A digital-twin-based multi-source monitoring data fusion system for a water conservancy plant

Abstract

The application relates to a kind of water conservancy plant multi-source monitoring data fusion systems based on digital twin, belong to water conservancy engineering intelligent control technical field.The system obtains multi-source heterogeneous monitoring data through data twin synchronization module, constructs water conservancy space-time data cube and drives water conservancy digital twin;Water conservancy decision calculation flow module presets water conservancy operator library containing multiple operators, dynamically selects, sorts and topologically connects operators according to event type, instantiates water conservancy business decision calculation flow;Cooperative scheduling module receives and analyzes real-time cooperative scheduling scheme output by decision calculation flow;Efficiency module compares actual response with simulation, evaluates prediction result, and generates efficiency evaluation report containing deviation quantization index.The application realizes efficient fusion of multi-source data of water conservancy plant, accurate mapping of twin and intelligent decision scheduling, improves water conservancy plant control precision and efficiency, and guarantees safe, stable and efficient operation of engineering.

Description

Technical Field

[0001] This invention belongs to the field of intelligent management and control technology for water conservancy projects, specifically relating to a multi-source monitoring data fusion system for water conservancy plants based on digital twins. Background Technology

[0002] With the continuous improvement of the intelligence level of water conservancy projects, water conservancy plants, as the core hubs for water resource regulation, flood control and disaster reduction, and water supply security, have seen their monitoring systems continuously improve. They have gradually formed a multi-dimensional and multi-type heterogeneous monitoring data network covering hydrology, structural safety, equipment status, meteorology, and video. Hydrological monitoring data is used to capture changes in core hydrological elements such as water level, flow rate, and velocity within the basin; structural safety monitoring data is used to monitor the stress, strain, settlement, and crack development of water conservancy structures; equipment operation monitoring data is used to collect operating parameters and fault signals of core facilities such as gates and pumping stations; and meteorological monitoring data is used to understand changes in the surrounding environment such as rainfall, temperature, and wind speed. These multi-type heterogeneous monitoring data collectively constitute the core data foundation for the safety management and scheduling optimization of water conservancy plants, and their level of integration and utilization directly determines the management effectiveness of the water conservancy plants.

[0003] However, the current collection and processing of multi-source monitoring data in water conservancy plants still faces many technical bottlenecks: data collection from different monitoring dimensions relies on independent monitoring terminals and transmission links, lacking unified collection standards and data interaction specifications, resulting in inconsistent data formats and large differences in dimensions, forming a "data silo" phenomenon; the monitoring data contains a large amount of abnormal, missing, and redundant data, and existing data processing methods are mostly simple screening and removal, lacking a systematic and standardized preprocessing process, which makes it impossible to achieve deep integration and collaborative utilization of multi-source data, resulting in the data value not being fully realized and making it difficult to provide comprehensive and accurate data support for the management and control of water conservancy plants.

[0004] Existing water conservancy plant scheduling strategies are mostly based on preset fixed patterns, lacking dynamic adaptive capabilities: scheduling schemes are largely formulated based on historical data and human experience, and cannot be dynamically adjusted according to real-time hydrological risk conditions, equipment operating status, and engineering safety requirements, resulting in problems such as scheduling lag and untimely response; at the same time, the deduction and evaluation of scheduling strategies lack scientific simulation verification carriers, making it impossible to predict the implementation effects of different scheduling schemes in advance, resulting in insufficient rationality and optimization of scheduling schemes, making it difficult to take into account the multiple objectives of water conservancy plants such as flood control and disaster reduction, engineering safety, and operational efficiency, and failing to meet the needs of intelligent and refined management and control of modern water conservancy plants.

[0005] Digital twin technology, as a core technology for realizing real-time mapping and dynamic interaction between physical entities and virtual models, has been widely used in various fields such as intelligent manufacturing, transportation, and energy. It can achieve real-time replication, simulation, and optimized control of the operating status of physical entities by constructing high-fidelity virtual models. However, in the field of multi-source monitoring data fusion and management in water conservancy plants, the application of digital twin technology still faces many unresolved technical challenges: how to construct a high-fidelity digital twin that is precisely synchronized with the physical water conservancy plant to achieve real-time linkage between multi-source heterogeneous monitoring data and virtual models; how to establish a scientific multi-source data fusion mechanism to achieve standardized integration and efficient utilization of data of different types and dimensions; how to combine the digital twin to conduct high-precision hydrological evolution simulation and structural safety diagnosis to achieve accurate quantification of risk status; and how to generate adaptive collaborative scheduling schemes based on real-time risk status and construct a complete closed-loop optimization mechanism to improve the control accuracy and efficiency of water conservancy plants. These problems all restrict the further improvement of the intelligent level of water conservancy plants. Therefore, developing a multi-source monitoring data fusion system for water conservancy plants based on digital twins has significant practical significance and application value. Summary of the Invention

[0006] To address the aforementioned problems in the existing technology, this invention provides a multi-source monitoring data fusion system for water conservancy plants based on digital twins. The objective of this invention can be achieved through the following technical solutions: A multi-source monitoring data fusion system for a water conservancy plant based on a digital twin includes: The data twin synchronization module acquires multi-source heterogeneous monitoring data of water conservancy plants in real time, constructs a water conservancy spatiotemporal data cube containing a real-time monitoring layer, a feature derivation layer, and a forecast assimilation layer, and drives the construction of a water conservancy digital twin that maps the physical state of the physical plant to the hydraulic relationship. The water conservancy decision-making computational flow module has a pre-built water conservancy operator library, including a data assimilation operator that assimilates multi-source data to update the state of the water conservancy digital twin, a hydrological simulation operator that performs one- or two-dimensional coupled hydrodynamic flood evolution simulation on the water conservancy digital twin, a safety assessment operator that performs safety status diagnosis and trend prediction based on structural monitoring time series data, and a scheduling optimization operator that optimizes the coordinated operation strategy of gate groups and pumping station units under multiple objectives. Based on the event type, a set of operators are dynamically selected, sorted, and topologically connected from the water conservancy operator library to instantiate a water conservancy business decision-making computational flow. The collaborative scheduling module receives and parses the real-time collaborative scheduling scheme, which contains a sequence of specific operation instructions, output by the water conservancy business decision calculation stream. The performance module compares the actual response with the prediction results of the hydrological simulation operator and the safety assessment operator, and generates a performance assessment report that includes deviation quantification indicators.

[0007] Specifically, the construction process of the water conservancy spatiotemporal data cube includes: Based on the multi-source heterogeneous monitoring data, a real-time monitoring layer, a feature derivation layer, and a forecast assimilation layer are constructed according to data characteristics. The real-time monitoring layer stores the unprocessed raw monitoring data and uses a time-series database for partitioned storage. The feature derivation layer, based on real-time monitoring data, uses feature engineering algorithms to derive hydraulic correlation features, structural state features, and operational trend features, thus constructing a multi-dimensional feature dataset. The forecast assimilation layer integrates feature-derived layer data with external forecast data, and uses a data assimilation algorithm to complete the assimilation calibration of forecast data and real-time monitoring data to generate forecast assimilation data; The three layers of data are linked and bound together through a spatiotemporal index to construct the water conservancy spatiotemporal data cube.

[0008] Specifically, the data assimilation operator's working process includes: The preprocessed multi-source heterogeneous monitoring data and the state data of the water conservancy digital twin are fused together to construct a data assimilation objective function. The extreme value of the objective function is solved by the least squares method, and the deviation quantification value between the multi-source monitoring data and the state data of the twin is calculated. Based on the deviation quantization value, the gradient descent algorithm is used to dynamically adjust the geometric, hydraulic and physical parameters of the water conservancy digital twin, forming a parameter adjustment feedback link.

[0009] Specifically, the execution process of the one- or two-dimensional coupled hydrodynamic flood evolution simulation includes: By calling the spatial coordinate parameters of the geometric model, the terrain elevation parameters, and the fluid dynamic parameters of the hydraulic physical model of the hydraulic digital twin, a one-dimensional coupled hydrodynamic simulation model is constructed based on the Saint-Venant equations to clarify the coupling boundary conditions between the one-dimensional channel and the two-dimensional region. The finite volume method is used to discretize the hydrodynamic control equations, and the continuous flood evolution process is discretized into discrete spatial nodes and time steps to construct a discretized solution model. By combining one-dimensional river hydraulic boundary conditions with two-dimensional regional hydraulic parameters, the spatiotemporal coupling simulation of flood evolution is completed through numerical iteration. The flood level and flow parameters of each discrete node are calculated in real time, and the temporal variation dataset and spatial distribution heat map data of flood level and flow are output.

[0010] Specifically, the working process of the security assessment operator includes: Receive the structural monitoring time-series data output by the data twin synchronization module, extract trend features, mutation features, and periodic features, and construct a structural safety status feature vector; The extracted feature vectors are compared and analyzed with the preset structural safety level thresholds. The diagnostic accuracy is calculated through the confusion matrix to complete the graded diagnosis of structural safety status. The system performs fitting analysis on the time series data of structural monitoring, sets the prediction step size, predicts the trend of structural safety status changes within a preset time period, and outputs diagnostic results and trend prediction reports including diagnostic level, characteristic parameters, and trend curves.

[0011] Specifically, the specific process of the hydrological evolution simulation includes: The hydrological evolution process is solved by calling the hydrological monitoring parameters, topographic parameters and boundary condition parameters in the standardized fusion dataset. The continuous hydrological evolution process is transformed into computable discrete nodes through discretization processing, and the complex coupling relationship in the water flow process is handled. The hydrological evolution time series change data, spatial distribution data and key node hydrological parameters are output in real time.

[0012] Specifically, the specific working process of the scheduling optimization operator includes: The multi-objective optimization objectives cover three core objectives: flood control safety, engineering operation efficiency, and energy consumption economy. The weight ratio of each objective is determined, and a multi-objective optimization objective function is constructed. Load the real-time operating parameters of the gate group and pumping station units of the water conservancy plant, as well as the operating limits of water conservancy facilities, hydrological environment, and engineering safety constraints, and construct a constraint matrix; The decision variables of gate group opening and pump station unit operating power are optimized and calculated. The set of non-dominated optimal solutions that satisfy all optimization objectives and meet the constraints is generated by iterative solution. Based on the entropy weight method, a comprehensive evaluation of the non-dominated optimal solution set is conducted. Combined with the current operation scenario and management requirements of the water conservancy plant, the optimal coordinated operation strategy of the gate group and pump station units is selected.

[0013] Specifically, the process of dynamically selecting operators in the water conservancy decision-making computational flow module includes: The system acquires real-time operational status data and monitoring and early warning data of the water conservancy plant, uses event recognition algorithms to perform feature matching on the collected data, identifies the current event type or control requirements, and outputs event type identifiers. The mapping relationship library is invoked based on the event type identifier to match the optimal combination of operators corresponding to the event type; the validity of the matched operators is verified, the fitness is calculated using a verification function, and operators with fitness below a preset threshold are removed.

[0014] Specifically, the process of sorting operators in the water conservancy decision-making computational flow module includes: Based on the logic of the entire water conservancy business process, an operator execution logic graph is constructed to clarify the functional dependencies and data interaction relationships of each operator. The selected operators are sorted in an ordered manner using a topological sorting algorithm to generate an operator execution sequence. When there are operators that are independent and can be executed synchronously, they are marked as parallel operator groups. A multi-threaded scheduling mechanism is used to allocate independent execution threads to the parallel operator groups.

[0015] Specifically, the process of topology connection includes: A directed acyclic graph topology connection algorithm is adopted. Based on the pre-defined standardized input and output interfaces of each operator, a data interaction link between operators is constructed, and the transmission direction, data format and transmission rate of the link are defined. Establish a one-to-one correspondence between the output data of the preceding operator and the input data of the following operator, and set the data transmission trigger conditions; at the same time, construct an exception feedback link to monitor the operator execution status and data transmission status in real time.

[0016] Specifically, the process by which the collaborative scheduling module receives and parses the real-time collaborative scheduling scheme includes: The system receives real-time collaborative scheduling schemes output from the computational stream of water conservancy business decisions, performs structured parsing of the schemes, and extracts the operation instruction sequence, execution subject, execution timing, parameter thresholds, and anomaly handling rules from the schemes. The parsed sequence of operation instructions is validated, and the valid operation instructions are sorted according to their execution sequence, converted into a control instruction format that can be recognized by the actuators of the physical hydraulic plant, and transmitted to the corresponding actuators.

[0017] Specifically, the process by which the performance module generates a performance evaluation report includes: The actual response data of the physical hydraulic plant after executing the scheduling plan is obtained in real time. The actual response data is compared with the flood evolution prediction data output by the hydrological simulation operator and the structural safety prediction data output by the safety assessment operator on a parameter-by-parameter basis. The absolute deviation, relative deviation and root mean square error of each parameter are calculated, and a deviation quantification index system is constructed. Based on the deviation quantification index, the execution efficiency and simulation prediction accuracy of the scheduling scheme are comprehensively evaluated, and the efficiency level is divided; an efficiency evaluation report containing deviation quantification index, comprehensive efficiency score and efficiency level is generated.

[0018] The beneficial effects of this invention are as follows: This invention constructs a water conservancy spatiotemporal data cube containing a real-time monitoring layer, a feature derivation layer, and a forecast assimilation layer through a data twin synchronization module. It realizes hierarchical management, feature derivation, and assimilation calibration of multi-source heterogeneous monitoring data, solving the technical pain points of existing water conservancy monitoring data such as dispersion, poor correlation, and insufficient accuracy. It provides multi-dimensional and high-precision data support for the construction and updating of water conservancy digital twins, ensuring that the digital twin can map the physical state and hydraulic connection of the physical factory.

[0019] The water conservancy decision-making computational flow module of this invention has a pre-built diverse operator library, covering four core operators: data assimilation, hydrological simulation, safety assessment, and scheduling optimization. It can dynamically select, sort, and topologically connect operators according to different event types of water conservancy plants, and instantiate and adapt the business decision-making computational flow to the current scenario. This solves the problem of fixed decision-making processes and poor adaptability in traditional water conservancy management and control, and realizes flexible adaptation and efficient execution of decision-making computational processes.

[0020] This invention achieves dynamic optimization of the core parameters of the water conservancy digital twin through the deviation quantization and parameter feedback adjustment mechanism of the data assimilation operator, thereby improving the synchronization accuracy between the twin and the physical entity. Combined with the one- and two-dimensional coupled hydrodynamic simulation of the hydrological simulation operator, it can replicate the flood evolution process and provide scientific support for flood risk prediction.

[0021] This invention's safety assessment operator extracts multi-dimensional features from structural monitoring time-series data and combines them with hierarchical diagnosis and trend prediction to identify structural safety hazards and predict changing trends in a timely and accurate manner. Compared with traditional manual inspection, it improves the efficiency, accuracy, and foresight of structural safety diagnosis, providing a basis for decision-making in the structural safety management of water conservancy projects.

[0022] This invention's scheduling optimization operator takes multi-objective optimization as its core, taking into account flood control safety, engineering operation efficiency, and energy consumption economy. Through optimization calculation and comprehensive evaluation, it selects the optimal collaborative scheduling strategy, solving the problem that traditional scheduling strategies are singular and difficult to balance multi-objective needs. It realizes the collaborative optimization operation of gate groups and pumping station units, reducing energy consumption while improving scheduling safety and efficiency.

[0023] The hydraulic decision-making computation module of this invention optimizes the execution efficiency of the decision-making process and reduces computation time through dynamic operator selection, topology sorting and parallel scheduling mechanisms. At the same time, the abnormal feedback link in the topology connection process ensures the stability and reliability of the collaborative work of operators and avoids the interruption of the entire decision-making process due to the abnormality of a single operator. Attached Figure Description

[0024] To facilitate understanding by those skilled in the art, the present invention will be further described below with reference to the accompanying drawings.

[0025] Figure 1This is an overall architecture diagram of a multi-source monitoring data fusion system for a water conservancy plant based on a digital twin, according to the present invention. Figure 2 This is a diagram of the water conservancy spatiotemporal data cube structure in this invention; Figure 3 This is a flowchart of the generation and execution of the water conservancy decision-making computation flow in this invention. Detailed Implementation

[0026] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided.

[0027] Please see Figures 1-3 A multi-source monitoring data fusion system for a water conservancy plant based on a digital twin, comprising: The data twin synchronization module acquires multi-source heterogeneous monitoring data of water conservancy plants in real time, constructs a water conservancy spatiotemporal data cube containing a real-time monitoring layer, a feature derivation layer, and a forecast assimilation layer, and drives the construction of a water conservancy digital twin that maps the physical state of the physical plant to the hydraulic relationship. The water conservancy decision-making computational flow module has a pre-built water conservancy operator library, including a data assimilation operator that assimilates multi-source data to update the state of the water conservancy digital twin, a hydrological simulation operator that performs one- or two-dimensional coupled hydrodynamic flood evolution simulation on the water conservancy digital twin, a safety assessment operator that performs safety status diagnosis and trend prediction based on structural monitoring time series data, and a scheduling optimization operator that optimizes the coordinated operation strategy of gate groups and pumping station units under multiple objectives. Based on the event type, a set of operators are dynamically selected, sorted, and topologically connected from the water conservancy operator library to instantiate a water conservancy business decision-making computational flow. The collaborative scheduling module receives and parses the real-time collaborative scheduling scheme, which contains a sequence of specific operation instructions, output by the water conservancy business decision calculation stream. The performance module compares the actual response with the prediction results of the hydrological simulation operator and the safety assessment operator, and generates a performance assessment report that includes deviation quantification indicators.

[0028] The data twin synchronization module acquires various types of heterogeneous monitoring data, specifically including hydrological monitoring data for characterizing hydrological elements such as water level, flow rate, and velocity within the watershed of the water conservancy plant; structural safety monitoring data for monitoring structural characteristics such as stress, strain, settlement, and crack development of water conservancy structures; equipment operation monitoring data for collecting operating indicators such as operating speed, load, energy consumption, and fault signals of various water conservancy equipment; video monitoring data for real-time capture of on-site operating conditions, equipment operating status, and surrounding environment of the water conservancy plant; and meteorological monitoring data for collecting surrounding meteorological elements such as rainfall, temperature, wind speed, and humidity. These various types of heterogeneous monitoring data are acquired using a distributed deployment of multi-source acquisition terminals. Depending on the actual monitoring coverage, control priorities, and monitoring accuracy requirements of the water conservancy plant, one or more of the above data types can be flexibly selected for combined acquisition and processing.

[0029] Specifically, the process of constructing the water conservancy spatiotemporal data cube by the data twin synchronization module includes: Based on multi-source heterogeneous monitoring data, a three-tiered system of real-time monitoring, feature derivation, and forecast assimilation layers is constructed according to data characteristics. The real-time monitoring layer stores unprocessed raw monitoring data and uses a time-series database for partitioned storage to ensure real-time writing and rapid reading. The feature derivation layer, based on real-time monitoring data, uses feature engineering algorithms to derive hydraulic correlation features, structural state features, and operational trend features, constructing a multi-dimensional feature dataset. The forecast assimilation layer integrates data from the feature derivation layer with external forecast data and uses data assimilation algorithms to complete the assimilation calibration of forecast data and real-time monitoring data, generating high-precision forecast assimilation data. The three layers of data are linked and bound through a spatiotemporal index to construct a hydraulic spatiotemporal data cube with spatiotemporal correlation and data traceability.

[0030] Specifically, the data assimilation operator's working process includes: The system receives multi-source heterogeneous monitoring data and current status data of the hydraulic digital twin from the data twin synchronization module. It performs standardized preprocessing on both types of data to eliminate dimensional differences and data redundancy. An ensemble Kalman filter assimilation algorithm is then used to fuse the preprocessed multi-source monitoring data and the twin status data, constructing a data assimilation objective function. The extremum of the objective function is solved using the least squares method, and the deviation quantization value between the multi-source monitoring data and the twin status data is calculated. Based on the deviation quantization value, a gradient descent algorithm is used to dynamically adjust the core parameters of the twin, such as geometric, hydraulic, and physical parameters, forming a parameter adjustment feedback loop.

[0031] Specifically, the process by which the hydrological simulation operator performs a one- or two-dimensional coupled hydrodynamic flood evolution simulation includes: By utilizing the spatial coordinate parameters of the geometric model, topographic elevation parameters, and fluid dynamic parameters of the hydraulic physics model of the hydraulic digital twin, a one-dimensional coupled hydrodynamic simulation model is constructed based on the Saint-Venant equations, clarifying the coupling boundary conditions between the one-dimensional river channel and the two-dimensional region. The finite volume method is used to discretize the hydrodynamic control equations, discretizing the continuous flood evolution process into discrete spatial nodes and time steps, and constructing a discretized solution model. Combining the one-dimensional river channel hydraulic boundary conditions (including upstream and downstream water levels and flow boundaries) with the two-dimensional region hydraulic parameters (including roughness and impermeability coefficient), a numerical iterative algorithm is used to complete the spatiotemporal coupling simulation of the flood evolution process, calculate the flood water level and flow parameters of each discrete node in real time, and finally output the temporal variation dataset and spatial distribution heat map data of the flood water level and flow.

[0032] Specifically, the working process of the security assessment operator includes: The system receives structural monitoring time-series data (including time-series sequences of stress, strain, settlement, crack width, etc.) output from the data twin synchronization module. A wavelet denoising algorithm is used to denoise the time-series data, eliminating environmental interference noise. A moving average algorithm is then used to smooth the denoised data, improving data stability. Based on the preprocessed time-series data, a feature extraction algorithm is used to extract core parameters such as trend features, abrupt change features, and periodic features, constructing a structural safety status feature vector. A machine learning-based structural safety status diagnostic model is built, comparing the extracted feature vector with preset structural safety level thresholds. The diagnostic accuracy is calculated using a confusion matrix, completing the structural safety status classification diagnosis (including safe, sub-safe, and dangerous levels). An ARIMA trend fitting algorithm is used to fit and analyze the preprocessed time-series data, setting a prediction step size to predict the structural safety status change trend within a preset time period. The system outputs diagnostic results and a trend prediction report, including the diagnostic level, feature parameters, and trend curve, providing a basis for structural safety management decisions.

[0033] Specifically, the specific working process of the scheduling optimization operator includes: The multi-objective optimization objectives are clearly defined, covering three core objectives: flood control safety, engineering operation efficiency, and energy consumption economy. The weight ratio of each objective is determined using the analytic hierarchy process (AHP), and a multi-objective optimization objective function is constructed. Real-time operating parameters of the hydraulic plant's gate group (including gate opening and opening / closing speed) and pump station units (including operating power and speed) are loaded, along with constraints such as the operating limits of hydraulic facilities, hydrological environment, and engineering safety, to construct a constraint matrix. The NSGA-III multi-objective optimization algorithm is used to optimize decision variables such as gate group opening and pump station unit operating power. Through iterative solving, a set of non-dominated optimal solutions that satisfy all optimization objectives and meet the constraints is generated. Based on the entropy weight method, the set of non-dominated optimal solutions is comprehensively evaluated. Combining the current operating scenario of the hydraulic plant (e.g., flood season, daily operation period) and control requirements, the optimal coordinated operation strategy for the gate group and pump station units is selected, and the strategy parameters and optimization effect evaluation data are output.

[0034] The water conservancy decision computing module pre-sets a mapping relationship library between event types and operator combinations, using a key-value pair storage structure. The event types are divided into four categories based on the water conservancy plant operation scenario: flood warning events, structural safety anomaly events, daily scheduling events, and equipment maintenance events. Each event type corresponds to a pre-set optimal operator combination, which clarifies the composition, execution order, and parameter configuration of the operator combination.

[0035] Specifically, the process of dynamically selecting operators in the water conservancy decision-making computational flow module includes: The system collects real-time operational status and monitoring / early warning data from the water conservancy plant via a data interface. An event recognition algorithm is used to match the features of the collected data, identifying the current event type or control requirement and outputting an event type identifier. Based on the event type identifier, a mapping database is invoked, and a fuzzy matching algorithm is used to find the optimal combination of operators corresponding to that event type. The matched operators are validated for validity, including their operational status, parameter configuration completeness, and compatibility with the current twin state. A validation function is used to calculate the compatibility degree, and operators with compatibility degrees below a preset threshold, those unable to operate normally, or those with insufficient compatibility are removed. Redundancy is eliminated from the remaining valid operators, retaining only the core functional operators to determine the final set of operators to be connected.

[0036] Specifically, the process of sorting operators in the water conservancy decision-making computational flow module includes: Based on the entire process logic of water conservancy operations, an operator execution logic graph is constructed to clarify the functional dependencies and data interaction relationships of each operator. Following the core logic of "data input - data processing - simulation analysis - evaluation and optimization - result output", a topological sorting algorithm is used to sort the selected operators in an ordered manner to generate an operator execution sequence. When there are operators with no dependencies and that can be executed synchronously, they are marked as parallel operator groups. A multi-threaded scheduling mechanism is used to allocate independent execution threads to the parallel operator groups to ensure that parallel operators can be executed synchronously, reduce the overall execution time of the computation flow, and improve the execution efficiency of the decision computation flow. At the same time, the operator sorting results are recorded to form a sorting log.

[0037] Specifically, the process of topology connection includes: A Directed Acyclic Graph (DAG) topology connection algorithm is adopted. Based on the pre-defined standardized input and output interfaces of each operator, a data interaction link between operators is constructed, clarifying the transmission direction, data format, and transmission rate of the link. Through an interface adaptation protocol, a one-to-one correspondence is established between the output data of the preceding operator and the input data of the following operator. Data transmission triggering conditions are set (including the signal indicating the completion of the preceding operator and the signal indicating that the data integrity has passed). A message queue mechanism is used to realize data caching and real-time transmission, ensuring that the output data of the preceding operator can be transmitted to the following operator in real time and accurately after the execution of the preceding operator, realizing the collaborative linkage between operators. At the same time, an anomaly feedback link is constructed to monitor the execution status of operators and the data transmission status in real time. When an operator executes abnormally or data transmission fails, an anomaly feedback signal is immediately triggered and fed back to the anomaly handling unit of the water conservancy decision computing flow module to execute the anomaly retry, operator replacement, or computing flow interruption processing mechanism.

[0038] Specifically, the process by which the collaborative scheduling module receives and parses the real-time collaborative scheduling scheme includes: The system receives real-time collaborative scheduling schemes from the computational flow of water conservancy business decisions via a standardized data interface. A JSON parsing algorithm is used to structure and parse the schemes, extracting the operation instruction sequences, execution entities, execution timing, parameter thresholds, and anomaly handling rules. The parsed operation instruction sequences are validated for accuracy, including instruction format standardization, execution entity matching, and parameter threshold rationality. Invalid instructions are removed, and abnormal instructions are marked. Valid operation instructions are sorted according to execution timing and converted into a control instruction format recognizable by the physical water conservancy plant actuators, which is then transmitted to the corresponding actuators via an industrial bus.

[0039] Specifically, the process by which the performance module generates a performance evaluation report includes: The system collects real-time actual response data from physical hydraulic plants after executing scheduling plans, covering actual hydrological changes, actual structural safety status, and actual equipment operating parameters. It then compares the actual response data with flood evolution prediction data output by the hydrological simulation operator and structural safety prediction data output by the safety assessment operator, parameter by parameter. A deviation calculation algorithm is used to calculate the absolute deviation, relative deviation, and root mean square error of each parameter, constructing a deviation quantification index system. Based on the deviation quantification index, the analytic hierarchy process (AHP) is used to comprehensively evaluate the execution effectiveness and simulation prediction accuracy of the scheduling plan, classifying effectiveness levels. Finally, an effectiveness evaluation report is generated, including deviation quantification indexes, comprehensive effectiveness score, effectiveness level, and optimization suggestions, which is synchronously fed back to the hydraulic decision-making computational flow module, providing data support for operator optimization and computational flow adjustment.

[0040] This embodiment discloses a multi-source monitoring data fusion system for a water conservancy plant based on a digital twin, including a data twin synchronization module, a water conservancy decision computing flow module, a collaborative scheduling module, and an efficiency module.

[0041] 1. Overall system coordination logic The data twin synchronization module acquires multi-source heterogeneous monitoring data (denoted as S, including hydrological, structural safety, equipment operation, meteorological, and other types of data) from the hydraulic plant in real time. It then constructs a hydraulic spatiotemporal data cube comprising a real-time monitoring layer, a feature derivation layer, and a forecast assimilation layer. This cube drives the construction of a hydraulic digital twin (denoted as DT) that maps the physical state of the physical plant to its hydraulic relationship. The module also integrates standardized monitoring data (denoted as S') and the current state data of the twin (denoted as DT). t The data is synchronized to the water conservancy decision-making computational flow module, providing core data input for this module. Based on the input data and the identified event types, the water conservancy decision-making computational flow module dynamically completes operator selection, sorting, and topology connection, instantiates the water conservancy business decision-making computational flow, and outputs a real-time collaborative scheduling scheme (denoted as Plan_opt) to the collaborative scheduling module. The collaborative scheduling module parses the scheme and transmits it to the physical execution mechanism. The efficiency module collects the actual response data of the execution mechanism (denoted as S_act), compares it with the prediction results of the hydrological simulation operator and the safety assessment operator, generates an efficiency assessment report (denoted as Report_eff) containing deviation quantification indicators, and synchronously feeds it back to the water conservancy decision-making computational flow module, providing data support for operator optimization and computational flow adjustment, forming a closed-loop collaborative logic.

[0042] 2. Specific Implementation of the Hydraulic Decision Flow Module As the core decision-making unit of the system, the water conservancy decision computing flow module has the core function of dynamically selecting, sorting and connecting a group of operators based on the pre-built water conservancy operator library and the current event type of the water conservancy plant, instantiating the water conservancy business decision computing flow adapted to the current operating scenario, and outputting a real-time collaborative scheduling scheme that meets the requirements.

[0043] 2.1 Presetting and Initialization of the Hydraulic Operator Library The water conservancy decision-making computational flow module has a pre-built complete water conservancy operator library (denoted as OL). It adopts a modular and scalable architecture and includes four core operators categorized by function. Each operator has a pre-built unified standardized input / output interface, parameter configuration interface, and data interaction protocol to ensure seamless connection and data interoperability between different operators. The specific pre-built and initialization process is as follows: (1) Data assimilation operator (denoted as O1): Its core function is to assimilate multi-source data to update the state of the water conservancy digital twin. It has a built-in ensemble Kalman filter assimilation algorithm and loads preset assimilation parameters (denoted as Para_kal) during initialization. It can receive preprocessed multi-source heterogeneous monitoring data (S') and the current state data (DT) of the water conservancy digital twin output by the data twin synchronization module. t After completing data fusion, deviation quantization, and twin parameter adjustment, it enters a standby state after initialization, waiting for data input and execution commands.

[0044] (2) Hydrological simulation operator (denoted as O2): The core function is to perform a one- or two-dimensional coupled hydrodynamic flood evolution simulation on the hydraulic digital twin. During initialization, a one- or two-dimensional coupled hydrodynamic simulation model, a finite volume discretization algorithm and a numerical iteration algorithm based on the Saint-Venant equations are loaded. The discretization step size (denoted as Δt, Δx) and the coupled boundary condition parameters (denoted as Para_bou) are preset. The geometric model spatial coordinate parameters, topographic elevation parameters and hydrodynamic parameters of the hydraulic physical model can be called. After execution, the flood evolution time series change data and spatial distribution data are output.

[0045] (3) Safety assessment operator (denoted as O3): The core function is to perform safety status diagnosis and trend prediction based on structural monitoring time series data. During initialization, the feature extraction algorithm, structural safety classification diagnosis model and ARIMA trend fitting algorithm are loaded. The structural safety level threshold (denoted as Thr_safe1, Thr_safe2, Thr_safe1>Thr_safe2) and prediction step size (denoted as Step_pre) are preset. It can receive structural monitoring time series data (denoted as S_stru, which includes time series sequences such as stress, strain, settlement, and crack width) output by the data twin synchronization module, and output the diagnosis results and trend prediction report.

[0046] (4) Scheduling optimization operator (denoted as O4): The core function is to optimize the coordinated operation strategy of gate group and pump station unit under multiple objectives. During initialization, the analytic hierarchy process, NSGA-Ⅲ multi-objective optimization algorithm and entropy weight method are loaded. The multi-objective optimization weights are preset (denoted as W_safe, W_eff, W_eco, which correspond to flood scheduling safety, engineering operation efficiency and energy consumption economy, respectively, and W_safe+W_eff+W_eco=1). The operating parameters and constraints of gate group and pump station unit can be loaded, and the optimal coordinated scheduling strategy is output.

[0047] Meanwhile, the operator library has a built-in mapping relationship library (denoted as Map) between event types and operator combinations, which adopts a key-value pair storage structure. The preset event types include flood warning events (T1), structural safety anomaly events (T2), and daily scheduling events (T3). Each event type corresponds to a preset optimal operator combination. The mapping relationship can be dynamically updated according to actual control needs. The specific preset mapping relationship is as follows: T1 (flood warning event) corresponds to {O1,O2,O4}, T2 (structural safety anomaly event) corresponds to {O1,O3}, and T3 (daily scheduling event) corresponds to {O1,O4}, ensuring the adaptability of operator selection.

[0048] 2.2 Dynamic selection of operators The core of dynamic operator selection is to filter out suitable core operators from the operator library based on the current event type of the hydraulic plant: Event type identification: The water conservancy decision-making computational flow module acquires monitoring and early warning data and water conservancy digital twin status data (DT) transmitted in real time from the data twin synchronization module through a standardized data interface. t Combined with its built-in status monitoring unit, it collects the operating status parameters of the water conservancy plant (denoted as P_run, including equipment operating parameters, hydrological parameters, and structural status parameters); it uses an event recognition algorithm (denoted as A_rec) to perform feature matching on the collected data, extracts the event feature vectors from the data (denoted as V_evt=[v1,v2,v3], where v1 is the hydrological feature, v2 is the structural feature, and v3 is the equipment operating feature), and compares V_evt with the preset event type feature library for similarity. The similarity calculation uses the cosine similarity algorithm. When the similarity is higher than the preset threshold (denoted as Thr_sim), it identifies the current event type or control requirement and outputs the event type identifier (denoted as ID_T, with values ​​T1, T2, and T3 corresponding to different event types).

[0049] Operator combination matching: Based on the event type identifier ID_T, call the mapping relationship library Map, and match the optimal operator combination corresponding to the event type through the fuzzy matching algorithm (denoted as A_mat) to obtain the initial operator combination set (denoted as OL_init). For example: when ID_T=T1, OL_init={O1,O2,O4}; when ID_T=T2, OL_init={O1,O3}.

[0050] Operator validity verification and redundancy elimination: Perform validity verification on each operator in OL_init, and adopt a verification function (denoted as F_che=a·Fit_dt+b·Fit_data, where a and b are verification weights, a+b=1, Fit_dt is the fitness of the operator to the current digital twin state, and Fit_data is the fitness of the operator to the current monitoring data), calculate the fitness of each operator (denoted as Fit, with a value range of [0,1]); preset a fitness threshold (denoted as Fit0, Fit0∈[0,1]), when the Fit of an operator is less than Fit0, it is determined that the operator has insufficient fitness and is eliminated; perform redundancy elimination on the remaining adaptive operators, and eliminate auxiliary operators with repeated functions and non-core functions (for example, when O1 already includes the data preprocessing function, the independent preprocessing auxiliary operator is eliminated), and finally determine the set of operators to be connected (denoted as OL_final) to complete the dynamic selection of operators.

[0051] 2.3 Sorting of Operators The core of operator sorting is to clarify the execution order of each operator based on the whole process logic of water conservancy business, so as to realize efficient collaboration. The steps are as follows: Construction of operator execution logic graph: Based on the whole process logic of water conservancy business (data input - data processing - simulation analysis - evaluation and optimization - result output), construct the operator execution logic graph (denoted as G_ope), and clarify the functional dependency and data interaction relationship of each operator in OL_final. The specific dependency relationship is as follows: the data assimilation operator O1 needs to be executed first, and the updated digital twin state data (DT t+1 ) is the core input data for all subsequent operators, providing basic data support for O2, O3 and O4; the hydrological simulation operator O2 and the safety assessment operator O3 have no direct dependency, and both only depend on the output data of O1, so they can be executed in parallel; the scheduling optimization operator O4 can only perform optimization calculation after O2 and O3 are executed, and receives the output flood evolution data and structural safety diagnosis data.

[0052] Topological sorting and parallel labeling: The operators in OL_final are ordered using a topological sorting algorithm (denoted as A_topo). The operator execution logic graph G_ope is traversed, and an ordered operator execution sequence (denoted as Seq_ope) is generated according to the core logic of "data input - data processing - simulation analysis - evaluation and optimization - result output". For example, when OL_final={O1,O2,O3,O4}, Seq_ope=[O1,{O2,O3},O4]; when OL_final={O1,O3}, Seq_ope=[O1,O3]. When there are operators with no dependencies and that can be executed synchronously (such as O2 and O3), they are marked as a parallel operator group (denoted as Group_par). A multi-threaded scheduling mechanism (denoted as M_thread) is used to allocate independent execution threads (denoted as Thread1 and Thread2) to the parallel operator group, ensuring that parallel operators can be executed synchronously, reducing the overall execution time of the water conservancy business decision-making computation flow, and improving decision-making efficiency.

[0053] Sorting result verification: The generated operator execution sequence Seq_ope is logically verified to confirm that there are no dependency conflicts and no execution order errors, and to ensure that the output data of the preceding operator can meet the input requirements of the subsequent operator. After the verification is passed, the final operator execution sequence is determined for subsequent topology connection.

[0054] 2.4 Topological Connections of Operators The core of topology connectivity is to build data interaction links between operators to achieve collaborative operation between them. The steps are as follows: Constructing data interaction links: Using a directed acyclic graph (DAG) topology connection algorithm (denoted as A_dag), based on the standardized input and output interfaces preset by each operator in OL_final, a data interaction link (denoted as Link_data) between operators is constructed. The transmission direction (from the output operator to the input operator), data format (denoted as Form_data, using standardized JSON format), and transmission rate (denoted as Rate_data, dynamically adjusted according to the data volume) of each link are clearly defined to ensure the standardization and real-time performance of data transmission and avoid problems such as data format incompatibility and transmission delay.

[0055] Establish data correspondence and triggering conditions: Through the interface adaptation protocol (denoted as Pro_adap), establish a one-to-one correspondence between the output data of the preceding operator and the input data of the following operator. The specific correspondence is as follows: O1 output data (updated twin state data DT) t+1The data from O2 (flood evolution data S_flood, including time-series variation data and spatial distribution heatmap data) and the output data from O3 (structural safety diagnostic data S_safe, including diagnostic level, characteristic parameters, and trend curves) are used as input data for O4. Simultaneously, data transmission triggering conditions (denoted as Cond_trig) are set, including a signal indicating the completion of the preceding operator (Sig_fin) and a signal indicating successful data integrity verification (Sig_chk). A message queue mechanism (denoted as Q_msg) is used to achieve data caching and real-time transmission, avoiding data loss or transmission delays, and ensuring that the output data of the preceding operator can be transmitted to the subsequent operator in real time and accurately after its execution.

[0056] Constructing an anomaly feedback link: An anomaly feedback link (denoted as Link_err) is constructed synchronously to monitor the execution status of each operator (denoted as State_ope, with values ​​of "running", "execution completed", and "execution exception") and the data transmission status (denoted as State_link, with values ​​of "transmitting", "transmission successful", and "transmission failed") in real time. When an operator experiences an execution anomaly (such as a calculation error or execution timeout) or data transmission fails, an anomaly feedback signal (Sig_err) is immediately triggered and fed back to the anomaly handling unit (denoted as Unit_err) of the hydraulic decision-making computational flow module. This unit then performs handling mechanisms such as anomaly retry, replacement with a backup operator, or interruption of the computational flow to ensure the stability and reliability of the topology connection and avoid the interruption of the entire decision-making process due to an anomaly of a single operator.

[0057] 2.5 Instantiation and Execution of the Computational Flow for Water Conservancy Business Decisions Based on the above operator selection, sorting, and topology connection results, the instantiation and execution of the water conservancy business decision-making computation flow are completed, and a real-time collaborative scheduling scheme is output. The specific process is as follows: Computational flow topology model construction and resource allocation: Based on the sorted operator execution sequence Seq_ope and topology connection link Link_data, a computational flow topology model for water conservancy business decisions is constructed (denoted as Model_flow). A resource scheduling algorithm (denoted as A_res) is used to allocate independent execution resources (denoted as Res_ope, including CPU, memory, etc.) to each operator. The execution priority is set according to the operator function priority (denoted as Pri_ope), where simulation operators (O2) and evaluation operators (O3) have higher priority than data processing operators (O1) and optimization operators (O4) to ensure that core functions are executed first and improve the execution efficiency of the computational flow.

[0058] Execution environment initialization and interface integration: Load the preset parameters, threshold configurations, and built-in algorithm models of each operator, initialize the operator execution environment (denoted as Env_ope), complete the data interface integration between each operator and the water conservancy digital twin DT and data twin synchronization module to ensure smooth data interaction; at the same time, initialize the computation flow execution monitoring mechanism (denoted as Mon_flow), adopt the heartbeat detection algorithm (denoted as A_heart), and monitor the execution progress (denoted as Pro_ope), data transmission status State_link, and execution results (denoted as Res_ope) of each operator in real time at a preset detection frequency (denoted as Freq_heart), and record key data such as operator execution time and calculation accuracy (denoted as Data_rec) for subsequent performance evaluation and operator optimization.

[0059] Computational flow execution and result output: Operators are started according to the operator execution sequence Seq_ope, and the operator functions are executed sequentially (or in parallel).

[0060] 1. O1 Execution: Receive S' and DT t The two types of data are fused to construct a data assimilation objective function (denoted as F_kal=Σ(S'_i-DT)). t _i)², where S'_i is the i-th parameter of the preprocessed multi-source monitoring data, and DT t Let _i be the i-th parameter of the current state of the digital twin. The objective function's extreme value is solved using the least squares method, and the quantized deviation value (denoted as Dev_kal) between the multi-source monitoring data and the digital twin's state data is calculated. Based on Dev_kal, the geometric, hydraulic, and physical parameters of the hydraulic digital twin are dynamically adjusted using the gradient descent algorithm, forming a parameter adjustment feedback loop, and outputting the updated digital twin state data DT. t+1 .

[0061] 2. O2 Execution: Based on DT t+1 By utilizing the spatial coordinate parameters of the geometric model, topographic elevation parameters, and fluid dynamic parameters of the hydraulic physics model of the hydraulic digital twin, a one-dimensional coupled hydrodynamic simulation model is constructed based on the Saint-Venant equations, clarifying the coupling boundary conditions between the one-dimensional river channel and the two-dimensional region. The finite volume method is used to discretize the hydrodynamic control equations, discretizing the continuous flood evolution process into discrete spatial nodes and time steps, and constructing a discretized solution model. Combining the one-dimensional river channel hydraulic boundary conditions and the two-dimensional region hydraulic parameters, the spatiotemporal coupling simulation of the flood evolution process is completed through numerical iteration, and the flood water level and flow parameters of each discrete node are calculated in real time, outputting the temporal variation dataset of flood water level and flow and the spatial distribution heat map data (S_flood).

[0062] 3. Execution of O3: receive S_stru, extract trend features, mutation features and periodic features, and construct a structural safety state feature vector (denoted as V_safe); perform comparative analysis on V_safe and preset structural safety level thresholds (Thr_safe1, Thr_safe2), calculate diagnostic accuracy through a confusion matrix, and complete the hierarchical diagnosis of structural safety status (V_safe ≥ Thr_safe1 indicates safe, Thr_safe2 < V_safe < Thr_safe1 indicates sub-safe, and V_safe ≤ Thr_safe2 indicates dangerous); perform ARIMA fitting analysis on S_stru, set the prediction step size Step_pre, predict the change trend of structural safety status in a preset future period, and output a diagnostic result and trend prediction report (S_safe) including diagnostic grade, characteristic parameters and trend curve.

[0063] 4. Execution of O4: define the multi-objective optimization objectives (flood dispatching safety, project operation efficiency, energy consumption economy), and construct a multi-objective optimization objective function based on preset weights (W_safe, W_eff, W_eco) (denoted as F_opt=W_safe·F_safe+W_eff·F_eff+W_eco·F_eco, where F_safe is the safety objective function, F_eff is the efficiency objective function, and F_eco is the economic objective function); load real-time operation parameters of gate groups and pump station units in water conservancy projects, as well as operation limits of water conservancy facilities, hydrological environment and engineering safety constraints, and construct a constraint condition matrix (denoted as Mat_con); perform optimization calculation on the decision variables of gate group opening and pump station unit operation power, iteratively solve through the NSGA-Ⅲ algorithm, and generate a non-dominated optimal solution set that meets all optimization objectives and conforms to the constraint conditions; perform comprehensive evaluation on the non-dominated optimal solution set based on the entropy weight method, combine the current operation scenario and control requirements of the water conservancy project, screen out the cooperatively operation strategy of gate groups and pump station units with optimal adaptability, and output the real-time cooperative dispatching scheme Plan_opt.

[0064] Completion check of calculation flow execution: when all operators have been executed and data transmission is correct, complete the instantiation and execution of the water conservancy business decision calculation flow, synchronize Plan_opt to the cooperative dispatching module through a standardized interface to realize decision output, and synchronize the calculation flow execution data (Data_rec) to the effectiveness module to provide data support for subsequent effectiveness evaluation.

[0065] 2.6 Operator optimization and dynamic adjustment of calculation flow To ensure the adaptability and execution effectiveness of the water conservancy decision calculation flow module, operator optimization and dynamic adjustment of calculation flow are realized in combination with Report_eff fed back by the effectiveness module, and the specific process is as follows: Operator performance analysis: The hydraulic decision-making computational flow module receives the Report_eff feedback from the performance module, extracts the deviation quantification index (denoted as Ind_dev, including absolute deviation, relative deviation, and root mean square error) from the report, and analyzes the execution performance, computational accuracy, and adaptability of each operator. When the computational accuracy of a certain operator is lower than the preset threshold (denoted as Acc0) or the adaptability decreases, the built-in algorithm parameters of the operator are dynamically adjusted. For example, the data assimilation algorithm weight of O1, the discretization step size (Δt, Δx) of O2, the feature extraction threshold of O3, and the optimization target weight (W_safe, W_eff, W_eco) of O4 are adjusted to improve the execution accuracy and adaptability of the operator.

[0066] Dynamic adjustment of the computational flow: When the operating status of the water conservancy plant changes abruptly, the event type is updated (ID_T changes), or the control requirements are adjusted, the water conservancy decision computational flow module immediately interrupts the currently executing computational flow, saves the current execution state and intermediate data (denoted as Data_mid), re-executes the dynamic selection, sorting, and topology connection process of operators, and instantiates a new water conservancy business decision computational flow (denoted as Flow_new); after the new computational flow is instantiated, Data_mid is restored and execution continues, realizing seamless switching of the computational flow, ensuring real-time adaptation with the actual operation scenario, and further improving the core functions of the water conservancy decision computational flow module.

[0067] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A multi-source monitoring data fusion system for a water conservancy plant based on digital twins, characterized in that, include: The data twin synchronization module acquires multi-source heterogeneous monitoring data from water conservancy plants in real time and constructs a water conservancy spatiotemporal data cube containing a real-time monitoring layer, a feature derivation layer, and a forecast assimilation layer. A data assimilation objective function is constructed for the preprocessed multi-source heterogeneous monitoring data and the state data of the water conservancy digital twin. The extreme value of the objective function is solved to obtain the deviation quantification value between the multi-source monitoring data and the state data of the twin. Based on the deviation quantification value, the geometric parameters, hydraulic parameters and physical parameters of the water conservancy digital twin are dynamically adjusted to form a parameter adjustment feedback link, which drives the construction of a water conservancy digital twin that maps the physical state and hydraulic relationship of the physical plant. The water conservancy decision-making computational flow module has a pre-built water conservancy operator library, including a data assimilation operator that assimilates multi-source data to update the state of the water conservancy digital twin, a hydrological simulation operator that performs one- or two-dimensional coupled hydrodynamic flood evolution simulation on the water conservancy digital twin, a safety assessment operator that performs safety status diagnosis and trend prediction based on structural monitoring time series data, and a scheduling optimization operator that optimizes the coordinated operation strategy of gate groups and pumping station units under multiple objectives. Based on the event type, a set of operators are dynamically selected, sorted, and topologically connected from the water conservancy operator library to instantiate a water conservancy business decision-making computational flow. The specific process of dynamically selecting operators in the water conservancy decision computational flow module includes: The system acquires real-time operational status data and monitoring and early warning data of the water conservancy plant. It uses an event recognition algorithm to perform feature matching on the collected data, identifies the current event type or control requirements, and outputs an event type identifier. Based on the event type identifier, it calls the mapping relationship library to match the optimal combination of operators corresponding to the event type. It performs validity verification on the matched operators, calculates the fit, and removes operators with a fit below a preset threshold. The specific process of sorting operators in the water conservancy decision-making computational flow module includes: Based on the logic of the entire water conservancy business process, an operator execution logic graph is constructed to clarify the functional dependencies and data interaction relationships of each operator. The selected operators are ordered by a topological sorting algorithm to generate an operator execution sequence. When there are operators with no dependencies and that can be executed synchronously, they are marked as parallel operator groups. A multi-threaded scheduling mechanism is used to allocate independent execution threads to the parallel operator groups. The specific process of topology connection includes: Based on the standardized input and output interfaces preset by each operator, a data interaction link between operators is constructed, and the transmission direction, data format and transmission rate of the link are defined; a one-to-one correspondence between the output data of the preceding operator and the input data of the following operator is established, and data transmission trigger conditions are set; at the same time, an exception feedback link is constructed to monitor the operator execution status and data transmission status in real time. The collaborative scheduling module receives and parses the real-time collaborative scheduling scheme, which contains a sequence of specific operation instructions, output by the water conservancy business decision calculation stream. The performance module compares the actual response with the prediction results of the hydrological simulation operator and the safety assessment operator, and generates a performance assessment report that includes deviation quantification indicators.

2. The system according to claim 1, characterized in that, The specific construction process of the water conservancy spatiotemporal data cube includes: Based on the aforementioned multi-source heterogeneous monitoring data, a real-time monitoring layer, a feature derivation layer, and a forecast assimilation layer are constructed according to data characteristics. The real-time monitoring layer stores the original monitoring data and uses a time-series database for partitioned storage. The feature derivation layer extracts hydraulic correlation features, structural state features, and operational trend features based on real-time monitoring data to construct a multi-dimensional feature dataset. The forecast assimilation layer fuses feature-derived layer data with external forecast data to complete the assimilation calibration of forecast data and real-time monitoring data, and generates forecast assimilation data; The three layers of data are linked and bound together through a spatiotemporal index to construct the water conservancy spatiotemporal data cube.

3. The system according to claim 1, characterized in that, The specific execution process of the one- or two-dimensional coupled hydrodynamic flood evolution simulation includes: By calling the geometric model spatial coordinate parameters, topographic elevation parameters, and fluid dynamic parameters of the hydraulic physical model of the hydraulic digital twin, a one-dimensional coupled hydrodynamic simulation model is constructed to clarify the coupling boundary conditions between the one-dimensional river channel and the two-dimensional region. Discretize the hydrodynamic control equations, and deconstruct the continuous flood evolution process into discrete spatial nodes and time steps to construct a discretized solution model. By combining one-dimensional river hydraulic boundary conditions with two-dimensional regional hydraulic parameters, the spatiotemporal coupling simulation of flood evolution is completed through numerical iteration. The flood level and flow parameters of each discrete node are calculated in real time, and the temporal variation dataset and spatial distribution heat map data of flood level and flow are output.

4. The system according to claim 1, characterized in that, The specific working process of the security assessment operator includes: Receive the structural monitoring time-series data output by the data twin synchronization module, extract trend features, mutation features, and periodic features, and construct a structural safety status feature vector; The extracted feature vectors are compared and analyzed with the preset structural safety level thresholds. The diagnostic accuracy is calculated through the confusion matrix to complete the graded diagnosis of structural safety status. The system performs fitting analysis on the time series data of structural monitoring, sets the prediction step size, predicts the trend of structural safety status changes within a preset time period, and outputs diagnostic results and trend prediction reports.

5. The system according to claim 1, characterized in that, The water conservancy decision computing flow module has a preset mapping relationship library between event types and operator combinations. The event types are divided based on the operation scenarios of water conservancy plants. Each event type corresponds to a preset optimal operator combination, which clarifies the composition, execution order and parameter configuration of the operator combination.

6. The system according to claim 1, characterized in that, The specific working process of the scheduling optimization operator includes: The multi-objective optimization objectives cover three core objectives: flood control safety, engineering operation efficiency, and energy consumption economy. The weight ratio of each objective is determined, and a multi-objective optimization objective function is constructed. Load the real-time operating parameters of the water conservancy plant's gate group and pump station units, and construct a constraint matrix; The decision variables of gate group opening and pump station unit operating power are optimized and calculated. The set of non-dominated optimal solutions that satisfy all optimization objectives and meet the constraints is generated by iterative solution. A comprehensive evaluation of the non-dominated optimal solution set is conducted, and the optimal coordinated operation strategy of gate group and pump station unit is selected based on the current operation scenario and management requirements of the water conservancy plant.

7. The system according to claim 1, characterized in that, The specific process by which the performance module generates a performance evaluation report includes: The actual response data of the physical hydraulic plant after executing the scheduling plan is obtained in real time. The actual response data is compared with the flood evolution prediction data output by the hydrological simulation operator and the structural safety prediction data output by the safety assessment operator on a parameter-by-parameter basis. The absolute deviation, relative deviation and root mean square error of each parameter are calculated, and a deviation quantification index system is constructed. Based on the deviation quantification index, the execution efficiency and simulation prediction accuracy of the scheduling scheme are comprehensively evaluated, and the efficiency level is divided; an efficiency evaluation report containing deviation quantification index, comprehensive efficiency score and efficiency level is generated.

Images