising: The model encapsulation module is used to encapsulate multiple water conservancy professional models according to the unified interface specification and parameter mapping rules, so as to obtain model components that can be called through the unified interface. The parameter mapping rules are used to establish the correspondence between the input and output parameters of each water conservancy professional model and the platform's standardized parameters. The topology construction module is used to acquire water conservancy simulation task and basic working condition data, determine the execution dependency and data transmission relationship between the model components according to the water conservancy simulation task, and obtain the model linkage topology. The mode determination module is used to determine the target coupling mode according to the accuracy requirements, response delay requirements and current computing power usage of the water conservancy simulation task, and according to the preset coupling mode matching rules. The target coupling mode is either a loose coupling mode that synchronizes data in batches according to a preset simulation cycle, or a tight coupling mode that transmits incremental data and updates the model input in real time. The linkage execution module is used to call the model components through the unified interface according to the basic working condition data, the model linkage topology and the target coupling mode, and when transferring data between model components, convert the output data of the preceding model components into standardized data according to the parameter mapping rules, and then adapt it into input data that meets the format, unit and time-space reference requirements of the subsequent model components, drive the subsequent model components to run, and obtain linkage simulation results. When the target coupling mode is loosely coupled, the linkage execution module is used to execute the following process: Based on the preset simulation step size and batch synchronization steps, the simulation time range between two adjacent batch synchronizations is determined, and the batch synchronization cycle is obtained. According to the batch synchronization cycle, each model component runs independently and asynchronously under the condition that its own input data is available, so as to obtain the model output data within the cycle. Based on the model linkage topology and the parameter mapping rules, the model output data of the preceding model components within the cycle are collected, standardized, and adapted for input to obtain the batch input data of the corresponding subsequent model components. According to the batch synchronization cycle, the batch input data is synchronized to the corresponding subsequent model components, and the subsequent model components with ready input data are called according to the execution dependency to perform calculations and obtain the staged linkage simulation results. When the target coupling mode is a tightly coupled mode, the linkage execution module is used to execute the following process: Obtain the new output data generated by the preceding model component at the current data interaction moment to obtain the incremental output data; Based on the model linkage topology and the parameter mapping rules, the incremental output data is standardized and adapted for input to subsequent model components to obtain incremental input data. Based on the incremental input data, update the input parameters corresponding to the subsequent model components to obtain the updated model input; Based on the updated model input, the subsequent model component is invoked to perform calculations, obtaining model output data for subsequent incremental interactions, thereby advancing the calculation of the water conservancy simulation task.

[0014] Compared with the prior art, this application has at least the following beneficial effects: This application, based on further analysis and research of existing technical problems, recognizes the issues in multi-disciplinary model co-simulation of water conservancy digital twin platforms, including the large workload of heterogeneous model access and replacement adaptation, complex cross-model data connection, and difficulty in adapting model linkage methods to the accuracy and timeliness requirements of different tasks. By encapsulating different water conservancy professional models according to unified interface specifications and parameter mapping rules, each model becomes a uniformly callable model component, reducing the customized adaptation work caused by differences in model calling methods and input / output parameters, and improving the convenience of model access, replacement, and reuse. Furthermore, by determining the execution dependencies and data transfer relationships between models according to the water conservancy simulation task, a model linkage topology is constructed to provide a basis for the execution order and data flow of multi-model calculations. Finally, by combining task accuracy requirements, allowable response latency, and current computing power usage, the target coupling mode is determined, enabling the data interaction method between models to adapt to the accuracy and timeliness requirements of different tasks, thus improving the overall system performance. Fixed coupling mode struggles to balance simulation requirements and computational overhead. Therefore, this approach organizes model calls according to model linkage topology and target coupling mode. The output of preceding models is converted into standardized data and then adapted to input data conforming to the format, unit, and spatiotemporal reference requirements of subsequent models. This ensures that preceding calculation results can seamlessly connect to subsequent calculations, reducing errors and delays caused by manual data conversion and transmission. Simultaneously, in loose coupling mode, data is collected and adapted in batches according to synchronization cycles, and subsequent models with ready inputs are called based on execution dependencies to reduce the overhead of frequent exchanges and coordination calls. In tight coupling mode, new outputs are adapted as incremental inputs, updating subsequent model inputs and driving calculations. This ensures that changes in operating conditions are continuously propagated along the model linkage topology, reducing information lag caused by waiting for batch synchronization. This achieves automated linkage simulation of multiple professional models, reduces model integration and data adaptation costs, improves the continuity of simulation execution, and enhances adaptability to different water conservancy business scenarios. Attached Figure Description

[0015] Figure 1 This is a flowchart illustrating a pluggable coupling scheduling method for a multi-disciplinary water conservancy model provided in one embodiment of this application. Detailed Implementation

[0016] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0017] In one embodiment, such as Figure 1As shown, a pluggable coupling scheduling method for multi-disciplinary water conservancy models is provided, which organizes different water conservancy professional models to conduct joint simulations based on the WIM water conservancy digital twin platform. In this embodiment, WIM refers to the water conservancy digital twin platform that carries out water conservancy data management, professional model integration, and simulation applications; a specific deployment example is the Yuyu WIM water conservancy digital twin platform. The WIM water conservancy digital twin platform provides the basic operating condition data, model operating environment, and simulation result application environment required for water conservancy operations. A unified adaptation and coupling scheduling middleware layer for multiple models is set between the platform's business functions and each water conservancy professional model, undertaking model access, mode selection, data adaptation, and coordinated execution. This method can be used for long-term basin simulation, historical operating condition review, multi-scheme comparison, flood control emergency drills, cascade pumping station scheduling, water resource allocation, and engineering operation analysis.

[0018] During model encapsulation, a unified interface specification defines the entry point, parameter organization method, and result return method used by the platform when calling model components. Parameter mapping rules define the correspondence between the model's own input and output parameters and the platform's standardized parameters. Based on the above specifications and rules, the model adapter encapsulates the calling process of different models within a unified interface, enabling the platform to drive each model using a consistent calling method, while the existing specialized calculation processes within the models remain executed by the respective models.

[0019] Model components are callable units formed from hydraulic engineering models after interface encapsulation and parameter adaptation. Platform-standardized parameters simultaneously associate the physical meaning, field names, units of measurement, time information, and spatial object information of the parameters. Parameter mapping rules include the correspondence between model outputs and platform-standardized parameters, as well as the correspondence between platform-standardized parameters and model inputs. Different models can retain their own internal field names and data organization formats, and connect to the unified interface through model adapters.

[0020] The hydraulic simulation task is used to determine the hydraulic operational process to be analyzed, the simulation time range, the models involved in the calculation, and the accuracy and response delay requirements. Basic operating condition data is used to form the initial and boundary inputs for model calculations, and may include rainfall and hydrological data, watershed and river channel basic data, hydraulic engineering parameters, and engineering operation status, depending on the specific task. Each data point is provided to the required model components according to its temporal and spatial assignments; it is not required that each model receive all basic operating condition data.

[0021] Based on the hydraulic simulation task, the platform determines the execution dependencies and data transfer relationships between model components and organizes them into a model linkage topology. Execution dependencies indicate which prerequisite results are required before subsequent model calculations, while data transfer relationships indicate which parameters from the prerequisite results should be passed to which subsequent model. Models without mutual input dependencies can be computed in parallel, while models with input dependencies run in the order of prerequisite result formation, data adaptation, and subsequent calculations.

[0022] After obtaining the accuracy requirements, allowable response latency, and current computing power utilization, the platform determines the target coupling mode according to preset coupling mode matching rules. Loose coupling mode synchronizes data in batches according to a preset simulation cycle, suitable for tasks involving phased analysis over a longer timeframe; tight coupling mode transmits incremental data and updates model inputs in real time, suitable for tasks requiring timely responses to changes in operating conditions. Here, response latency refers to the timeliness requirements specified by the task, representing different information from the model's simulation time step.

[0023] During the execution of the linkage calculation, the platform first provides basic operating condition data to the model components with available input conditions and obtains the model output through a unified interface. For data that needs to be passed to subsequent models, standardized data is first generated according to parameter mapping rules, and then input data is formed according to the format, units, and spatiotemporal reference requirements of the subsequent models. After the data transfer is completed and the necessary inputs are available, the platform calls the subsequent model components to perform calculations and continues to process their outputs along the model linkage topology until the linkage simulation results corresponding to the hydraulic simulation task are formed.

[0024] In this process, data adaptation is embedded in the actual data exchange between models. In loosely coupled mode, the batch synchronization period is determined according to the preset simulation step size and batch synchronization steps. The model output within the period is collected and adapted as batch input. After synchronization according to the period, the subsequent model with ready input is called according to the execution dependency. In tightly coupled mode, the new output generated at the current interaction moment is adapted as incremental input, the corresponding input parameters of the subsequent model are updated, and then the subsequent model is called to calculate, forming the output for subsequent incremental interactions. Both modes complete data transfer and input adaptation based on the model linkage topology and corresponding parameter mapping rules. For tasks whose results still need to participate in the calculation of subsequent simulation periods, data input, model calculation, result transfer, and subsequent calculation continue to advance according to the model linkage topology, forming a continuous linkage simulation process.

[0025] This embodiment establishes a common calling method for different water conservancy professional models through unified interface encapsulation. It eliminates the connection barriers caused by inconsistent data expression between models through parameter mapping and input adaptation. It also organizes data exchange and calculation execution through model linkage topology and target coupling mode, so that model access, data processing and operation scheduling cooperate with each other, thereby reducing the work of model customization and integration and manual data conversion, reducing data connection errors and update lag, and adapting to the requirements of different simulation tasks on operation overhead and result timeliness.

[0026] In one implementation, the platform obtains the existing calling protocol and input / output parameter descriptions for the hydrological model, hydrodynamic model, or water conservancy project scheduling model to be connected. The calling protocol is used to determine the model's calling entry point, startup parameters, and result acquisition method, while the input / output parameter descriptions are used to determine the parameter names, data types, units, and corresponding time and space information.

[0027] Based on the above information, a model encapsulation configuration is formed. This configuration maps the calls made in the platform's unified interface to the model's original calling methods, and maps the parameters passed to the platform during the call to the parameters required by the model. The model adapter organizes the calling process according to this configuration, enabling the platform to convert computation requests sent through the unified interface into a call format recognizable by the corresponding model, and to obtain the returned results after the model completes the computation. This process uses the model's existing callable capabilities for external encapsulation, without requiring changes to the model's internal hydraulic calculation algorithms for unified access.

[0028] Parameter mapping rules are configured for each model and stored in a model-specific parameter mapping rule library. For fields with the same physical meaning but different names, the correspondence between the model field and the platform field is recorded. For parameters represented by different units, time stamps, or spatial objects, corresponding transformation and association rules are configured. For example, the flow field in the model output can be mapped to the platform's standard flow field, while retaining the corresponding time and cross-sectional information to determine which model it can be used as input for later.

[0029] After the model adapter is configured, the model component to be integrated is obtained. When standardizing its output parameter set, the model adapter reads the model-specific parameter mapping rule library and the platform's preset parameter specifications, and performs adaptation according to the following formula:

[0030] In the formula, The set of output parameters from the original hydraulic engineering model received by the model adapter; This indicates that the platform has preset standardized parameters for the hydraulic model. These parameters include field, unit, time series, and spatial reference requirements, and are represented by a set of parameters that meet these requirements. A model-specific parameter mapping rule library; An adapter function that performs the transformation according to the standardized parameter specification and parameter mapping rules; This is a standardized set of parameters that, after conversion, can be used by the platform for cross-model data exchange. The subscripts in the parameter set are used to distinguish the inputs of the adaptation function, and do not indicate that the parameter set is necessarily the input of the hydraulic engineering model itself.

[0031] The converted parameter set is validated according to platform specifications. Field validation determines whether required fields are present and whether the data types are consistent; unit validation determines whether the units used for parameter values ​​are consistent with the standardized parameter specifications; time series validation determines whether the time stamps and time ranges meet the requirements; spatial reference validation determines whether the spatial objects and references associated with the data can be recognized by the platform. Parameter specification compliance can be expressed as:

[0032] In this context, membership represents the set of parameters that conform to the platform's standardized parameter specifications after transformation, i.e., it passes parameter specification compliance verification. Based on the verification results, the parameter verification outcome is determined, and the verified model components are then integrated into the platform. Parameter sets that do not meet the required fields or baseline requirements cannot be used as successfully adapted data in subsequent models. This verification focuses on parameter specification compliance; the model's runtime environment and entry point still need to be configured according to the model encapsulation.

[0033] After integration, similar models can participate in the same type of computational tasks through the same unified interface. When replacing an existing model component with another, the replacement model is configured with the appropriate calling method and parameter mapping rules, and its input and output are used for subsequent calls after meeting the requirements of the task. Model unloading is used to remove the corresponding component from the platform's callable models, and model reuse is used to use the adapted component for other tasks that meet its applicable conditions; pluggability here does not require runtime state migration during model execution.

[0034] This embodiment centralizes the differences in calling protocols and parameters into the model encapsulation configuration and parameter mapping rules, and verifies the standardized parameter set during access. This enables the platform to reuse a unified interface to complete the access and calling of heterogeneous models, reducing the work of repeatedly developing business connection programs for different models, and providing a consistent parameter connection basis for model replacement, uninstallation and cross-task reuse, thereby reducing the workload of platform model integration and iterative maintenance.

[0035] In one implementation, after obtaining the water conservancy simulation task and basic operating condition data, the platform determines the model components to participate in the calculation based on the required water conservancy business processes. For example, a watershed inflow and engineering scheduling analysis task may simultaneously involve runoff generation and confluence calculation, hydrodynamic evolution calculation, and water conservancy engineering scheduling calculation. Based on this, the platform determines the corresponding model components and the input and output parameters of each component, forming a model task configuration.

[0036] Model task configuration records the computational processes undertaken by model components, along with the inputs and outputs required for those processes. Input parameters can come from baseline load data or the outputs of other model components. For parameters directly provided by the baseline load data, their data source is recorded; for parameters passed between models, the preceding model providing the parameter and the subsequent model receiving the parameter are identified, thus avoiding the mixing of data from different sources but with similar names as the same input.

[0037] When establishing data transfer relationships between models, first determine the correspondence based on the physical meaning of the parameters, and then verify the time range and spatial object corresponding to the data. For example, the runoff generation and confluence results intended as input for river inflows should be associated with the river inlet receiving the inflow or the corresponding calculation location; data from different watershed units or different cross-sections cannot be directly transferred simply because the field names are the same. Parameter mapping rules are responsible for the specific transformation, while data transfer relationships between models are responsible for determining which component and input location the data is transmitted to.

[0038] Execution dependencies are formed based on the data transfer relationships between models. When a subsequent model requires the output of a preceding model, it should wait for the corresponding output to be formed and the input adaptation to be completed. When the inputs of different models are already available and there are no dependencies between them required for current computation, they can be executed in parallel. Execution priority is used to determine the scheduling order among executable model components. Components whose data dependencies are not met will not use results that have not yet been formed in advance due to higher priority.

[0039] Model components are treated as computational units within the topology, data transfer relationships are used as connections between units, and execution dependencies and priorities are associated with the corresponding computational units to obtain a model-linked topology. During platform runtime, this topology determines which components can be invoked and sends the generated outputs to the corresponding subsequent components. For components whose inputs include both basic operating condition data and preceding results, these two elements together form the input required for that invocation.

[0040] The same model component can be connected to different preceding or succeeding components in different task configurations, and the data transformation is still completed by the parameter mapping rules corresponding to the model. Therefore, when adjusting business tasks, the model linkage topology can be reorganized without writing all business dependencies into the internal calculation process of a single model. The topology can represent either a sequential link or the relationship where multiple branch results converge into subsequent models; the specific connections depend on the input-output dependencies adopted by the task.

[0041] This application organizes model selection, parameter sources, and execution dependencies into a model linkage topology, enabling subsequent models to execute according to predetermined relationships once the required data is available. It also enables independent computational processes to run in parallel, thereby reducing the need for manual arrangement of model calls and result transfer, avoiding input loss or result mismatch due to inconsistencies between execution order and data dependencies, and improving the manageability and execution efficiency of multi-model linkage processes.

[0042] In one implementation, the accuracy requirement index is derived from the precision requirements of the hydraulic simulation task for the calculation results, the allowable response latency is derived from the task's requirements for the result return time, and the current computing power utilization rate is derived from the operating status of the computing resources used to perform the model calculation. These parameters collectively constitute the mode determination parameters, representing the task's accuracy requirements, timeliness requirements, and current resource availability, respectively.

[0043] The preset coupling mode matching rules include the correspondence between conditions and modes. The platform compares the accuracy requirement index value, allowable response latency, and current computing power utilization rate with the corresponding thresholds, and then combines the comparison results according to the logical conditions in the rules to obtain the coupling mode matching conditions. Based on these conditions, the matching rules are determined, and the loosely coupled mode or tightly coupled mode recorded in the rules is obtained as the target coupling mode for the current task. The following describes the specific determination process for tasks that satisfy the following two sets of conditions:

[0044] In the formula, The target coupling mode is defined as Loose, which indicates a loosely coupled mode, and Tight, which indicates a tightly coupled mode. For accuracy requirements, To allow for response delay, This represents the current computing power utilization rate. , and These are the preset accuracy threshold, latency threshold, and computing power utilization threshold, respectively. The "∩" in the formula indicates that all connected comparison conditions are met simultaneously. Each indicator and its corresponding threshold uses the same evaluation criteria and measurement method.

[0045] A loosely coupled mode is achieved when the accuracy requirement is no greater than the accuracy threshold, the allowable response latency is no less than the latency threshold, and the current computing power utilization is no less than the computing power utilization threshold. This condition is suitable for tasks with relatively relaxed accuracy and response time requirements and currently heavy resource usage. Periodic batch synchronization reduces high-frequency data exchange and computational coordination overhead. This mode can be used for large-scale historical review or multi-scheme comparison tasks when this condition is met.

[0046] When the accuracy requirement index value is greater than the accuracy threshold and the allowable response delay is less than the delay threshold, a tightly coupled mode matching result is obtained. This condition is used to express tasks that have high requirements for both the granularity of the results and the timeliness of the response. Incremental data interaction is used to update the subsequent model input in a timely manner, reducing the information lag caused by waiting for batch synchronization. This mode can be adopted for flood control emergency drills or real-time scheduling of cascade pumping stations when this condition is met.

[0047] In the above judgment example, the accuracy requirement index values ​​are compared according to the principle that larger values ​​indicate higher accuracy requirements. Permissible error values ​​with opposite meanings cannot be directly substituted and the same comparison direction cannot be used. A smaller allowable response delay indicates that the task requires a faster return of results. Thresholds are preset configuration parameters and must be consistent with the specific task evaluation indicators; they do not require all services to use the same set of values.

[0048] Once the target coupling mode is determined, it is used for model-wide execution, enabling data exchange to adopt the batch synchronization or incremental interaction method corresponding to that mode. When subsequent tasks or inference phases use the re-acquired mode determination parameters, the matching process can be executed again to obtain the target coupling mode corresponding to that task or phase. Mode selection represents the determination of the data exchange and invocation organization method; whether the actual computation can meet the response requirements also depends on the model's computational load and available resources.

[0049] This embodiment converts task accuracy requirements, response latency requirements, and current computing power usage into comparable mode determination parameters, and maps the comparison results to specific data exchange methods through preset rules. This provides clear parameter basis for coupling mode selection, reduces the work of manually configuring linkage methods for each task, and thereby improves the adaptability of the operation mode to different water conservancy simulation tasks by using batch synchronization to reduce coordination overhead and incremental interaction to reduce data update lag under applicable working conditions.

[0050] In one implementation, platform-standardized parameter specifications serve as the basis for data interaction between different models. Output transformation configurations are used to convert the output of preceding models into standardized data, while input adaptation configurations are used to process the standardized data into inputs acceptable to subsequent models. These two types of configurations are directed to the source model and the target model respectively, avoiding the solidification of data transformation between each pair of models into independent business procedures.

[0051] When determining the output conversion configuration, the correspondence between source fields and standard fields, as well as the unit conversion relationship, are read according to the parameter mapping rules, while preserving the physical meaning of the parameters. For example, the field names used in the source model for flow rate and water level may differ from the platform name, but after conversion, they should still represent flow rate and water level respectively. Different physical quantities cannot be treated as the same parameter simply by renaming them. Unit conversion applies when the same physical quantity uses different units, and the conversion result should be consistent with the units specified in the standard fields.

[0052] After completing field mapping and unit conversion, time information and spatial object information are associated according to platform requirements to form standardized data. Time information is used to determine the corresponding moment or time interval, while spatial object information is used to determine the corresponding watershed unit, river segment, cross section, or water conservancy project. For parts that already conform to platform specifications, their values ​​and associated information are retained, without the need for repeated numerical changes.

[0053] For configurations that are numerically normalized according to parameter value ranges, the following normalized fusion expression is used:

[0054] In the formula, Output data for the original model. and These are the upper and lower limits of the industry extreme values ​​for the corresponding parameters. The dimension conversion factor is... This is a time or spatial offset correction amount. To standardize and unify data, when executing this expression, the upper and lower limits should belong to the same parameter and have a consistent measurement caliber; they cannot be the same. The conversion coefficient and correction amount are provided by the conversion configuration of the corresponding parameters, and their numerical meanings should be consistent with the converted data expression.

[0055] This expression describes numerical transformations in the configuration. Matching of field semantics, time location, and spatial objects requires separate processing by corresponding rules. For subsequent models requiring inputs with actual physical units, input adaptation must ensure that the input values ​​and units conform to the model's requirements. Values ​​normalized to their range cannot be used indiscriminately as actual flow rates or water levels. Temporal or spatial relationships cannot be replaced simply by adding corrections to individual values.

[0056] The time base offset uses the following timing alignment correction expression:

[0057] In the formula, This is the corrected synchronization time. The base time to be corrected. This represents the offset between the two time representations. After adjusting the time stamps based on the offset, the source and target models can identify data for the same physical moment. This processing is used for time base offset and does not directly replace data resampling between different time steps.

[0058] When determining the input adaptation configuration, read the fields, units of measurement, time steps, and spatial ranges required by the subsequent model. For continuously changing parameters with different time steps, the target time series can be formed using time interpolation or interval summarization according to the parameter configuration; the method used depends on whether the parameter represents an instantaneous value, an interval average, or an interval cumulative value. For example, interval cumulative values ​​need to be processed according to the time range and cannot be interpreted as instantaneous values. For data lacking the required time range, simply adjusting the timestamps does not mean the input is ready.

[0059] Spatial matching is performed based on the spatial objects associated with the source data and the spatial objects required by the subsequent model. When the source data and the target input correspond to the same location or the same cell, they can be passed according to the object correspondence; when different spatial divisions are involved, the target input should be formed according to the association or aggregation relationship in the configuration. Spatial benchmark transformation is used to unify coordinates or elevation references, while spatial range matching is used to select the corresponding data within the target range. The two are processed separately, and time offset or numerical normalization is not equated with spatial range matching.

[0060] After completing the above processing, the data is organized into the field arrangement, record structure, or input file format required by the subsequent model to obtain the input data for the subsequent model. Output adaptation and input adaptation constitute cross-model bidirectional adaptation capability, that is, the model can convert its own results into platform expressions, and can also receive platform expressions and convert them into its own input; this bidirectional adaptation does not mean that every aggregation, completion, or filtering operation is mathematically reversible.

[0061] This embodiment configures the source model output conversion and target model input adaptation separately, so that the data can participate in subsequent calculations under the condition that the field semantics, units, time and spatial affiliation are all corresponding. This reduces the work of repeatedly building dedicated conversion programs between different models, reduces input errors caused by mixed units, time misalignment or spatial object mismatch, and provides a consistent data processing foundation for batch synchronization and incremental interaction.

[0062] In one implementation, before the output data of the preceding model enters the field mapping and unit conversion, it undergoes integrity and anomaly detection. Data integrity rules are determined based on the fields, time coverage, and spatial object range that the parameters to be transmitted should have. These rules are used to identify situations where required fields are missing, data is missing at expected time locations, or parameter values ​​are missing from specified object records, and the corresponding record locations and parameter locations are used to form missing data locations.

[0063] Anomaly detection rules are configured based on the meaning of the parameters. For example, they can detect whether the field data type meets the requirements, whether the value is within the allowed range of the parameter, and whether there are obvious conflicts between records of the same time and object. The location where the anomaly detection rule is hit is recorded as the location of the abnormal data. Missing data detection and anomaly detection results are retained separately to avoid treating invalid values ​​as normal observations in subsequent unit conversions or time processing.

[0064] The output data is filtered based on the location of outliers, preventing values ​​deemed abnormal from being processed and resulting in filtered output data. During filtering, the necessary information for identifying the original time and spatial objects is preserved, ensuring that missing data locations can still be mapped to the appropriate parameters. The purpose of outlier filtering is to exclude outliers from calculations; it does not mean that every filtered location automatically forms a reliably complete target.

[0065] Based on the location of missing data, missing parameter values ​​are generated according to preset completion rules. For continuous time-series parameters that allow interpolation, data from adjacent valid time points can be used for interpolation; for parameters for which the model explicitly allows the use of preset values, completion can be performed according to the corresponding configuration. The completion method should be consistent with the physical meaning of the parameter and the input format allowed by the model, and discrete parameters such as engineering operation status should not be treated as continuous values. The above methods are the feasible data processing implementations, and the specific selection is determined by the corresponding parameter configuration.

[0066] The completed data is then transformed into data to be processed by the field mapping and unit conversion processes. If the necessary input is still unavailable, it is not considered ready; the execution of subsequent models remains subject to input dependencies in the model linkage topology. In this way, data quality processing and subsequent scheduling are connected through available inputs, rather than unconditionally starting subsequent computations after only completing filtering or completion operations.

[0067] This embodiment identifies missing and abnormal locations before cross-model conversion, eliminates outliers, and completes missing data that can be completed according to parameter rules. This ensures that the data entering the subsequent mapping and adaptation process has better integrity and effectiveness, reduces the propagation of abnormal values ​​in the model chain, and lowers the possibility of interrupted linkage calculations or incorrect inputs due to missing data.

[0068] In loosely coupled mode, the preset simulation step size represents the basic advance interval on the linkage calculation time axis, and the batch synchronization step number represents the basic advance step number between adjacent synchronizations. The batch synchronization period is determined based on these two parameters, and the simulation time range requiring batch switching is divided accordingly. The batch synchronization period is a simulation time organization parameter and is not equivalent to the actual time consumed by the computer to complete the batch calculation.

[0069] Within a batch synchronization cycle, each model component runs independently and asynchronously, producing model outputs for the corresponding time frame, provided input data is available. Independent asynchronous execution allows different models to have different computation times, while still adhering to input dependencies; models requiring preceding results can only perform their computations after those results are generated and adapted. Therefore, loose coupling does not require all models to run simultaneously when input is lacking.

[0070] The platform aggregates the outputs of preceding model components within their respective timeframes according to the model linkage topology, performs standardized transformations based on parameter mapping rules, and then generates batch input data for the time steps and spatial ranges required by the corresponding subsequent models. During data aggregation, the temporal order and spatial object correspondence are maintained, ensuring that each value in a batch of data can be associated with the input positions required by the subsequent models.

[0071] During the data exchange phase of the batch synchronization cycle, batch input data is transferred to subsequent model components. For multi-level links with serial dependencies, data adaptation and model invocation are completed step by step according to execution dependencies; for independent branches with existing inputs, parallel computation can be organized according to execution priority. After obtaining the simulation results of the current stage, subsequent synchronization cycles are advanced, allowing data to continuously flow between models in batches.

[0072] For the model task configuration where hydrological and hydrodynamic calculation results are combined into pump station scheduling calculations, loosely coupled batch linkage can be expressed by the following formula:

[0073] In the formula, This is the simulation result for the current batch linkage phase; This represents the time series of basic operating conditions associated with this batch phase; To synchronize steps in batches, This refers to the step position corresponding to the current stage; This is a function for calculating the runoff in batch hydrological data. For batch hydrodynamic evolution calculation functions, This is the pump station scheduling calculation function that receives the results from the first two types. This formula represents the relationship between the two calculation results and the merging of the scheduling calculation; the specific range of data used in each batch should be consistent with the configured synchronization period.

[0074] Both calculation functions in the formula take the corresponding basic working condition sequence as explicit input. Therefore, this expression is used to describe the above result aggregation configuration. When using preceding hydrological results to further drive the serial task of the hydrodynamic model, the input should be transferred according to the corresponding model linkage topology. Both types of configurations can use batch synchronization, but the model connection relationship is determined by their respective tasks and is not replaced by changing the formula symbols.

[0075] This embodiment reduces the overhead of frequent data exchange and repeated coordination calls between models by aggregating model results and performing batch synchronization within a predetermined simulation time range. At the same time, it retains input readiness and execution dependency constraints, enabling simulation tasks with a longer time range to proceed stably in stages. This adapts to the requirements of historical working condition review, large-scale long-term simulation, and multi-scheme comparison tasks on the utilization of computing resources and batch processing efficiency.

[0076] The incremental interaction process in the tightly coupled mode is further explained as follows: The platform obtains the new output data generated by the preceding model component at the current data interaction moment, forming incremental output data. The incremental output data carries its parameter content and corresponding time and space object information, used to distinguish the data that needs to be provided to the subsequent model in this interaction from the data that has already been interacted with.

[0077] Based on the model's interconnected topology, the receiving component and input location of the incremental output data are determined. Then, according to the parameter mapping rules, it is converted into standardized data, and input adaptation is performed according to the format, unit, and spatiotemporal reference requirements of the receiving component to obtain the incremental input data. If the data only involves some parameters or some spatial objects, the update range is based on the actual input corresponding to the incremental data to avoid mistakenly applying changes from one location to other locations.

[0078] The incremental input data is used to update the corresponding input parameters of subsequent model components, resulting in the updated model input. During input updates, the time and object correspondence between the basic operating condition data, existing valid inputs, and the current incremental input is maintained. When a subsequent model requires data from multiple sources, the availability of the required inputs for the current calculation is determined based on the execution dependencies. The timeliness of data interaction and the consistency of inputs jointly determine the timing of subsequent calls.

[0079] Once the necessary inputs are available, subsequent model components are called through a unified interface to calculate and obtain their current outputs. For outputs from subsequent receiving components, the next incremental interaction is formed according to the model linkage topology; for the final result of the current calculation stage, the corresponding linkage simulation result is generated. This process can be coordinated with second-level incremental interactions to organize continuous updates, and the actual completion time depends on the model solution time and runtime resources.

[0080] For the combined configuration of hydrological and hydrodynamic calculation results provided to the pump station scheduling model, the tightly coupled linkage adopts the following expression:

[0081] In the formula, for Watershed basic working condition data at any given time. This is a function for calculating hydrological runoff. For hydrodynamic evolution calculation functions, Optimize the calculation function for pump station scheduling. For the simulation results of the next moment, This represents the simulation time interval between adjacent linked calculation moments. The platform transmits each calculation output according to the selected incremental interaction method, enabling subsequent models to form subsequent results based on the updated input.

[0082] This expression explicitly indicates that hydrological and hydrodynamic results are jointly incorporated into the pump station scheduling calculation, without separately listing the internal state update of the model or the feedback convergence process at the same moment. The iteration in the tightly coupled mode is manifested here as continuous, linked calculations advancing with input updates; second-level interactions are not equated with the model having completed bidirectional feedback solutions. For tasks with other input dependencies, the actual data transmission direction is still determined by the corresponding model linkage topology.

[0083] This embodiment generates incremental inputs and updates subsequent models in a timely manner for new outputs, enabling changes in operating conditions to be continuously transmitted along the model linkage topology. This reduces information lag caused by waiting for the synchronization of the entire batch of results and avoids repeatedly transmitting all the data that has already been exchanged. As a result, it improves the timeliness of model input updates and linkage result formation in tasks such as flood control emergency drills and cascade pumping station scheduling.

[0084] In one implementation, the model task configuration includes a hydrological model component, a hydrodynamic model component, and a hydraulic engineering scheduling model component connected sequentially along the data transmission direction. This implementation uses a serial link, first generating runoff and confluence results, then using these results to drive hydrodynamic evolution, and finally applying the hydrodynamic evolution results to engineering scheduling. The data connection at each stage is illustrated below using a pump station scheduling application as an example.

[0085] The platform provides basic operating condition data based on the simulation time range and target area of ​​the task. Data used for runoff generation and confluence calculations includes relevant rainfall and watershed conditions. A unified interface is used to call the hydrological model component, and the pre-configured runoff generation and confluence calculation process of this component generates the calculation results. The results are associated with watershed units or confluence locations and corresponding time information according to the model definition, serving as the data source for subsequent hydrodynamic input adaptation.

[0086] Based on the data transfer relationships between models, the inflow parameters and their receiving locations that need to be provided to the hydrodynamic model in the runoff generation and confluence calculation results are determined. The model adapter converts these parameters into platform-standardized data, and then organizes the input according to the fields, units, time steps, and spatial locations required by the hydrodynamic model to obtain the hydrodynamic input data. This processing enables the runoff generation and confluence results to participate in the calculation in the form of boundary inputs required by the hydrodynamic model.

[0087] The platform provides hydrodynamic input data and corresponding basic operating condition data to the hydrodynamic model component. The basic operating condition data provides the river channel conditions, initial state, or other configured boundary conditions required by the model, while the hydrodynamic input data provides the inflow information formed by previous runoff calculations. The hydrodynamic model solves for the flow evolution according to its existing calculation process, obtaining the hydrodynamic evolution results such as water level and flow rate for the corresponding river section or cross-section.

[0088] Based on the parameters required by the engineering scheduling model, data corresponding to the engineering location and scheduling period are determined from the hydrodynamic evolution results. These data are then standardized, converted, and adapted to form the engineering scheduling input data. In the pump station scheduling example, parameters such as water level and flow rate at associated pump station locations can be provided to the pump station scheduling calculation process. The engineering operation capacity, operational constraints, and scheduling objectives are used by the configured hydraulic engineering scheduling model according to the task conditions.

[0089] By calling the water conservancy project scheduling model component through a unified interface, the system performs its existing scheduling calculations based on the project scheduling input data to obtain the project scheduling simulation results. The results, as defined by the model, include the project operation arrangements and corresponding evaluation results, which the platform can then use for scheduling scheme analysis or comparison. This generates simulation and decision analysis results; whether or not to issue commands to the actual project control equipment is a matter for the specific application interface to execute.

[0090] In this serial implementation, data transmission sequentially passes through the hydrological model, hydrodynamic model, and hydraulic engineering scheduling model, organizing batch synchronization or incremental interaction according to the selected target coupling mode. The specific runoff generation and confluence, hydrodynamic solution, and engineering scheduling algorithms within each model are provided by the connected professional model, with a unified adaptation and scheduling middleware layer responsible for invocation, connection, and cross-model data processing. The engineering scheduling results are then fed back to other connections such as the hydrodynamic boundary, requiring explicit configuration by the task employing this feedback relationship; it cannot be inferred from the serial link itself.

[0091] This embodiment connects runoff generation and runoff calculation, hydrodynamic evolution, and engineering scheduling along a clear data dependency, and automatically completes parameter and temporal-spatial adaptation between each stage. This enables water conservancy engineering scheduling to use the hydrological evolution information formed by the preceding model, reducing the manual export, sorting, and re-entry of intermediate results, lowering the risk of cross-stage data mismatch, and improving the continuity and automation of watershed hydrological analysis to engineering scheduling simulation.

[0092] This embodiment also provides a pluggable coupled scheduling system for multi-disciplinary water conservancy models, based on the WIM water conservancy digital twin platform. It includes a model encapsulation module, a topology construction module, a mode determination module, and a linkage execution module. Each module works collaboratively as a functional unit in the platform's unified adaptation and coupled scheduling intermediate layer for multiple models, establishing connections through model components, model linkage topology, target coupling modes, and simulation data.

[0093] The model encapsulation module encapsulates hydraulic engineering models according to a unified interface specification and parameter mapping rules, outputting model components that can be called through the unified interface. This module maintains the model encapsulation configuration and model-specific parameter mapping rules, enabling other modules to access professional computing capabilities via the unified interface and use parameter rules matching the corresponding model during data exchange. The model encapsulation module can also perform the aforementioned parameter validation to determine whether the model components to be integrated meet the platform's standardized parameter specifications.

[0094] The topology construction module acquires the hydraulic simulation task and basic operating condition data, determines the model components involved in the calculation, their execution dependencies and data transfer relationships, and outputs the model linkage topology. This module transforms the business process into an executable model organization relationship and provides the basic operating condition data and the resulting topology to the linkage execution module, enabling the latter to determine the initial calling object, the sequential relationship, and the subsequent data receiving location.

[0095] The pattern determination module obtains the accuracy requirements and response delay requirements of the hydraulic simulation task, and performs preset rule matching based on the current computing power occupancy, outputting the target coupling pattern. Upon receiving this pattern, the linkage execution module adopts either a loosely coupled execution method that synchronizes data in batches according to a preset simulation cycle, or a tightly coupled execution method that transmits incremental data in real time and updates the model input. The pattern determination process is distinct from the internal solution process of the specific hydraulic model.

[0096] The linkage execution module organizes model calls through a unified interface based on basic operating condition data, model linkage topology, and target coupling mode. When transferring data between models, the linkage execution module uses an automatic data conversion engine to read parameter mapping rules, converting preceding outputs into standardized data, adapting it as input for subsequent models, and driving the subsequent model to run when input conditions are met. The automatic data conversion engine can perform the aforementioned field mapping, unit conversion, temporal alignment, spatial matching, and missing / anomaly handling; the processing results are then used for model calls. Specifically, in loosely coupled mode, the linkage execution module determines the batch synchronization period based on the preset simulation step size and batch synchronization steps, collects model outputs within the period, completes standardization conversion and input adaptation, synchronizes batch inputs according to the batch synchronization period, and calls subsequent models with ready inputs based on execution dependencies. In tightly coupled mode, the linkage execution module converts and adapts new outputs at the current interaction moment into incremental inputs, updates the corresponding input parameters of subsequent models, calls subsequent models to calculate and form the outputs used for subsequent incremental interactions.

[0097] From a layered functional perspective, the model encapsulation module serves as a standardized, pluggable encapsulation layer; the mode determination module serves as a dual-mode adaptive coupling scheduling layer; the automatic data conversion engine in the linkage execution module serves as a fully automatic intelligent data adaptation layer; and the topology construction module and the linkage execution module jointly serve as a multi-model chain-linked scheduling layer. These four layers are connected through a unified calling interface, parameter mapping rules, and model linkage topology, forming a continuous operational process of data input, model computation, data flow, simulation output, and subsequent calculations.

[0098] This system can be deployed on the Yuyu WIM water conservancy digital twin platform, providing multi-model calculation services for the platform's water conservancy "seven conditions and four predictions," flood control scheduling, pump station operation and maintenance, and water resource allocation. Specific business operations provide tasks and operating conditions to the system, which returns linked simulation results for the platform to analyze and display. When the model library is expanded, new models are configured through the model encapsulation module, and the topology construction module organizes them into tasks that meet their input and output conditions, continuing to use existing patterns to determine and link the execution process.

[0099] This application provides a unified computing entry point through a model encapsulation module, an execution and data dependency provision through a topology construction module, and an interaction method adapted to task requirements through a pattern determination module. The linkage execution module completes data conversion and model-driven operations, enabling heterogeneous hydraulic models to run collaboratively on the same platform according to tasks. This reduces model access and maintenance costs, minimizes manual processing of cross-model data, improves the execution efficiency and timeliness of result updates in multi-model simulations, and provides reusable linkage simulation capabilities for hydraulic digital twin operations.

[0100] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

Claims

1. A pluggable coupling scheduling method for a multi-disciplinary water conservancy model, characterized in that, include: Based on the unified interface specification and parameter mapping rules, multiple water conservancy professional models are encapsulated to obtain model components that can be called through the unified interface. The parameter mapping rules are used to establish the correspondence between the input and output parameters of each water conservancy professional model and the platform's standardized parameters. Acquire water conservancy simulation task and basic working condition data, determine the execution dependency and data transmission relationship between the model components based on the water conservancy simulation task, and obtain the model linkage topology; Based on the accuracy requirements, response delay requirements, and current computing power usage of the water conservancy simulation task, the target coupling mode is determined according to the preset coupling mode matching rules. The target coupling mode is either a loosely coupled mode that synchronizes data in batches according to a preset simulation cycle, or a tightly coupled mode that transmits incremental data and updates the model input in real time. Based on the basic working condition data, according to the model linkage topology and the target coupling mode, the model components are called through the unified interface. When data is transferred between model components, the output data of the preceding model components is converted into standardized data according to the parameter mapping rules, and then adapted into input data that meets the format, unit and spatiotemporal reference requirements of the subsequent model components, driving the subsequent model components to run and obtaining the linkage simulation results. When the target coupling mode is a loose coupling mode, the execution process includes: Based on the preset simulation step size and batch synchronization steps, the simulation time range between two adjacent batch synchronizations is determined, and the batch synchronization cycle is obtained. According to the batch synchronization cycle, each model component runs independently and asynchronously under the condition that its own input data is available, so as to obtain the model output data within the cycle. Based on the model linkage topology and the parameter mapping rules, the model output data of the preceding model components within the cycle are collected, standardized, and adapted for input to obtain the batch input data of the corresponding subsequent model components. According to the batch synchronization cycle, the batch input data is synchronized to the corresponding subsequent model components, and the subsequent model components with ready input data are called according to the execution dependency to perform calculations and obtain the phased linkage simulation results. When the target coupling mode is a tightly coupled mode, the execution process includes: Obtain the new output data generated by the preceding model component at the current data interaction moment to obtain the incremental output data; Based on the model linkage topology and the parameter mapping rules, the incremental output data is standardized and adapted for input to subsequent model components to obtain incremental input data. Based on the incremental input data, update the input parameters corresponding to the subsequent model components to obtain the updated model input; Based on the updated model input, the subsequent model component is invoked to perform calculations, obtaining model output data for subsequent incremental interactions, thereby advancing the calculation of the water conservancy simulation task.

2. The method according to claim 1, characterized in that, The process involves encapsulating multiple hydraulic engineering models according to a unified interface specification and parameter mapping rules to obtain model components that can be called through a unified interface, including: Based on the calling protocol and input / output parameters of each water conservancy professional model, the calling method and parameter configuration are determined according to the unified interface specification to obtain the model encapsulation configuration; Based on the model encapsulation configuration, the calling process of each water conservancy professional model is encapsulated through the model adapter, and the parameter mapping rules are configured to obtain the model components to be connected; According to the parameter mapping rules, the output parameter set of the model component to be connected is standardized and transformed, and verified according to the platform's preset field, unit, time series and spatial reference specifications to obtain the parameter verification result; Based on the parameter verification results, the verified model components to be connected are connected to the WIM water conservancy digital twin platform to obtain model components that can be called through a unified interface.

3. The method according to claim 1, characterized in that, The process of acquiring water conservancy simulation task and basic working condition data, determining the execution dependencies and data transfer relationships between the model components based on the water conservancy simulation task, and obtaining the model linkage topology includes: Obtain the water conservancy simulation task and basic working condition data, and determine the model components participating in the linkage calculation and the input and output parameters of each model component based on the water conservancy business processes involved in the water conservancy simulation task, so as to obtain the model task configuration. Based on the model task configuration, establish the correlation between the output parameters of the preceding model components and the input parameters of the following model components to obtain the data transfer relationship between models; Based on the data transfer relationships between the models, the execution priority and serial or parallel execution relationship of each model component are determined, and the execution dependency relationship is obtained. Based on the execution dependencies and the data transfer relationships between models, the model components in the model task configuration are organized to obtain the model linkage topology.

4. The method according to claim 1, characterized in that, The step of determining the target coupling mode according to the accuracy requirements, response delay requirements, and current computing power usage of the water conservancy simulation task, and in accordance with preset coupling mode matching rules, includes: Based on the accuracy requirements, response latency requirements, and current computing power utilization, the accuracy requirement index value, allowable response latency, and current computing power utilization rate are obtained to obtain the mode determination parameters. Based on the mode determination parameters, the results are compared with preset accuracy thresholds, delay thresholds, and computing power utilization thresholds to obtain the comparison results for each threshold. The coupling pattern matching conditions are obtained by combining the results of the comparison of various thresholds. Based on the coupling mode matching conditions, a matching rule is determined from the preset coupling mode matching rules, and the loose coupling mode or tight coupling mode corresponding to the rule is determined as the target coupling mode.

5. The method according to claim 1, characterized in that, The step of converting the output data of the preceding model component into standardized data according to the parameter mapping rules, and then adapting it into input data that conforms to the format, unit, and spatiotemporal reference requirements of the subsequent model component, includes: Based on the parameter mapping rules, determine the field correspondence and unit conversion relationship between the output parameters of the preceding model component and the platform standardized parameters, and obtain the output conversion configuration; According to the output transformation configuration, the output data of the preceding model component is mapped to fields and transformed to units, and aligned according to the time and space references specified by the platform to obtain the standardized data; Based on the correspondence between the input parameters of the subsequent model components and the platform standardization parameters, the data fields, units of measurement, time steps and spatial range required by the subsequent model components are determined to obtain the input adaptation configuration; Based on the input adaptation configuration, the standardized data is subjected to format conversion, unit conversion, temporal alignment, and spatial matching to obtain input data that meets the input requirements of the subsequent model components.

6. The method according to claim 5, characterized in that, Before performing field mapping and unit conversion on the output data of the preceding model component, the following is also included: Based on preset data integrity rules and anomaly detection rules, the output data of the preceding model component is detected to obtain the location of missing data and the location of abnormal data. Based on the location of the abnormal data, the output data of the preceding model component is filtered to obtain the filtered output data. Based on the location of the missing data, the filtered output data is padded with missing data to obtain the data to be converted for the field mapping and unit conversion.

7. The method according to claim 3, characterized in that, The model task configuration includes a hydrological model component, a hydrodynamic model component, and a water conservancy project scheduling model component connected sequentially according to the data transmission direction. The process of performing linked calculations according to the linked topology of the model includes: Based on the basic working condition data, the hydrological model component is invoked to perform runoff generation and runoff calculation, and the runoff generation and runoff calculation results are obtained. Based on the data transfer relationship between the models and the parameter mapping rules, the runoff generation and confluence calculation results are standardized and adapted for the input of the hydrodynamic model components to obtain hydrodynamic input data. Based on the hydrodynamic input data and the basic working condition data, the hydrodynamic model component is invoked to perform hydrodynamic evolution calculations and obtain hydrodynamic evolution results. Based on the data transfer relationship between the models and the parameter mapping rules, the hydrodynamic evolution results are standardized and adapted to the input of the water conservancy project scheduling model components to obtain the engineering scheduling input data. Based on the engineering scheduling input data, the water conservancy engineering scheduling model component is invoked to perform engineering scheduling calculations and obtain engineering scheduling simulation results.

8. A pluggable coupled scheduling system for multi-disciplinary water conservancy models, characterized in that, The system includes: The model encapsulation module is used to encapsulate multiple water conservancy professional models according to the unified interface specification and parameter mapping rules, so as to obtain model components that can be called through the unified interface. The parameter mapping rules are used to establish the correspondence between the input and output parameters of each water conservancy professional model and the platform's standardized parameters. The topology construction module is used to acquire water conservancy simulation task and basic working condition data, determine the execution dependency and data transmission relationship between the model components according to the water conservancy simulation task, and obtain the model linkage topology. The mode determination module is used to determine the target coupling mode according to the accuracy requirements, response delay requirements and current computing power usage of the water conservancy simulation task, and according to the preset coupling mode matching rules. The target coupling mode is either a loose coupling mode that synchronizes data in batches according to a preset simulation cycle, or a tight coupling mode that transmits incremental data and updates the model input in real time. The linkage execution module is used to call the model components through the unified interface according to the basic working condition data, the model linkage topology and the target coupling mode, and when transferring data between model components, convert the output data of the preceding model components into standardized data according to the parameter mapping rules, and then adapt it into input data that meets the format, unit and time-space reference requirements of the subsequent model components, drive the subsequent model components to run, and obtain linkage simulation results. When the target coupling mode is loosely coupled, the linkage execution module is used to execute the following process: Based on the preset simulation step size and batch synchronization steps, the simulation time range between two adjacent batch synchronizations is determined, and the batch synchronization cycle is obtained. According to the batch synchronization cycle, each model component runs independently and asynchronously under the condition that its own input data is available, so as to obtain the model output data within the cycle. Based on the model linkage topology and the parameter mapping rules, the model output data of the preceding model components within the cycle are collected, standardized, and adapted for input to obtain the batch input data of the corresponding subsequent model components. According to the batch synchronization cycle, the batch input data is synchronized to the corresponding subsequent model components, and the subsequent model components with ready input data are called according to the execution dependency to perform calculations and obtain the phased linkage simulation results. When the target coupling mode is a tightly coupled mode, the linkage execution module is used to execute the following process: Obtain the new output data generated by the preceding model component at the current data interaction moment to obtain the incremental output data; Based on the model linkage topology and the parameter mapping rules, the incremental output data is standardized and adapted for input to subsequent model components to obtain incremental input data. Based on the incremental input data, update the input parameters corresponding to the subsequent model components to obtain the updated model input; Based on the updated model input, the subsequent model component is invoked to perform calculations, obtaining model output data for subsequent incremental interactions, thereby advancing the calculation of the water conservancy simulation task.

Images