Water conservancy dispatching processing method and system based on digital twinning

Abstract

The application provides a water conservancy dispatching processing method and system based on digital twinning, which comprises the following steps: acquiring multi-source sensing data corresponding to historical extreme climate events, constructing a historical event resource library, acquiring real-time water conservancy monitoring data, matching the real-time water conservancy monitoring data with the historical event resource library, acquiring current water conservancy project operation data for mapping and adaptation, generating a candidate set of dispatching instructions, inputting the candidate set of dispatching instructions into a digital twinning river basin model, simulating a whole-process operation process of the water conservancy project under the guidance of the candidate set of dispatching instructions, obtaining simulation operation results, identifying abnormal change nodes with parameter change rates exceeding a preset range, locating dispatching operation items causing the abnormal change nodes and correcting parameter values of the dispatching operation items, inputting the corrected candidate set of dispatching instructions into the digital twinning river basin model again for simulation operation calculation, and outputting a water conservancy dispatching scheme. The application can gradually optimize the dispatching scheme and improve the stability and effectiveness of the scheme in actual application.

Description

Technical Field

[0001] This application relates to the field of data processing, and in particular to a water conservancy scheduling and processing method and system based on digital twins. Background Technology

[0002] Water conservancy scheduling is a key technology for ensuring the rational allocation of water resources in a river basin, flood control and disaster reduction, and the safe operation of engineering projects. In existing technologies, the formulation of water conservancy scheduling schemes largely relies on summaries of historical experience and simple comparisons of real-time monitoring data. This involves analyzing the correspondence between historical hydrological data and scheduling cases, combining real-time monitoring information such as water level and flow, and generating operational instructions based on preset scheduling rules. Some technologies attempt to introduce digital twin technology to construct virtual models of the river basin to simulate the execution process of scheduling schemes. However, in practical applications, the lack of correlation between the evolutionary patterns and spatial impact range of historical events and scheduling operations, coupled with the limited accuracy of matching real-time data with historical cases, limits the reference value. Furthermore, the generation of scheduling schemes often fails to fully consider the actual operating status and constraints of current water conservancy projects. The simulation verification process lacks the ability to reproduce the entire process of engineering operation and dynamic hydrological changes, meaning that the adaptability and reliability of scheduling schemes in response to extreme weather events or complex engineering conditions need improvement. Summary of the Invention

[0003] This invention provides a water conservancy scheduling and processing method and system based on digital twins.

[0004] In a first aspect, embodiments of the present invention provide a water conservancy scheduling processing method based on digital twins. The method includes: acquiring multi-source sensing data corresponding to historical extreme climate events; based on the multi-source sensing data, sorting out the continuous change patterns of the events on a time scale and the coverage characteristics on a spatial scale; associating the continuous change patterns, spatial coverage characteristics, and corresponding water conservancy scheduling operation records; constructing a historical event resource library containing the event's time change patterns, spatial coverage characteristics, and associated scheduling operation records; wherein the multi-source sensing data includes meteorological sensing data, hydrological sensing data, and water conservancy project operation sensing data; acquiring real-time water conservancy monitoring data; matching the real-time water conservancy monitoring data with the event's time change patterns and spatial coverage characteristics in the historical event resource library; calculating the matching degree between the real-time water conservancy monitoring data and each historical event through comparison of the time and spatial dimensions; filtering similar historical events whose matching degree meets preset requirements; extracting the scheduling operation records associated with similar historical events and organizing them into a reference scheduling operation sequence in chronological order; acquiring current water conservancy project operation data; and so on. The water conservancy project operation data includes the current operating status parameters and operable constraints of the water conservancy facilities. A reference scheduling operation sequence is mapped and adapted to the current water conservancy project operation data. Based on the current operating status parameters and operable constraints, the parameter values ​​of each operation item in the reference scheduling operation sequence are adjusted to generate a candidate set of scheduling instructions that meets the current project operation conditions. This candidate set of scheduling instructions is then input into a digital twin watershed model to perform simulation calculations, simulating the entire process of the water conservancy project's operation under the guidance of the candidate set of scheduling instructions. Real-time recording of the water conservancy project's operating status changes and the watershed's hydrological dynamic changes output by the digital twin watershed model yields simulation results containing parameter change trends. The simulation results are analyzed to identify abnormal change nodes where the parameter change rate exceeds a preset range. The scheduling operation items causing these abnormal change nodes are located and their parameter values ​​are corrected. The corrected candidate set of scheduling instructions is then input back into the digital twin watershed model for simulation calculations. This correction and simulation process is repeated until no abnormal change nodes are found in the simulation results, at which point a water conservancy scheduling scheme is output.

[0005] Secondly, embodiments of the present invention provide a water conservancy scheduling system, comprising: a memory for storing computer-executable instructions or computer programs; and a processor for executing the computer-executable instructions or computer programs stored in the memory to implement the above-mentioned water conservancy scheduling processing method based on digital twins.

[0006] The embodiments of this application have the following beneficial effects:

[0007] This invention acquires multi-source sensing data of historical extreme weather events, analyzes the continuous temporal variation patterns and spatial coverage characteristics of these events, and associates these spatiotemporal characteristics with corresponding scheduling operation records to construct a historical event resource library. This systematically integrates scattered historical data and operational experience, providing a structured reference for the formulation of scheduling schemes. By comparing real-time water conservancy monitoring data with the temporal variation patterns and spatial coverage characteristics in the historical event resource library in both time and space dimensions to calculate the matching degree, similar historical events are screened out and reference scheduling operation sequences are extracted. This allows for a more comprehensive capture of the similarity of event characteristics and improves the fit between the reference sequences and the current actual situation. By acquiring current water conservancy project operation data, the reference scheduling operation sequences are compared with current operating status parameters and operable parameters. By mapping and adapting the scheduling range and adjusting the operation item parameters to generate a candidate set of scheduling instructions, historical experience can be combined with the actual operating conditions of the current project, ensuring that the scheduling instructions are in line with the current operating capacity of the project. By inputting the candidate set of scheduling instructions into the digital twin watershed model to perform simulation calculations, and recording the changes in the operating status of the water conservancy project and the dynamic changes in the watershed hydrology in real time, the entire process of the scheduling scheme can be reproduced in the virtual environment, providing detailed data support for evaluating the effectiveness of the scheme. By analyzing the simulation results to identify abnormal change nodes where the parameter change rate exceeds the preset range, the scheduling operation item causing the abnormality can be located and its parameter value corrected. By repeating the correction and simulation process until there are no abnormal nodes, the scheduling scheme can be gradually optimized, improving the stability and effectiveness of the scheme in practical applications. Attached Figure Description

[0008] Figure 1 This is a schematic diagram of the architecture of the application scenario provided in the embodiments of this application;

[0009] Figure 2 This is a schematic diagram of the structure of the water conservancy dispatching system provided in the embodiments of this application;

[0010] Figure 3 This is a schematic flowchart of the water conservancy scheduling and processing method based on digital twins provided in the embodiments of this application. Detailed Implementation

[0011] See Figure 1 , Figure 1 This is a schematic diagram of the application scenario provided in the embodiments of this application. The data acquisition terminal 400 is connected to the water conservancy dispatching system 200 through the network 300. The network 300 can be a wide area network or a local area network, or a combination of the two, and data transmission is achieved using wireless or wired links.

[0012] The data acquisition terminal 400 is used to collect multi-source sensing data, such as meteorological sensing data, hydrological sensing data and water conservancy project operation data, and send the multi-source sensing data to the water conservancy dispatch system 200.

[0013] In some embodiments, the water conservancy dispatching system 200 can be a standalone physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms. The data acquisition terminal 400 can be various sensors or databases.

[0014] The following describes the water conservancy dispatching system that implements the digital twin-based water conservancy dispatching processing method provided in the embodiments of this application. See also Figure 2 , Figure 2 This is a schematic diagram of the structure of the water conservancy dispatching system provided in the embodiments of this application. Figure 2 The illustrated water conservancy dispatching system includes at least one processor 210, a memory 250, at least one network interface 220, and an external interface 230. The various components in the water conservancy dispatching system 200 are coupled together via a bus system 240. It is understood that the bus system 240 is used to implement communication between these components. In addition to a data bus, the bus system 240 also includes a power bus, a control bus, and a status signal bus. However, for clarity, in… Figure 2 The general labeled all buses as Bus System 240.

[0015] Processor 210 can be an integrated circuit chip with signal processing capabilities, such as a general-purpose processor, a digital signal processor (DSP), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Among them, the general-purpose processor can be a microprocessor or any conventional processor, etc.

[0016] External interface 230 may include, for example, one or more speakers and / or one or more visual displays. External interface 230 may also include one or more input devices 432, such as a keyboard, mouse, microphone, touch screen display, camera, etc.

[0017] The memory 250 may be removable, non-removable, or a combination thereof. Exemplary hardware devices include solid-state storage, hard disk drives, optical disk drives, etc. The memory 250 may optionally include one or more storage devices physically located away from the processor 210.

[0018] The memory 250 may include volatile memory or non-volatile memory, or both. The non-volatile memory may be read-only memory (ROM), and the volatile memory may be random access memory (RAM). The memory 250 described in this application embodiment is intended to include any suitable type of memory.

[0019] In some embodiments, memory 250 is capable of storing data to support various operations, examples of which include programs, modules, and data structures or subsets or supersets thereof, as illustrated below.

[0020] Operating system 251 includes system programs for handling various basic system services and performing hardware-related tasks, such as the framework layer, core library layer, driver layer, etc., for implementing various basic business functions and handling hardware-based tasks;

[0021] The network communication module 252 is used to reach the data acquisition end via one or more (wired or wireless) network interfaces 220. Exemplary network interfaces 220 include Bluetooth, WiFi, and Universal Serial Bus (USB), etc.

[0022] Presentation module 253 is configured to enable the display of information (e.g., external interface for operating peripheral devices and displaying content and information) via one or more output devices 231 (e.g., display screen, speaker, etc.) associated with external interface 230;

[0023] The input processing module 254 is used to detect and translate one or more user inputs or interactions from one or more input devices 232.

[0024] Based on the above description of the application scenarios and water conservancy dispatching systems provided in the embodiments of this application, the following describes the water conservancy dispatching processing method based on digital twins provided in the embodiments of this application. In actual implementation, the water conservancy dispatching processing method based on digital twins provided in the embodiments of this application can be implemented by a water conservancy dispatching system. See also Figure 3 , Figure 3 This is a schematic flowchart of the water conservancy scheduling and processing method based on digital twins provided in the embodiments of this application. Next, it will be combined with... Figure 3 The steps shown are explained.

[0025] Step S100: Obtain multi-source sensing data corresponding to historical extreme climate events. Based on the multi-source sensing data, sort out the continuous change patterns of the events on the time scale and the coverage characteristics on the spatial scale. Associate the continuous change patterns, spatial coverage characteristics and corresponding water conservancy scheduling operation records to construct a historical event resource library containing the event's time change patterns, spatial coverage characteristics and associated scheduling operation records. The multi-source sensing data includes meteorological sensing data, hydrological sensing data and water conservancy project operation sensing data.

[0026] Multi-source sensing data refers to data acquired from various channels that reflects information related to historical extreme climate events. Meteorological sensing data describes climate and meteorological characteristics, such as observed values ​​of meteorological elements like temperature, precipitation, and wind speed, reflecting the meteorological conditions during extreme climate events. Hydrological sensing data covers hydrological conditions, including measurements of hydrological elements such as water level, flow rate, and water quality, reflecting changes in rivers, lakes, and other water bodies during extreme climate events. Water conservancy project operation sensing data pertains to the operational status of water conservancy facilities, such as reservoir storage capacity, sluice gate opening, and pumping station operating power. Obtaining multi-source sensing data corresponding to historical extreme climate events can be achieved in various ways. For meteorological sensing data, historical meteorological observation records can be obtained from meteorological monitoring stations. These stations utilize various meteorological sensors, such as thermometers, rain gauges, and anemometers, to collect, record, and store meteorological element data in real time. For hydrological sensing data, measurement data such as water level and flow rate can be collected from hydrological monitoring stations, which can use equipment such as water level gauges and flow meters for data acquisition. For the operational sensing data of water conservancy projects, information such as reservoir water storage and sluice gate opening can be extracted from the automated control system of the water conservancy project. These systems record the operational status parameters of the engineering facilities in real time.

[0027] In one implementation, step S100 may specifically include the following steps S110 to S160:

[0028] Step S110: Perform hierarchical spatiotemporal fusion of multi-source sensing data. Perform time series alignment and outlier suppression within the respective modalities of meteorological sensing data, hydrological sensing data, and water conservancy project operation sensing data. Calculate the correlation weights of different modal data through a cross-modal attention mechanism to generate a fusion data cube with spatiotemporal consistency. Each data unit of the fusion data cube contains multimodal sensing values ​​and correlation confidence.

[0029] Hierarchical spatiotemporal fusion integrates multi-source sensing data across time and space, enabling the analysis of different data types within a unified spatiotemporal framework. Time series alignment matches and unifies timestamps from different data sources, ensuring comparability of data at the same point in time. Outlier suppression identifies and handles outliers to prevent them from affecting subsequent analysis results. Statistical methods, such as standard deviation-based methods, can be used to identify outliers. Data points exceeding a certain standard deviation range are considered outliers and processed accordingly, such as replacing them with the average of adjacent normal data. Cross-modal attention mechanisms automatically focus on important correlations between different modalities, assigning different weights based on data relevance. Generating a spatiotemporally consistent fused data cube involves organizing the time-series aligned and outlier-suppressed multi-source sensing data along time and space dimensions to form a three-dimensional data cube. Each data unit contains multimodal sensing values ​​and correlation confidence scores, where correlation confidence scores represent the reliability of the correlation between different modalities within that data unit.

[0030] Step S120: The dynamic time warp and multi-scale window joint analysis method is used to mine the time features of the fused data cube. The pattern sequence of data change is extracted at multiple time granularities. The stage transformation characteristics of the event development are identified by sequence pattern matching. Combining the duration and magnitude of the transformation characteristics, an event time stage division model including the initiation period, development period, peak period and decline period is constructed to sort out the continuous change pattern of the event.

[0031] In this step, dynamic time warping is used to match and compare data sequences within the fused data cube at different time granularities to discover patterns of data change. Multi-scale window joint analysis analyzes the data across multiple different time windows to capture data characteristics at different time scales. By setting time windows of different sizes, short-term and long-term trends in data can be observed.

[0032] Extracting pattern sequences of data changes at multiple time granularities involves dividing the data in the fused data cube according to different time intervals, and then using a dynamic time warping algorithm to find patterns of data change at each time granularity. Sequence pattern matching involves matching the extracted pattern sequences with predefined event development stage patterns to identify the phase transition characteristics of the event's development. For example, if predefined pattern sequences represent the initiation, development, peak, and decline phases, comparing the extracted pattern sequences with these predefined patterns determines which development stage the current data is in. Combining the duration and magnitude of change characteristics takes into account both the duration of the event's development stage and the magnitude of data change. For example, during the peak phase, the data change magnitude may be larger and the duration relatively longer; while during the initiation phase, the data change magnitude is smaller and the duration shorter. By comprehensively considering both duration and magnitude of change, the time stages of the event can be more accurately segmented.

[0033] Step S130: Extract spatial topological features from the fused data cube, generate a state classification map of the watershed monitoring unit using an adaptive threshold segmentation algorithm, calculate the morphological parameters and spatial adjacency relationships of the abnormal state regions in the classification map, construct a spatial topological network of the abnormal regions using graph theory, where network nodes represent abnormal sub-regions and edge weights represent the connectivity strength between sub-regions, and use the overall structural features of the topological network as the spatial coverage features of the event.

[0034] Spatial topology feature extraction extracts feature information related to spatial structure and relationships from the fused data cube. A watershed monitoring unit is a monitoring area obtained after dividing a watershed, and each monitoring unit contains multi-source sensing data such as meteorological, hydrological, and water conservancy project operation data within that area. The adaptive threshold segmentation algorithm is an algorithm that automatically determines the segmentation threshold based on the statistical characteristics of the data. In this step, this algorithm is used to segment the data in the fused data cube, classifying the watershed monitoring units into different state categories, such as normal and abnormal states. By analyzing the data of each monitoring unit, an appropriate threshold is automatically determined based on the data distribution. Monitoring units with data values ​​greater than the threshold are marked as abnormal states, and those less than the threshold are marked as normal states, thus generating a state classification map of the watershed monitoring units.

[0035] Calculating the morphological parameters and spatial adjacency relationships of anomalous regions in a classification map involves calculating the shape, size, area, and other morphological parameters of these regions, and analyzing their spatial relationships. For example, this includes calculating the area, perimeter, and roundness of anomalous regions, as well as determining whether and to what extent they are adjacent.

[0036] Using the overall structural characteristics of the topological network as the spatial coverage characteristics of an event involves analyzing the spatial topological network of anomaly regions to extract the overall structural characteristics of the network, such as the degree distribution of nodes, clustering coefficients, and shortest path lengths. These characteristics can reflect the spatial coverage and degree of influence of the event.

[0037] Step S140: Analyze the water conservancy dispatch operation records corresponding to historical extreme climate events, extract the engineering state change data before and after the operation, identify the direct correlation between the operation and the state change through causal inference methods, and construct an operation impact map including the operation object, operation intensity, impact time and chain reaction path, wherein the chain reaction path is represented by the transmission delay and amplitude attenuation rate of the state change.

[0038] Analyzing historical records of water conservancy scheduling operations corresponding to extreme weather events involves reading and understanding the information in these records to extract key details such as the time of execution, the object of the operation, and the content of the operation. Causal inference methods are statistical methods used to determine causal relationships between variables. In this step, causal inference methods are used to identify the direct correlation between water conservancy scheduling operations and changes in the state of the project. For example, the Pearl causal graph method is used to construct a causal model and analyze the causal relationship between operational variables (such as adjustments to sluice gate openings) and state variables (such as changes in reservoir water levels). By analyzing multiple historical data points, it is determined under what circumstances the operation caused the state change.

[0039] Constructing an operational impact map that includes the operational object, operational intensity, impact duration, and chain reaction path integrates the above analysis results into a graphical representation. The operational object refers to the hydraulic engineering facilities subject to scheduling operations, such as reservoirs, sluices, and pumping stations. Operational intensity refers to the magnitude of the operation, such as the adjustment range of sluice gate opening or changes in pumping station operating power. Impact duration is the duration of the operation's impact on the engineering status. Chain reaction path is the path of the chain reaction impact of the operation on other related engineering facilities or statuses, represented by the propagation delay and amplitude decay rate of status changes. Propagation delay is the time delay before the operation affects a related facility or status, and amplitude decay rate is the proportion by which the amplitude of the status change gradually decreases during the chain reaction process.

[0040] Step S150: Dynamically match the features of each stage of the event time stage segmentation model with the timestamps of the operation influence map. At the same time, calculate the spatial correlation between each path and the abnormal sub-region in the operation influence map based on the node connectivity strength of the spatial topology network, and generate a two-dimensional correlation matrix containing time dependency weights and spatial correlation weights.

[0041] Dynamically matching the characteristics of each stage of the event time-phase segmentation model with the timestamps of the operational impact map involves mapping and matching the characteristics of the event at different time stages (such as the initiation, development, peak, and decline stages) with the execution time of water conservancy scheduling operations. This matching allows us to understand which scheduling operations were taken at different stages of the event and their impact on the event's development. Calculating the spatial correlation between each path and anomaly sub-region in the operational impact map based on the node connectivity strength of the spatial topology network involves analyzing the spatial relationships between chain reaction paths and anomaly sub-regions in the operational impact map based on the connectivity strength between nodes in the spatial topology network of the anomaly region. If a chain reaction path of a certain operation is spatially close to an anomaly sub-region and the connectivity strength of the nodes in the topology network is high, it indicates a high spatial correlation between the operation and that anomaly sub-region.

[0042] Generating a two-dimensional correlation matrix that includes time dependency weights and spatial correlation weights involves integrating the temporal matching results and spatial correlation calculation results into a single two-dimensional matrix. The time dependency weights represent the degree of correlation between the operation and the event's time phase, while the spatial correlation weights represent the degree of spatial correlation between the operation and the abnormal sub-region. This two-dimensional correlation matrix comprehensively reflects the temporal and spatial relationships between water conservancy scheduling operations and events.

[0043] Step S160: Construct a dynamic index structure for the historical event resource library based on a two-dimensional association matrix. The index key includes the event time stage identifier, spatial topology network node ID, and operation impact path identifier. Through the multi-dimensional combination of the index key, collaborative retrieval of event patterns, spatial features, and operation records is achieved. At the same time, it supports dynamic updating of the weight values ​​of the association matrix based on newly accessed multi-source sensing data.

[0044] The index key contains an event time phase identifier, a spatial topology network node ID, and an operational impact path identifier, serving as a unique identifier for each index entry. The event time phase identifier distinguishes different time phases of an event, such as the initiation, development, peak, and decline phases. The spatial topology network node ID identifies nodes within the spatial topology network of the anomalous area, allowing location of specific anomalous sub-regions. The operational impact path identifier identifies the chain reaction path of water conservancy scheduling operations, revealing the impact path of the operation on other engineering facilities or their status.

[0045] The collaborative retrieval of event patterns, spatial characteristics, and operation records through multi-dimensional combinations of index keys allows users to simultaneously search the historical event resource database based on multiple dimensions, including event time phase, spatial topology network nodes, and operation impact paths. For example, users can query water conservancy scheduling operations that occurred near a specific abnormal sub-region during the peak of an event and their chain reactions to other regions. This multi-dimensional retrieval method provides a more comprehensive and accurate way to obtain the required information. Furthermore, this dynamic index structure supports dynamic updates to the weight values ​​of the correlation matrix based on newly accessed multi-source sensing data. When new meteorological, hydrological, or water conservancy project operation sensing data is accessed, the temporal and spatial correlation between operations and events needs to be re-analyzed and calculated, thereby updating the weight values ​​of the two-dimensional correlation matrix. The updated correlation matrix reflects the latest event situation and the impact of scheduling operations, ensuring that the information in the historical event resource database remains timely and accurate.

[0046] Step S200: Obtain real-time water conservancy monitoring data, match the real-time water conservancy monitoring data with the event time change patterns and spatial coverage characteristics in the historical event resource database, calculate the matching degree between the real-time water conservancy monitoring data and each historical event by comparing the time and spatial dimensions, filter similar historical events that meet the preset matching degree, extract the scheduling operation records associated with similar historical events and organize them into a reference scheduling operation sequence in chronological order.

[0047] Acquiring real-time water conservancy monitoring data involves collecting various water-related monitoring data at the current moment. This data includes meteorological sensing data (such as real-time temperature, precipitation, and wind speed), hydrological sensing data (such as water level, flow rate, and water quality), and water conservancy project operation sensing data (such as reservoir storage capacity, sluice gate opening degree, and pumping station operating power). This data can be collected in real time by various monitoring devices distributed throughout the watershed and transmitted to the data processing center.

[0048] In one implementation, step S200 may specifically include the following steps S210 to S260:

[0049] Step S210: Perform multi-dimensional data processing on real-time water conservancy monitoring data, separate meteorological, hydrological and engineering operation data according to data type, perform adjacent time-to-time trend analysis on various types of data, identify and remove isolated data points that deviate from the overall trend, and fill in the missing time periods in the data sequence by trend extension based on the slope of the changes in the preceding and following data segments, thereby generating a continuous and complete real-time monitoring data sequence.

[0050] Multi-dimensional data processing of real-time water conservancy monitoring data involves comprehensively sorting and classifying the collected data for subsequent analysis and application. Separating meteorological, hydrological, and engineering operation data by data type involves categorizing real-time data according to their respective categories, resulting in meteorological data (such as temperature, precipitation, and wind speed), hydrological data (such as water level, flow rate, and water quality), and water conservancy engineering operation data (such as reservoir storage capacity, sluice gate opening degree, and pumping station operating power).

[0051] Analyzing the changing trends of various types of data over adjacent time periods involves studying how each type of data changes between close moments, observing whether the data is rising, falling, or stabilizing. By calculating the difference or rate of change between adjacent time points, the direction and magnitude of the data change can be determined. For example, for water level data, calculating the difference between two adjacent time points indicates that the water level is rising rapidly if the difference is positive and large.

[0052] Identifying and removing isolated data points that deviate from the overall trend involves finding those data points in a data sequence that are significantly inconsistent with the overall trend and removing them from the data sequence. These isolated data points may be due to monitoring equipment malfunctions, data transmission errors, or other reasons, and if not addressed, they will affect subsequent analysis results. Statistical methods, such as those based on standard deviation, can be used to identify isolated data points. Data points that exceed a certain standard deviation range are considered isolated and removed.

[0053] Filling in gaps in a data sequence by trend extension based on the slope of changes in preceding and following data segments involves analyzing the changing trends of data segments before and after the gap, and using the slope of these changes to predict and fill in the missing data. For example, if there are gaps in water level data for a certain period, the slope of changes in water level data before and after the gap can be calculated, and the water level value within the gap can be predicted based on this slope, thus achieving data filling.

[0054] Step S220: Extract the monitoring data of the most recent observation period from the continuous and complete real-time monitoring data sequence, calculate the unit time change sequence of the data in the most recent observation period, the distribution of extreme value occurrence locations, and the cumulative consistency of the continuous change direction, and construct a real-time data feature set containing temporal dynamic features and spatial distribution features, wherein the cumulative consistency of the continuous change direction is represented by the number of continuous data points with the same change sign.

[0055] In one implementation, step S220 may specifically include the following steps S221 to S226:

[0056] Step S221: Extract the monitoring data of the most recent observation period from the continuous and complete real-time monitoring data sequence, and divide the period into multiple time segments of equal length according to the data acquisition time interval. Each time segment contains multiple continuous monitoring data points.

[0057] Extracting monitoring data from the most recent observation period from a continuous and complete real-time monitoring data sequence, as mentioned earlier, involves extracting data from the processed real-time data sequence within a recent time period based on a preset observation period. Dividing this period into multiple equal-length time segments according to the data acquisition time interval is for further refinement and analysis of the extracted observation period.

[0058] Step S222: Perform linear fitting on the monitoring data points within each time segment, calculate the slope of the fitted line as the unit time change of that segment, and arrange the unit time changes of all time segments in sequence to obtain the unit time change sequence.

[0059] Linear fitting is performed on the monitoring data points within each time segment. This involves using linear regression methods such as least squares to find a straight line that best fits the data points within that time segment. The purpose of linear fitting is to approximate the trend of data change within that time segment using a straight line. The slope of the fitted line is calculated as the unit-time change for that segment because the slope reflects the average rate of change of the data within that time segment. A positive slope indicates an upward trend, a negative slope indicates a downward trend, and a larger absolute value of the slope indicates a faster rate of change. The unit-time changes for all time segments are arranged sequentially to form a sequence of unit-time changes. This sequence visually demonstrates the unit-time changes of the data throughout the entire observation period.

[0060] Step S223: Traverse the monitoring data throughout the entire observation period, identify the maximum and minimum points in the data sequence, record the specific time location and value of each extreme point within the observation period, count the number of extreme points distributed in each time segment, and generate the distribution characteristics of extreme value occurrence locations.

[0061] Traversing the monitoring data throughout the entire observation period involves accessing and analyzing all monitoring data within the most recent observation period. During this process, maxima and minima are identified in the data sequence. A maximum is the point with the highest data value within a local range, and a minimum is the point with the lowest data value within a local range. For example, for water level data, if the water level at a certain moment is higher than the water levels at its adjacent moments, then the point corresponding to that water level is a maximum; conversely, if the water level at a certain moment is lower than the water levels at its adjacent moments, then the point corresponding to that water level is a minimum. Recording the specific time location and value of each extreme point within the observation period ensures accurate recording of the time and corresponding value of each extreme point. Statistically analyzing the distribution of extreme points across different time segments involves classifying and statistically analyzing the identified extreme points according to time segments, calculating the number of extreme points contained within each time segment. Through this statistical analysis, the distribution of extreme points across different time segments can be obtained, generating a distribution characteristic of extreme value occurrence locations.

[0062] Step S224: Analyze the direction of change of continuous data points. When the value of the next point of an adjacent data point is greater than the value of the previous point, it is marked as a positive change; when it is less than the previous point, it is marked as a negative change; when it is equal to the previous point, it is marked as a zero change. Calculate the maximum length and total length of continuous positive or negative changes, and generate a cumulative consistency index of continuous change direction.

[0063] Analyzing the direction of change in continuous data points involves comparing and judging the numerical relationship between adjacent data points in the monitoring data sequence. When the value of a subsequent data point is greater than the value of a preceding data point, it indicates an upward trend, marked as a positive change; when the value of a subsequent data point is less than the value of a preceding data point, it indicates a downward trend, marked as a negative change; when the value of a subsequent data point is equal to the value of a preceding data point, it indicates no change, marked as zero change.

[0064] The maximum and total lengths of consecutive positive or negative changes are calculated by separately counting the lengths of the longest segments in the data sequence exhibiting consecutive positive or negative changes, and the total length of all segments with consecutive positive or negative changes. A cumulative consistency index for the direction of continuous change is generated by combining the obtained information such as the maximum and total lengths of consecutive change directions to form an index used to measure the consistency of the direction of continuous change in the data. This index reflects whether the trend of data change is stable and continuous over a period of time.

[0065] Step S225: Grid the spatial distribution data of the continuous and complete real-time monitoring data sequence, calculate the difference between the value of each grid cell and the mean value of the surrounding adjacent grid cells, divide the grid cells into different state levels according to the size of the difference, count the number of grid cells and spatial clustering of each state level, and generate spatial distribution characteristics.

[0066] Gridding the spatial distribution of continuous and complete real-time monitoring data sequences involves dividing the spatial area where water conservancy monitoring data is recorded into multiple equally sized grid cells. These grid cells can cover the entire monitoring area, and each grid cell contains the monitoring data within that area. For example, for watershed water level monitoring data, the geographical space of the watershed is divided into multiple square or rectangular grid cells, and the average water level value of the area is recorded within each grid cell. Calculating the difference between the value of each grid cell and the mean value of its surrounding grid cells involves calculating the difference between the data value of each grid cell and the average value of the data values ​​of its surrounding grid cells. This difference reflects the degree of deviation of the data value of that grid cell relative to the surrounding area.

[0067] Dividing grid cells into different state levels based on the magnitude of the difference involves setting different thresholds based on the calculated mean difference. For example, with three thresholds, a grid cell is classified as low-difference if the mean difference is less than the first threshold; medium-difference if the mean difference is between the first and second thresholds; and high-difference if the mean difference is greater than the second threshold.

[0068] Statistical analysis of the number and spatial clustering of grid cells at each state level involves counting the number of grid cells divided into different state levels and analyzing their spatial clustering. For example, it involves counting the number of grid cells at low-difference state levels and whether these cells are concentrated in a specific area; and where the grid cells at high-difference state levels are distributed. Generating spatial distribution features involves organizing and describing the statistically obtained information on the number and spatial clustering of grid cells at each state level to obtain features reflecting the spatial distribution of real-time monitoring data.

[0069] Step S226: Combine the sequence of changes per unit time, the distribution characteristics of extreme value locations, the cumulative consistency index of continuous change direction, and the spatial distribution characteristics in the order of time dynamic characteristics first and spatial distribution characteristics last to construct a real-time data feature set containing multi-dimensional sub-features.

[0070] Combining the unit-time change sequence, extreme value location distribution characteristics, cumulative consistency index of continuous change direction, and spatial distribution characteristics in the order of temporal dynamic characteristics first and spatial distribution characteristics second integrates the temporal dynamic characteristics (unit-time change sequence, extreme value location distribution characteristics, and cumulative consistency index of continuous change direction) and spatial distribution characteristics calculated in the previous steps in a specific order. Temporal dynamic characteristics reflect the changes in data over time, while spatial distribution characteristics reflect the distribution of data in space. Combining them in the order of temporal dynamic characteristics first and spatial distribution characteristics second can more clearly show the multi-dimensional characteristics of the data. Constructing a real-time data feature set containing multi-dimensional sub-features involves forming a complete dataset from the above combined features to describe the characteristics and changes of the current real-time water conservancy monitoring data. This feature set contains multiple sub-features, each describing the data from a different perspective. For example, the unit-time change sequence describes the rate of change of the data per unit time, the extreme value location distribution characteristics describe the temporal distribution of data extreme values, the cumulative consistency index of continuous change direction describes the continuity of the data change direction, and the spatial distribution characteristics describe the spatial distribution differences of the data.

[0071] Step S230: Traverse the historical events in the historical event resource library. For each historical event, extract the trend characteristics of each stage in the event time change pattern and the abnormal area morphological characteristics in the spatial coverage characteristics. Construct a historical event feature template containing time trend sequence and spatial morphological parameters. Each historical event feature template corresponds to a unique historical event identifier.

[0072] Traversing the historical events in the historical event repository involves accessing and processing each historical event stored in the repository one by one. The historical event repository records relevant information about different extreme weather events, including the temporal variation patterns of the events, spatial coverage characteristics, and corresponding water conservancy dispatch operation records.

[0073] For each historical event, extracting the trend characteristics of each stage in its time change pattern is to find the changing trend of the event in different stages (such as the initiation period, development period, peak period and decline period) from the time change pattern of the historical event.

[0074] Extracting morphological features of anomalous regions from the spatial coverage characteristics of a historical event involves obtaining morphological information about these regions, such as their area, perimeter, and circularity. These morphological parameters reflect the shape and size of the anomalous regions and serve as spatial morphological parameters. Constructing a historical event feature template that includes a time trend sequence and spatial morphological parameters involves integrating the extracted time trend sequence and spatial morphological parameters to form a template describing the characteristics of the historical event. Each historical event feature template corresponds to a unique historical event identifier, which allows for unique identification and association with the corresponding historical event.

[0075] Step S240: In the time dimension, compare the unit time change sequence of the real-time data feature set with the time trend sequence of the historical event feature template segment by segment, calculate the number of times the change direction is consistent within the corresponding time period and the degree of closeness of the absolute value of the change, and generate a time matching score, wherein the degree of closeness is represented by the reciprocal of the difference in change.

[0076] In the time dimension, a segmented comparison is performed between the unit-time change sequence of the real-time data feature set and the time trend sequence of the historical event feature template. This involves dividing both the unit-time change sequence of the real-time data feature set and the time trend sequence of the historical event feature template into segments with equal time intervals, and then comparing their changes segment by segment. For example, the unit-time change sequence of real-time water level data and the time trend sequence of historical flood event feature templates are both divided into hourly segments, and their changes are compared hourly. Calculating the number of times the change direction is consistent within a corresponding segment involves counting the number of times the real-time data and historical event data show the same change direction within the same time period during the segmented comparison process. If both real-time data and historical event data show an upward trend or a downward trend within a certain time period, their change direction is considered consistent. Calculating the closeness of the absolute values ​​of change involves comparing the differences in the absolute values ​​of change between real-time data and historical event data within the corresponding time period. The closeness is represented by the reciprocal of the difference in change; that is, the smaller the difference in change, the higher the closeness. Generating a time-matching score involves comprehensively considering the number of times the change direction is consistent and the closeness of the absolute values ​​of the changes. A score is then calculated using a specific method to measure the degree of matching between real-time data and historical events over time. For example, the number of times the change direction is consistent can be multiplied by a weighting coefficient, and then added to the average of the closeness of the absolute values ​​of the changes to obtain the time-matching score.

[0077] Step S250: In the spatial dimension, calculate the regional similarity between the current spatial distribution pattern of the continuous and complete real-time monitoring data sequence and the spatial morphological parameters of the historical event feature template, count the number and area ratio of sub-regions in the current monitoring area that match the morphological characteristics of historical abnormal regions, and generate a spatial matching score.

[0078] In the spatial dimension, calculating the regional similarity between the current spatial distribution of continuous and complete real-time monitoring data sequences and the spatial morphological parameters of historical event feature templates involves comparing and analyzing the current spatial distribution reflected by the real-time monitoring data with the spatial morphological parameters in the historical event feature templates to calculate the degree of similarity between them. For example, comparing the spatial distribution of real-time water level data (such as the regional distribution of water level height) with the spatial morphological parameters such as the area, perimeter, and circularity of abnormal areas in historical flood event feature templates. Counting the number and area proportion of sub-regions in the current monitoring area that match the morphological characteristics of historical abnormal areas involves identifying those sub-regions in the current monitoring area whose spatial morphological characteristics are similar to those of historical abnormal areas, counting the number of these sub-regions, and calculating their area as a percentage of the total monitoring area. Generating a spatial matching score involves comprehensively considering the number and area proportion of sub-regions that match the morphological characteristics of historical abnormal areas, and obtaining a score through a specific calculation method. This score is used to measure the degree of matching between real-time data and historical events in the spatial dimension. For example, the number of sub-regions can be multiplied by a weighting coefficient, and the area proportion can be multiplied by another weighting coefficient to obtain the spatial matching score.

[0079] Step S260: Combine the time matching score and spatial matching score in a proportional manner to generate a comprehensive event matching degree. Select historical events with a comprehensive matching degree that meet the set conditions as similar historical events. Extract the scheduling operation records associated with similar historical events. Perform time validity verification on the operation records. Retain the operation records executed within the corresponding stage of the event time change pattern. Organize the operation records according to the order of operation execution time to generate a reference scheduling operation sequence.

[0080] The overall event matching score is generated by combining time-matching and spatial-matching scores in a proportional manner. This process comprehensively considers the matching between real-time water conservancy monitoring data and historical events in both time and spatial dimensions. The time-matching score reflects the similarity between real-time data and historical events in terms of temporal trends, while the spatial-matching score reflects the similarity between real-time data and historical events in terms of spatial distribution patterns. By assigning appropriate weights to the time-matching and spatial-matching scores and then summing them in a weighted manner, the overall event matching score is obtained.

[0081] The weighting can be adjusted according to specific application scenarios and needs. If, in this water conservancy scheduling scenario, changes in the time dimension have a more critical impact on events, then a higher weight can be assigned to the time matching score; conversely, if the similarity of spatial distribution is more important, then the weight of the spatial matching score can be increased.

[0082] Historical events whose overall matching degree meets the set criteria are selected as similar historical events. The set criteria can be a threshold for the overall matching degree. When the overall matching degree of a historical event exceeds the threshold, the historical event is considered to have a high similarity to the current real-time water conservancy monitoring data and is selected as a similar historical event.

[0083] Extracting scheduling operation records associated with similar historical events involves retrieving relevant water conservancy scheduling operation information from a pool of similar historical events. These operation records contain various scheduling measures taken during the occurrence of historical events, such as reservoir water storage and flood discharge operations, and the opening and closing of sluice gates.

[0084] Performing time validity checks on operation logs involves verifying whether these logs correspond to the current real-time event in time. Specifically, it involves comparing the current stage of the real-time event with the corresponding stage in the historical event's timeline, retaining only operation logs executed within the corresponding stage. For example, if the current real-time event is in its startup phase, then only scheduling operation logs executed during the startup phase of similar historical events will be retained.

[0085] The process of organizing and generating a reference scheduling operation sequence by arranging operation records that have passed time validity verification in the order of operation execution forms an operation sequence that can be used as a reference for current water conservancy scheduling.

[0086] Step S300: Obtain the current water conservancy project operation data, which includes the current operation status parameters and operable constraint range of the water conservancy project facilities. Map and adapt the reference scheduling operation sequence to the current water conservancy project operation data. Adjust the parameter values ​​of each operation item in the reference scheduling operation sequence according to the current operation status parameters and operable constraint range to generate a candidate set of scheduling instructions that meet the current project operation conditions.

[0087] In one implementation, step S300 may specifically include the following steps S310 to S360:

[0088] Step S310: Parse the reference scheduling operation sequence, extract the execution time stamp, target identifier and original operation parameters of each operation item, group the operation items according to the target identifier, calculate the time interval distribution of the operation items in the same group, identify consecutive operation items with decreasing time intervals, mark them as operation sequences with potential associations, and generate a basic list of operation items containing potential association markers.

[0089] In one implementation, step S310 may specifically include the following steps S311 to S316:

[0090] Step S311: Traverse each operation item in the reference scheduling operation sequence, extract the execution time stamp and convert it into a time offset relative to the start time of the sequence, extract the target identifier and the original operation parameters, assign a unique identifier to each operation item, and generate an operation item information table containing the identifier, time offset, target identifier, and original operation parameters.

[0091] Traversing each operation item in the reference scheduling operation sequence involves processing each operation item individually. The execution time stamp is extracted and converted into a time offset relative to the start of the sequence. This is done to standardize the time scale for easier subsequent time analysis and comparison. The target object identifier and original operation parameters are extracted. As mentioned earlier, the target object identifier identifies the hydraulic engineering facility targeted by the operation, and the original operation parameters are the specific operation values ​​specified in the operation item.

[0092] Each operation item is assigned a unique identifier, which can be a numerical number or a code with specific rules, used to uniquely identify each operation item in subsequent operations. An operation item information table is generated, containing the identifier, time offset, target identifier, and original operation parameters. This table organizes the extracted and processed information for easy data storage and retrieval.

[0093] Step S312: Group the operation items in the operation item information table according to the target object identifier. The operation items in the same group contain the same target object identifier, generating multiple target object operation groups. The operation items in each group are arranged in ascending order according to the time offset.

[0094] Grouping operation items in the operation item information table by their target object identifier is based on the classification of the water conservancy engineering facilities they correspond to. Operation items with the same target object identifier are grouped together, allowing for centralized management and analysis of operations on the same facility. Multiple target object operation groups are generated, each corresponding to a specific water conservancy engineering facility. Operation items within each group are sorted in ascending order by time offset to ensure they are arranged chronologically according to their execution time.

[0095] Step S313: For each group of operations, calculate the time offset difference between adjacent operations to obtain a time interval sequence. Perform trend analysis on the time interval sequence to identify subsequences that show a continuous decreasing trend. These subsequences must show the time interval decreasing feature multiple times consecutively and be marked as potential related subsequences.

[0096] For each group of operations, the time offset difference between adjacent operations is calculated by subtracting the time offsets of adjacent operations to obtain the time interval. The time intervals of all adjacent operations are arranged in sequence to obtain the time interval sequence.

[0097] Trend analysis of time interval sequences can be performed using methods such as moving averages and linear regression. Moving averages calculate the average value within a certain window of the time interval sequence and observe the trend of this average value. Linear regression, on the other hand, describes the trend of the time interval sequence by fitting a straight line. Identifying subsequences that exhibit a continuously decreasing trend requires that the time intervals show this decreasing characteristic multiple times consecutively. For example, in a time interval sequence [2, 1.5, 1, 0.8, 0.6], subsequences such as [2, 1.5, 1] ​​and [1, 0.8, 0.6] can be identified as exhibiting a continuously decreasing trend.

[0098] Step S314: Analyze the changing trend of the original operation parameters of each operation item in the potentially related subsequence, calculate the ratio of the parameter changes of adjacent operation items. If the ratio shows an increasing or decreasing characteristic, the subsequence is confirmed to be an operation sequence with correlation; otherwise, the marking is canceled.

[0099] Analyzing the changing trends of the original operating parameters of each operation item in a potentially correlated subsequence involves observing how the original operating parameters of the operation items change within the potentially correlated subsequence. For example, for the operation item of reservoir sluice gate opening, we observe whether its opening percentage gradually increases or gradually decreases.

[0100] The ratio of parameter changes between adjacent operations is calculated by dividing the original parameter changes of the adjacent operations by the ratio. If the ratio shows an increasing or decreasing trend, the subsequence is confirmed to be an associated sequence of operations. For example, if the ratios of parameter changes between adjacent operations are 1.2, 1.5, and 1.8, showing an increasing trend, it indicates that there is a certain correlation between these operations, possibly indicating continuous adjustments made to gradually achieve a certain scheduling goal. Conversely, if the ratio does not show a clear increasing or decreasing trend, such as 1.2, 0.8, and 1.3, the potential association marker for the subsequence is removed.

[0101] Step S315: For the confirmed sequence of associated operations, mark the order of the operation items according to their positions in the sequence. Mark the first operation item as the starting operation and the subsequent operation items as associated follow-up operations. Record the correspondence between the identifiers of the starting operation and the associated follow-up operations.

[0102] For a confirmed sequence of associated operations, marking the operations in order of their position within the sequence clarifies the role and sequence of each operation. Marking the first operation in the sequence as the starting operation is the initial action that triggers a series of associated operations. Recording the correspondence between the identifiers of the starting operation and subsequent associated operations involves associating the unique identifier of the starting operation with the unique identifier of each subsequent associated operation.

[0103] Step S316: Add an association marker field to the operation item information table, mark the operation item in the associated operation sequence with the associated sequence number and the position number in the sequence, and generate a basic list of operation items containing potential association markers. The association marker field of non-associated operation items is empty.

[0104] Adding a correlation marker field to the operation item information table is to create a dedicated field in the table to record the correlation information of operation items. Operation items in a correlated operation sequence are labeled with their correlation sequence number and position number within the sequence. The correlation sequence number links operation items within the same correlated operation sequence, while the position number specifies the exact location of each operation item within the sequence.

[0105] Generating a basic list of operation items containing potential association markers involves compiling the operation item information table with association marker fields added into a list. In this list, operation items in a related operation sequence have explicit association markers, while the association marker field for non-related operation items is empty.

[0106] Step S320: Extract the current operating status parameters corresponding to the target identification from the current water conservancy project operation data, including real-time equipment operation indicators, structural fatigue cumulative values ​​and surrounding hydrological environment parameters. Calculate the synergistic change coefficient between indicators through cross-analysis of the status parameters. The synergistic change coefficient is represented by the consistency of the change direction of the two parameters and the ratio of the change magnitude, and generate the target status correlation matrix.

[0107] Extracting current operational status parameters corresponding to the target object identifier from current water conservancy project operation data involves finding the corresponding water conservancy facility's operational status information from the current water conservancy project operation data based on the target object identifier in the operation item information table. Real-time equipment operation indicators reflect the current working status of the equipment, such as the actual opening degree of reservoir sluice gates and the actual operating power of pumping stations. Structural fatigue cumulative values ​​reflect the degree of fatigue of the structure during long-term operation of water conservancy facilities and are closely related to the service life and safety of the facilities. Surrounding hydrological environmental parameters include information such as water level, flow rate, and water quality; these parameters affect the operational efficiency and safety of water conservancy facilities.

[0108] Calculating the synergistic variation coefficient between indicators through cross-analysis of state parameters allows for in-depth research into the relationships between different state parameters. The synergistic variation coefficient is represented by the ratio of the consistency of the direction of change of the two parameters to the magnitude of change.

[0109] Generating the state correlation matrix of the affected objects involves organizing the calculated cooperative change coefficients into a matrix form. The rows and columns of the matrix correspond to different state parameters, and the elements in the matrix are the cooperative change coefficients between the corresponding parameters.

[0110] Step S330: Obtain the operable constraint range in the current water conservancy project operation data, organize the upper and lower limit thresholds, single adjustment amount boundary and operation interval threshold of the operation parameters according to the target object identifier, and adjust the boundary value of the constraint range in combination with the cooperative change coefficient in the target object state correlation matrix. The higher the cooperative change coefficient of the state parameter, the stricter the corresponding constraint boundary value, and generate a dynamic constraint parameter table.

[0111] Obtaining the operational constraints from the current water conservancy project operation data involves extracting limitations on operational parameters from this data. Upper and lower thresholds define the range of values ​​for operational parameters, such as the minimum and maximum opening values ​​of reservoir sluice gates. Single adjustment limits restrict the magnitude of parameter adjustments during each operation, preventing excessive adjustments from adversely affecting the project. Operation interval thresholds specify the minimum time interval between two adjacent operations to ensure sufficient time for the engineering facilities to respond to operations.

[0112] Organizing the upper and lower limits, single adjustment limits, and operation interval thresholds of the operation parameters according to the object of action is to classify and organize the operation constraints of different objects (hydraulic engineering facilities), so as to facilitate the subsequent targeted adjustment of the operation parameters of each facility.

[0113] By combining the co-variance coefficients in the state correlation matrix of the affected object, the boundary values ​​of the constraint range are adjusted. A higher co-variance coefficient indicates a stronger correlation between these parameters, meaning a change in one parameter may significantly impact others. Therefore, for parameters with high co-variance coefficients, the adjustment range of the operational parameters needs to be more strictly limited to ensure the safety and stability of the project.

[0114] Generating a dynamic constraint parameter table involves organizing the adjusted operational parameter constraints into a tabular format. This table contains information such as the upper and lower limits of the operational parameters for each affected object, the boundary of a single adjustment, and the threshold for operational intervals. These constraints are dynamically adjusted based on the coordinated changes in state parameters to adapt to different engineering operating states.

[0115] Step S340: For each operation item, compare its original operation parameters with the upper and lower limit thresholds of the dynamic constraint parameter table, calculate the parameter deviation, and at the same time trace the historical execution effect data of the preceding operation items in the potential related operation sequence. Calculate the residual impact value of the preceding operation on the current operation through effect propagation analysis. The residual impact value is calculated by the change amplitude and decay rate of the state parameters after the preceding operation.

[0116] For each operation item, its original operation parameters are compared with the upper and lower thresholds in the dynamic constraint parameter table to check whether the original operation parameters are within the allowable range. If the original operation parameters exceed the upper or lower thresholds, it indicates that the parameters have deviated and need to be adjusted. The parameter deviation is calculated by using the difference between the original operation parameters and the upper and lower thresholds to represent the degree of deviation. Tracing the historical execution effect data of the preceding operation items in the potentially related operation sequence involves finding the impact data of the preceding operation items on the engineering state parameters after execution in the potentially related operation sequence. This historical execution effect data includes changes in state parameters, such as changes in reservoir storage and water level fluctuations.

[0117] Effect propagation analysis calculates the residual impact of a preceding operation on the current operation. This analysis studies how the influence of a preceding operation is transmitted to the current operation. The residual impact value is calculated using the magnitude and decay rate of changes in state parameters after the preceding operation.

[0118] Step S350: Based on the co-change coefficient of the state correlation matrix of the target object, determine the influence weight of each state parameter on the operating parameter. The higher the co-change coefficient of the state parameter and the operating parameter, the greater the influence weight. Combined with the parameter deviation and considering the system state offset trend indicated by the residual influence value, calculate the adjustment direction and initial adjustment magnitude of the operating parameter.

[0119] In one implementation, step S350 may specifically include the following steps S351 to S356:

[0120] Step S351: Extract the co-change coefficients of the operation parameters and each state parameter from the state association matrix of the object of action. A positive co-change coefficient indicates that the change direction is consistent, and a negative coefficient indicates that the change direction is opposite. The absolute value indicates the degree of correlation of the change magnitude.

[0121] Extracting the co-variance coefficients between operational parameters and state parameters from the state association matrix of the target object involves obtaining information on the co-variance of state parameters related to the current operational parameter from this data structure. The co-variance coefficients reflect the relationship between the operational and state parameters. A positive co-variance coefficient indicates that the operational and state parameters change in the same direction; for example, increasing the opening of a reservoir sluice gate will cause the downstream water level to rise. A negative co-variance coefficient indicates that the operational and state parameters change in opposite directions; for example, increasing the opening of a reservoir sluice gate may decrease the reservoir's water storage capacity. The larger the absolute value of the co-variance coefficient, the stronger the correlation between the changes in the operational and state parameters; that is, a change in one parameter will cause a significant change in the other.

[0122] Step S352: Calculate the relative distance between the current value and the critical value of each state parameter. The relative distance is represented by the ratio obtained by comparing the difference between the current value and the critical value with the critical value. The smaller the relative distance, the closer the state parameter is to the critical state.

[0123] The relative distance between the current value and the critical value of each state parameter is calculated. The critical value is a key boundary for the state parameter; when the state parameter approaches or exceeds the critical value, it may affect the safety and normal operation of the hydraulic engineering project. The relative distance is expressed as a ratio obtained by comparing the difference between the current value and the critical value with the critical value. For example, for the state parameter of reservoir water level, the critical value is the warning water level, and the current water level is the actual measured water level value. The relative distance is obtained by calculating (current water level - warning water level) / warning water level. The smaller the relative distance, the closer the state parameter is to the critical state, and in this case, more careful consideration is needed when adjusting the operating parameters.

[0124] Step S353: Multiply the absolute value of the cooperative change coefficient by the relative distance to obtain the comprehensive influence value of the state parameter on the operating parameter. The larger the comprehensive influence value, the more significant the influence of the state parameter on the operating parameter.

[0125] Multiplying the absolute value of the co-variance coefficient by the relative distance yields the comprehensive influence of the state parameter on the operating parameter. The absolute value of the co-variance coefficient reflects the correlation between the changes in the operating parameter and the state parameter, while the relative distance reflects how close the state parameter is to the critical state. Multiplying the two allows for a comprehensive consideration of the influence of both factors on the operating parameter. The larger the comprehensive influence value, the more significant the influence of the state parameter on the operating parameter, and the more important it is to consider the changes in the state parameter when adjusting the operating parameter.

[0126] Step S354: Normalize the comprehensive influence value of all state parameters to obtain the influence weight of each state parameter. The sum of the influence weights is 1. The larger the comprehensive influence value, the greater the influence weight of the state parameter.

[0127] Normalizing the combined influence value of all state parameters involves processing the combined influence values ​​of all state parameters so that their sum equals 1. The purpose of normalization is to facilitate comparison of the degree of influence of each state parameter on the operating parameter and to use this as an influence weight for subsequent calculations. A min-max normalization method can be used, which involves subtracting the minimum value among all combined influence values ​​from the combined influence value of each state parameter, and then dividing by the difference between the maximum and minimum values ​​to obtain the normalized influence weight. A state parameter with a larger combined influence value has a larger normalized influence weight, indicating that the influence of this state parameter on the operating parameter accounts for a greater proportion of all state parameters.

[0128] Step S355: Analyze the positive and negative directions of the parameter deviation. If the original operating parameter is higher than the upper limit threshold, the adjustment direction is to decrease; if it is lower than the lower limit threshold, the adjustment direction is to increase. Combine the influence weights to calculate the adjustment component contributed by each state parameter. The adjustment component is obtained by multiplying the parameter deviation by the influence weight.

[0129] Analyze the positive and negative directions of the parameter deviation. The parameter deviation is the difference between the original operating parameter of the operation item and the upper and lower limit thresholds of the dynamic constraint parameter table. If the original operating parameter is higher than the upper limit threshold, it means that the operating parameter is too large and needs to be reduced, and the adjustment direction is to decrease; if the original operating parameter is lower than the lower limit threshold, it means that the operating parameter is too small and needs to be increased, and the adjustment direction is to increase.

[0130] The adjustment component of each state parameter is calculated by combining its influence weights. This adjustment component reflects the magnitude of each state parameter's contribution to the adjustment of the operating parameters. The adjustment component of each state parameter is obtained by multiplying the parameter deviation by its influence weight.

[0131] Step S356: Add up the adjustment components of all state parameters to obtain the initial adjustment range based on the parameter deviation; analyze the sign of the residual influence value. If the residual influence value is positive, it means that the historical operation has caused the system state to shift in the same direction as the adjustment direction, so decrease the initial adjustment range. If it is negative, it means that it has shifted in the opposite direction, so increase the initial adjustment range. Generate the final operation parameter adjustment direction and initial adjustment range.

[0132] The adjustment components of all state parameters are summed to obtain the initial adjustment magnitude based on the parameter deviation. Each state parameter's adjustment component reflects its contribution to the adjustment of the operating parameters. Summing all adjustment components yields the initial adjustment magnitude after comprehensively considering the influence of each state parameter. The sign of the residual influence value is analyzed; this value reflects the residual effect of the preceding operation on the current operation. If the residual influence value is positive, it indicates that the preceding operation has shifted the system state in the same direction as the adjustment, and the initial adjustment magnitude can be appropriately reduced to avoid over-adjustment. If the residual influence value is negative, it indicates that the preceding operation has shifted the system state in the opposite direction to the adjustment, and the initial adjustment magnitude needs to be increased to compensate for this reverse shift. By considering the residual influence value, the final adjustment direction and initial adjustment magnitude of the operating parameters are generated.

[0133] Step S360: Based on the single adjustment amount boundary and operation interval threshold of the dynamic constraint parameter table, correct the initial adjustment range so that the adjusted operation parameters do not exceed the upper and lower limit thresholds and the adjustment amount is within the single adjustment amount boundary. Sort all corrected operation items according to the execution time mark to generate a scheduling instruction candidate set.

[0134] The initial adjustment range is corrected based on the single adjustment limit and operation interval threshold in the dynamic constraint parameter table. The single adjustment limit specifies the maximum adjustment range of the parameter in each operation, and the operation interval threshold specifies the minimum time interval between two adjacent operations. If the initial adjustment range exceeds the single adjustment limit, the adjustment range needs to be reduced to within the limit range; at the same time, it must be ensured that the adjusted operation parameters are within the upper and lower limit thresholds of the dynamic constraint parameter table. Ensuring that the adjusted operation parameters do not exceed the upper and lower limit thresholds and that the adjustment range is within the single adjustment limit is to guarantee the safe and stable operation of water conservancy facilities. If the operation parameters exceed the upper and lower limit thresholds or the adjustment range is too large, it may cause damage to the project or lead to safety accidents. All corrected operation items are sorted by execution time marker, which arranges the adjusted and corrected operation items in chronological order of execution time. This ensures that dispatch instructions are executed in a reasonable time sequence, improving the efficiency and accuracy of dispatching.

[0135] Step S400: Input the candidate set of scheduling instructions into the digital twin watershed model, perform simulation operation calculations, simulate the entire process of water conservancy project operation under the guidance of the candidate set of scheduling instructions, record the water conservancy project operation status change data and watershed hydrological dynamic change data output by the digital twin watershed model in real time, and obtain simulation operation results including parameter change trends.

[0136] In one implementation, step S400 may specifically include the following steps S410 to S460:

[0137] Step S410: Parse the candidate set of scheduling instructions, extract the time offset, target identifier and correction operation parameters of each operation item, analyze the correspondence between the time offset and parameter adjustment for operation items with the same target identifier, identify the continuous pattern of parameter adjustment over time by the ratio of parameter adjustment of adjacent operations, and generate a model input instruction stream containing pattern features.

[0138] In one implementation, step S410 may specifically include the following steps S411 to S416:

[0139] Step S411: Traverse each operation item in the scheduling instruction candidate set, extract the time offset, the target identifier, and the correction operation parameters, assign a unique instruction identifier to each operation item, and record it in the instruction basic information table.

[0140] Traversing each operation item in the candidate set of scheduling instructions involves processing each operation item one by one. The time offset, target identifier, and adjusted operation parameters are extracted. As mentioned earlier, the time offset determines the execution time of the operation, the target identifier clarifies the target facility, and the adjusted operation parameters are the specific operation values ​​after adjustment. A unique instruction identifier is assigned to each operation item. This identifier can be a numerical number or a code with specific rules, used to uniquely identify each operation item in subsequent operations. This information is recorded in the instruction basic information table, which organizes the extracted time offset, target identifier, adjusted operation parameters, and assigned instruction identifier into a table for convenient data storage and retrieval.

[0141] Step S412: Group the operation items in the instruction basic information table according to the target identifier, so that each group of operation items has the same target identifier, and the operation items within the group are arranged in natural order of time offset to form a target operation sequence.

[0142] Grouping operation items in the instruction basic information table by target object identifier is based on the classification of the water conservancy engineering facilities targeted by the operation items. Grouping operation items with the same target object identifier allows for centralized management and analysis of operations on the same facility. Arranging operation items within a group in natural order of time offset ensures that the operation items are arranged according to the chronological order of execution time. This arrangement facilitates subsequent analysis of the correspondence between the time offset and parameter adjustment amounts of operation items, as well as the identification of continuous parameter adjustment patterns. By arranging operation items in chronological order, the temporal sequence and changing trends of operations on the same facility can be clearly understood, forming a target object operation sequence.

[0143] Step S413: For each operation sequence of the target object, calculate the ratio of parameter adjustment of adjacent operation items, and identify the continuous pattern of parameter adjustment over time by the change pattern of the ratio. The continuous pattern is manifested as the ratio fluctuating within a set range.

[0144] For each sequence of operations, the ratio of parameter adjustments between adjacent operations is calculated by dividing the changes in the correction parameters of adjacent operations. The changing pattern of this ratio identifies a continuous pattern in parameter adjustments over time; a continuous pattern is indicated by a regular change in the ratio of parameter adjustments. If the ratio fluctuates within a set range, it suggests a continuous pattern in parameter adjustments.

[0145] Step S414: Divide operation items with the same continuous pattern into an operation stage, record the starting instruction identifier and the number of operation items contained in each operation stage, and generate the stage division result.

[0146] Grouping operation items with the same continuous pattern into one operation stage involves grouping operation items based on the identified parameter adjustment continuous pattern. If the parameter adjustment ratios of a group of operation items fluctuate within the same set range, it indicates that they have the same continuous pattern, and these operation items can be grouped into one operation stage. The starting instruction identifier and the number of operation items included in each operation stage are recorded. The starting instruction identifier uniquely identifies the starting operation item of the operation stage, while the number of included operation items reflects the operation scale of that stage.

[0147] Step S415: Within each operation phase, calculate the mean and fluctuation range of the parameter adjustment ratio, and add the mean and fluctuation range as the pattern characteristics of that phase to the phase division results.

[0148] Within each operational phase, the mean and fluctuation range of the parameter adjustment ratios are calculated. The mean is the average of all parameter adjustment ratios within that phase, reflecting the average trend of parameter adjustment changes during that phase. The fluctuation range is the difference between the maximum and minimum values ​​of the parameter adjustment ratios, reflecting the magnitude of parameter adjustment changes within that phase. Adding the mean and fluctuation range as pattern features to the phase segmentation results allows for a more detailed description of the parameter adjustment patterns in each operational phase. By incorporating these pattern features into the phase segmentation results, richer information can be provided for subsequent model input and analysis.

[0149] Step S416: Combine the operation sequence of the target object with the stage division result to generate a model input instruction stream containing instruction identifier, time offset, correction operation parameters, target object identifier and the mode characteristics of the stage to which it belongs.

[0150] Combining the sequence of operations applied to the target object with the stage division results involves integrating the sequence of operation items grouped by the target object identifier and arranged in chronological order with the division results and pattern characteristics of each operation stage. This combination allows association between operation items and their respective operation stages, as well as the pattern characteristics of those stages.

[0151] Generating a model input command stream that includes command identifiers, time offsets, correction operation parameters, target identifiers, and mode characteristics of the corresponding stage involves presenting the integrated information in the form of a command stream. This command stream contains detailed information for each operation item, including command identifiers, time offsets, correction operation parameters, target identifiers, and mode characteristics of the corresponding operation stage.

[0152] Step S420: Map the current operating status parameters in the current water conservancy project operation data to the digital twin watershed model, and load the watershed topographic data and initial hydrological observation data. Transmit the associated parameters through the data interface to enable the project operation data and hydrological environment data to form a dynamic response relationship.

[0153] In this embodiment of the invention, the digital twin watershed model is a virtual simulation system constructed based on watershed geospatial data and hydrological and hydrodynamic principles. During its construction, the watershed can be divided into multiple computational grids or sub-watershed units according to the digital elevation model and river network data. One-dimensional cross-sectional division is used for the river network, while two-dimensional structured or unstructured grids are used for spatial discretization of the floodplain. Each computational unit is assigned initial hydraulic parameters such as topographic elevation, Manning roughness coefficient, soil saturated hydraulic conductivity, and specific yield.

[0154] The model describes the unsteady flow motion in the river channel by solving the one-dimensional Saint-Venant equations, whose continuity equation is: The momentum equation is Where A is the cross-sectional area of ​​the water passage, Q is the flow rate, h is the water depth, q is the lateral inflow, and S is the cross-sectional area of ​​the water passage. f The model uses the Manning formula to calculate the friction slope. For two-dimensional surface runoff in floodplains, a two-dimensional shallow water equation system is used for simulation. The runoff generation process on the slope is described using a full-storage runoff or over-permeability runoff model combined with the kinematic wave approximation. The model integrates hydraulic engineering facility modules. For reservoirs, it establishes water level-storage capacity curves and flow formulas for discharge facilities (including general weir and orifice flow formulas). For sluice gates, it calculates the flow rate in real time using free outflow or submerged outflow formulas based on gate opening, orifice size, and upstream and downstream water levels. For pumping stations, it calculates the pumping volume based on pump characteristic curves and operating power. All scheduling operations of engineering facilities are input into the model as time-varying boundary conditions, and upstream and downstream hydraulic elements are updated in real time. The model parameters are calibrated using historical hydrological observation data (such as measured flow rate and water level hydrographs) through a combination of automated calibration algorithms (e.g., general SCE-UA, genetic algorithms) and manual trial and error, ensuring the simulation results reach the predetermined accuracy with measured data. The model is also validated using independent time-period data to guarantee its reliability. The model has a standard data interface that can access meteorological observation data (precipitation, evaporation), hydrological monitoring data (water level, flow rate), and water conservancy project operation data (gate opening, pumping station power, etc.) in real time, and can interact with external systems through dynamic link libraries or APIs.

[0155] Mapping current operational status parameters from existing water conservancy project data to a digital twin watershed model involves loading the current operational status information of various facilities within the actual water conservancy project into the digital twin watershed model. These current operational status parameters include reservoir storage capacity, sluice gate opening degree, and pumping station operating power. Through precise mapping operations, the digital twin watershed model can accurately reflect the actual operational status of the current water conservancy project.

[0156] Simultaneously, watershed topographic data and initial hydrological observation data are loaded. Watershed topographic data describes the geographical features of the watershed, such as the distribution of mountains, rivers, and lakes, and plays a crucial role in simulating hydrological processes. Initial hydrological observation data consists of actual observations of hydrological elements such as water level, flow rate, and water quality recorded at the start of the simulation. This data provides the foundation for the initialization of the digital twin watershed model. Related parameters are transmitted through a data interface, which serves as a bridge for data transmission between different systems or modules. Through the data interface, related parameters such as current operating status parameters, watershed topographic data, and initial hydrological observation data are transmitted to the digital twin watershed model.

[0157] This allows for a dynamic response relationship between engineering operation data and hydrological environmental data. In a digital twin watershed model, the operation of water conservancy facilities impacts the watershed's hydrological environment, and conversely, changes in the hydrological environment affect the operation of water conservancy projects. By rationally setting the model's algorithms and rules, the engineering operation data and hydrological environmental data can influence and respond to each other, forming a dynamic response relationship.

[0158] Step S430: Start the operation instructions sequentially according to the time offset of the model input instruction stream. Before triggering each instruction, obtain the real-time running data of the target object. If the real-time running data meets the environmental conditions for operation execution, execute the operation and record the execution time. If it does not meet the conditions, maintain the current state and wait for the next time interval to re-evaluate.

[0159] The operation instructions are initiated sequentially according to the time offset of the model input instruction stream. This means that the corresponding operation instructions are triggered in chronological order based on the time offset of each operation item in the model input instruction stream. The time offset determines the execution time of each operation instruction, ensuring that the operation instructions are executed in the digital twin watershed model in a predetermined time order.

[0160] Before triggering each instruction, real-time operational data of the target object is acquired. The target object is the hydraulic engineering facility to which the instruction is directed. Through the monitoring function of the digital twin watershed model, real-time operational status parameters of the target object are acquired, such as the real-time water storage of the reservoir and the actual opening degree of the sluice gate. If the real-time operational data meets the environmental conditions for operation execution, the operation is executed and the execution time is recorded. The environmental conditions for operation execution are a series of conditions set according to the actual situation and safety requirements of the hydraulic engineering project. If the real-time operational data meets these conditions, the operation instruction is executed, and the execution time of the operation instruction is recorded for subsequent analysis and evaluation.

[0161] If the conditions are not met, the current state is maintained, and the system waits for the next time interval to reassess. When the real-time running data does not meet the environmental conditions for operation execution, the operation instruction is not executed; instead, the current running state of the target object is maintained. The system waits for the next time interval, retrieves the real-time running data of the target object again, and reassesses whether the environmental conditions for operation execution are met. This cyclical assessment method ensures that operation instructions are executed under appropriate environmental conditions, improving the accuracy and security of the digital twin watershed model simulation.

[0162] Step S440: During the simulation operation, collect data on the changes in the operating status of the water conservancy project at set time intervals, including the status parameters of the object after operation, response delay data, and status linkage data of related facilities. Simultaneously collect dynamic changes in the watershed hydrology, including hydrological observation values ​​in different regions and hydrological transfer data between regions, and generate a multi-source simulation data set.

[0163] During the simulation, data on the operational status changes of water conservancy projects are collected at set time intervals. These time intervals can be adjusted according to the simulation's accuracy requirements and actual conditions. The collected data includes the status parameters of the affected objects after operation, such as changes in reservoir storage volume after operation, and adjustments to sluice gate opening after operation; response delay data reflects the response time delay of the affected objects to operational commands, such as the time required for a sluice gate to fully open after receiving an opening command; and status linkage data of related facilities reflects the status changes of other water conservancy facilities associated with the affected objects, such as the status linkage changes of downstream sluice gates during reservoir discharge. Simultaneously, watershed hydrological dynamic change data is collected, including hydrological observations from different regions, such as water level, flow rate, and water quality data at different locations; and inter-regional hydrological transfer data describes the transfer process and changes of hydrological elements between different regions, such as the transfer speed and volume changes of water flow from upstream to downstream in rivers. A multi-source simulation dataset is generated by integrating the collected water conservancy project operational status change data and watershed hydrological dynamic change data into a simulation dataset containing multiple data sources.

[0164] Step S450: By establishing the spatial correspondence between the geographical location of the target object and the spatial relationship between the hydrological observation area, the spatial correspondence between engineering data and hydrological data is established. At the same time, the collection times of different types of data are aligned according to the time interval to generate spatiotemporal correlation data pairs.

[0165] In one implementation, step S450 may specifically include the following steps S451 to S456:

[0166] Step S451: Extract the geographical location information of each object in the data on changes in the operational status of water conservancy projects, and at the same time extract the location information of each hydrological observation area in the data on dynamic changes in the watershed hydrology. Determine the scope of influence based on the type and parameters of the object, and adjust the scope of influence as the parameter values ​​of the object change.

[0167] Geographical location information of each object involved in the operation status change data of water conservancy projects is extracted. The geographic location information of the objects involved can be obtained through technologies such as Geographic Information System (GIS), which records the specific location coordinates of each water conservancy facility (such as reservoir, sluice gate, pumping station, etc.) in the actual geographic space.

[0168] At the same time, the location information of each hydrological observation area in the dynamic change data of the watershed is extracted. The location information of the hydrological observation area can also be obtained through GIS data, which clarifies the specific location of each hydrological observation point (such as water level station, flow station, etc.) in the watershed.

[0169] The scope of influence is determined by the type and parameters of the object being acted upon. Different types of objects have different impact ranges on the hydrological environment. For example, the impact range of a reservoir is mainly a certain area downstream, while the impact range of a pumping station may be mainly concentrated in a smaller surrounding area. Furthermore, the scope of influence of the object can also adjust with changes in its parameter values.

[0170] Step S452: Overlay the influence range of the target object with the location information of the hydrological observation area, calculate the hydrological observation area included in the influence range of each target object, and preliminarily associate the engineering data of the target object with the hydrological data of the corresponding observation area to form a preliminary associated data pair.

[0171] Overlaying the influence range of an object with the location information of hydrological observation areas allows for the visualization and analysis of these areas in geographic space. Using GIS technology, the influence range of the object is graphically plotted on a map and overlaid with the location information of the hydrological observation areas. The number and specific locations of the hydrological observation areas within the influence range of each object are calculated and tallied on the overlaid map. The engineering data of the object is initially correlated with the hydrological data of the corresponding observation areas, forming preliminary data pairs. Based on the calculated hydrological observation areas within the influence range of the object, the engineering operation data of the object (such as reservoir storage capacity and sluice gate opening) are paired with the hydrological data (such as water level and flow rate) of the corresponding observation areas. Through this preliminary correlation, a basic connection is established between the engineering data and the hydrological data, forming preliminary data pairs.

[0172] Step S453: Perform time characteristic analysis on the engineering data and hydrological data in the preliminary correlation data pair, identify the delay time of hydrological data changes after engineering operations, adjust the time correspondence of the data according to the delay time, and generate time-calibrated correlation data pairs.

[0173] Analyzing the temporal characteristics of engineering and hydrological data in preliminary correlation data pairs is crucial for studying the patterns and relationships between these data over time. By analyzing the time series of the data, we can observe the response of hydrological data (such as downstream water level changes) after engineering events (e.g., reservoir discharge operations).

[0174] Identifying the time lag in hydrological data changes after engineering operations is crucial. Due to the inertia and transmission process of hydrological systems, the impact of engineering operations on the hydrological environment is often not immediately apparent but rather exhibits a certain time lag. By analyzing time series data, the point at which hydrological data begins to change significantly after the engineering operation is determined, and the difference between this point and the engineering operation time is calculated; this difference represents the time lag.

[0175] The time correspondence of the data is adjusted based on the delay time, so that the time points of the hydrological data correspond to the time points of the engineering data. For example, if the engineering operation time is t and the delay time is Δt, then the time point of the hydrological data is adjusted from t' to t'-Δt, thus making the engineering data and hydrological data more matched in time.

[0176] Generate time-calibrated associated data pairs by re-pairing the time-adjusted engineering and hydrological data to form time-calibrated associated data pairs.

[0177] Step S454: Calculate the ratio of the change range of engineering data and hydrological data in the time-calibrated associated data pair. The ratio of the change range is expressed by the ratio of the change in engineering data to the change in hydrological data. Select associated data pairs with the ratio within a reasonable range.

[0178] The calculation of the ratio of changes in engineering data and hydrological data in time-calibrated correlated data pairs involves calculating the changes in both engineering and hydrological data for each pair. These changes can be represented by the data differences at different time points within the same data series. The ratio of changes is expressed as the ratio of the changes in engineering data to the changes in hydrological data. This ratio reflects the degree of impact of engineering operations on hydrological environmental changes. For example, if a reservoir's water storage decreases by a certain amount, causing a rise in downstream water levels, the ratio of these two changes can indicate the intensity of the impact of reservoir discharge on downstream water levels.

[0179] The selection criteria include data pairs with correlation ratios within a reasonable range, which is pre-defined based on the characteristics of the actual water conservancy projects and hydrological environment. If the ratio of changes is too large or too small, it may indicate an anomaly or unreasonable data correlation.

[0180] Step S455: Perform a stability test on the filtered associated data pairs. Determine the stability of the association by the consistency of the change direction over multiple consecutive time intervals, and retain the associated data pairs that meet the stability requirements.

[0181] A stability test is performed on the filtered correlated data pairs to ensure that the correlation between engineering data and hydrological data remains stable and reliable over a period of time. By observing the changes in engineering and hydrological data over multiple consecutive time intervals, it is determined whether the correlation between them remains consistent.

[0182] The stability of the correlation can be judged by the consistency of the direction of change over multiple consecutive time intervals. If the direction of change of engineering data and hydrological data remains consistent over multiple consecutive time intervals, it indicates that the correlation between them has high stability.

[0183] Retain correlation data pairs that meet the stability requirements. For correlation data pairs that do not meet the stability requirements, there may be data errors or other interference factors affecting the correlation between engineering data and hydrological data. Remove these unstable correlation data pairs, and only retain correlation data pairs that meet the stability requirements.

[0184] Step S456: Classify stable correlated data pairs according to the target of action and the hydrological observation area, record the target of action, the observation area, the time interval and the corresponding data value for each data pair, and generate a set of spatiotemporal correlated data pairs.

[0185] Classifying stable correlated data pairs by their target objects and hydrological observation areas allows for more organized and managed data pairs that have undergone screening and stability testing. Categorizing data pairs by target objects (such as different reservoirs, sluices, and other water conservancy projects) and hydrological observation areas (such as different water level monitoring stations and flow monitoring points) allows related data pairs to be grouped together, facilitating subsequent querying and analysis.

[0186] Each data pair is recorded, including the object identifier, observation area identifier, time interval, and corresponding data value. The object identifier uniquely identifies each water conservancy project facility, the observation area identifier uniquely identifies each hydrological observation area, the time interval specifies the time range for data collection, and the corresponding data value is the specific engineering and hydrological data. By recording this detailed information, each spatiotemporally correlated data pair can be accurately located and described. Generating a spatiotemporally correlated data pair set involves integrating all the categorized and recorded stable correlated data pairs into a complete dataset. This set contains rich spatiotemporal information, reflecting the close temporal and spatial relationship between water conservancy project operation and hydrological environmental changes.

[0187] Step S460: Analyze the direction of change and the amount of change per unit time of each parameter in the spatiotemporal correlation data pair. The direction of change is determined by comparing the parameter values ​​of adjacent time intervals, and the amount of change per unit time is calculated by the ratio of the parameter difference to the time interval. Based on the direction of change and the amount of change per unit time, generate the trend curves of each parameter. Combine all the trend curves and the corresponding spatiotemporal correlations to obtain the simulation results that include the parameter change trends.

[0188] Analyzing the changing direction of parameters in spatiotemporal correlation data pairs involves detailed observation and comparison of engineering and hydrological data within each pair. By comparing parameter values ​​across adjacent time intervals, it's determined whether the parameter is increasing, decreasing, or remaining stable. For example, regarding reservoir storage parameters, if the storage volume at a later time point is greater than that at a previous time point, the change is considered increasing; conversely, it's decreasing; and if they are equal, the storage is considered stable.

[0189] The change per unit time is calculated as the ratio of the parameter difference to the time interval, reflecting the rate of change of the parameter per unit time. For each parameter, the parameter difference between adjacent time intervals is calculated, and then divided by the length of the time interval to obtain the change per unit time. This change directly reflects how fast the parameter changes.

[0190] Trend curves for each parameter are generated based on the direction of change and the amount of change per unit time. With time on the horizontal axis and parameter values ​​on the vertical axis, the curves showing the change of each parameter over time are plotted according to the direction of change and the amount of change per unit time. These trend curves clearly demonstrate the changes in parameters throughout the simulation process, helping analysts intuitively understand the dynamic trends of water conservancy project operation and hydrological environment changes. All trend curves and their corresponding spatiotemporal relationships are combined, integrating the generated trend curves for each parameter while preserving their spatiotemporal correlations. These spatiotemporal correlations reflect the interrelationships and influences of different parameters in time and space, such as the correlation between the trend of reservoir water storage and the trend of downstream water level changes.

[0191] Step S500: Analyze the simulation results, identify abnormal change nodes where the parameter change rate exceeds the preset range, locate the scheduling operation item that caused the abnormal change node and correct its parameter value, input the corrected scheduling instruction candidate set back into the digital twin watershed model for simulation calculation, repeat the above correction and simulation process until there are no abnormal change nodes in the simulation results, and output the water conservancy scheduling scheme.

[0192] In one implementation, step S500 may specifically include the following steps S510-S560:

[0193] Step S510: Analyze the parameter change trend curves in the simulation results, extract the rate of change of each parameter in continuous time intervals, compare the rate of change with the preset normal change interval point by point, identify sampling points whose rate of change exceeds the upper limit of the normal interval or is lower than the lower limit of the normal interval, mark them as abnormal change nodes, and record the parameter type, time coordinate and the difference between the rate of change and the interval boundary of the abnormal change nodes.

[0194] The rate of change of each parameter over consecutive time intervals is extracted, reflecting the degree of change of the parameter per unit time. For the trend curve of each parameter, the parameter difference between adjacent time intervals is calculated and then divided by the length of the time interval to obtain the rate of change of the parameter within that time interval. By continuously calculating the rate of change over multiple time intervals, a sequence of the rate of change of the parameter throughout the entire simulation process is obtained. The rate of change is compared point by point with a preset normal change range, which is a reasonable range of parameter change rates determined based on factors such as the design requirements, safety standards, and actual operating experience of the water conservancy project. Each value in the calculated parameter rate of change sequence is compared with the normal change range one by one to determine whether it falls within the range. Sampling points where the rate of change exceeds the upper limit or falls below the lower limit of the normal range are identified. When the rate of change is greater than the upper limit or less than the lower limit of the normal range, it indicates that the parameter change corresponding to that sampling point is abnormal. These sampling points that exceed the range are marked as abnormal change nodes. Record the parameter type, time coordinate, and difference between the rate of change and the interval boundary for the nodes with abnormal changes. Identify the parameter type corresponding to the nodes with abnormal changes, such as water level, flow rate, and reservoir storage. Record the specific location of the nodes with abnormal changes on the time axis, i.e., the time coordinate. Calculate the difference between the rate of change and the normal interval boundary. The magnitude of the difference reflects the severity of the abnormal change.

[0195] Step S520: For each abnormal change node, trace the scheduling operation items executed within the preset time period before that time coordinate in the simulation operation, extract the execution time, target identifier and parameter adjustment amount of the operation item, and construct an abnormal associated operation pool by combining the relationship between parameter type and target. The associated operation pool contains all operation items that may affect the abnormal parameters.

[0196] For each abnormal change node, the scheduling operations executed within a preset time period prior to that time coordinate in the simulation are traced. The preset time period is a time range pre-defined based on the response time of the water conservancy project and the actual situation. Within this time range, all executed scheduling operations are searched, as these operations may be the cause of the abnormal change. The execution time, the target object identifier, and the parameter adjustment amount of each operation are extracted. The execution time clarifies the execution point of the operation, the target object identifier identifies the water conservancy facility targeted by the operation, and the parameter adjustment amount is the specific value adjusted by the operation on the target object.

[0197] By combining the correlation between parameter types and their affected objects, specific correlations exist between different parameter types and different affected objects. For example, the reservoir storage capacity parameter is associated with the reservoir itself, while the downstream water level parameter is associated with the reservoir's flood discharge operations and downstream sluice gates. By analyzing these correlations, it is determined which operational items may affect anomalous parameters. An anomalous correlation operation pool is constructed, containing all operational items that may affect anomalous parameters. The scheduling operational items traced back and executed within a preset time period are filtered and integrated according to the correlation between parameter type and affected object, forming a set of operational items that may lead to anomalous changes.

[0198] Step S530: Analyze the spatiotemporal correlation between the parameter adjustment amount of each operation item in the abnormal correlation operation pool and the difference in the rate of change of the abnormal change node. Calculate the time interval from the operation execution time to the abnormal node, the spatial distance between the operation target and the abnormal parameter monitoring point, and construct a two-dimensional correlation coefficient by the reciprocal of the time interval and the reciprocal of the spatial distance. The correlation value increases as the time interval or spatial distance decreases.

[0199] In one implementation, step S530 may specifically include the following steps S531 to S536:

[0200] Step S531: Extract the execution time of each operation item in the abnormal associated operation pool, and subtract it from the time coordinate of the abnormal change node to obtain the time interval. The time interval is calculated by subtracting the operation execution time from the time coordinate of the abnormal node. A positive value indicates that the operation was executed before the abnormal node.

[0201] Extract the execution time of each operation item in the exception-related operation pool, and obtain the specific execution time point of each operation item from the exception-related operation pool. These time points record the actual time when the scheduling operation occurred.

[0202] The time interval is obtained by subtracting the time coordinate of the abnormal change node from the time coordinate of the abnormal change node. The execution time of each operation is then subtracted from the time coordinate of the abnormal change node on the time axis. The time interval is calculated by subtracting the operation execution time from the time coordinate of the abnormal node. A positive difference indicates that the operation was executed before the abnormal node; a negative difference indicates that the operation was executed after the abnormal node. This latter case is generally illogical and may indicate an error in data recording or analysis, requiring inspection and correction.

[0203] Step S532: Obtain the geographic coordinates of the object to be operated on and the geographic coordinates of the abnormal parameter monitoring point. Calculate the straight-line distance between the two using the coordinates to obtain the spatial distance. The unit of the spatial distance should be consistent with the spatial scale of the watershed model.

[0204] The process involves obtaining the geographic coordinates of the target object of the operation and the geographic coordinates of the abnormal parameter monitoring points. Both the target object (such as water conservancy facilities like reservoirs and sluices) and the abnormal parameter monitoring points (such as water level observation stations and flow monitoring points) have specific geographic locations. Their geographic coordinates can be obtained through a Geographic Information System (GIS) database or related monitoring equipment. The straight-line distance between the two is calculated using these coordinates to obtain the spatial distance. Based on the obtained geographic coordinates, spatial distance calculation methods, such as the Euclidean distance formula, are used to calculate the straight-line distance between the target object and the abnormal parameter monitoring points.

[0205] Step S533: Take the reciprocal of the time interval to obtain the time correlation factor. The smaller the time interval, the larger the time correlation factor. If the time interval is zero, the time correlation factor takes the preset maximum value.

[0206] Taking the reciprocal of the time interval yields the time correlation factor, which reflects the degree of temporal correlation between the operation item and the abnormal change node. A smaller time interval indicates that the execution time of the operation item is closer to the time of the abnormal change, the stronger the temporal correlation, and the larger the time correlation factor. If the time interval is zero, the time correlation factor is set to a preset maximum value. The strongest temporal correlation occurs when the execution time of the operation item is exactly the same as the time coordinate of the abnormal change node. To avoid infinitely large values, the time correlation factor in this case is set to a preset maximum value, which is predetermined based on the actual situation and computational requirements.

[0207] Step S534: Take the reciprocal of the spatial distance to obtain the spatial correlation factor. The smaller the spatial distance, the larger the spatial correlation factor. If the spatial distance is zero, the spatial correlation factor takes the preset maximum value.

[0208] Taking the reciprocal of the spatial distance yields the spatial correlation factor, which reflects the degree of spatial correlation between the operation item and the abnormal change node. The smaller the spatial distance, the closer the geographical location of the operation item's target and the abnormal parameter monitoring point, indicating a stronger spatial correlation and a larger spatial correlation factor. If the spatial distance is zero, the spatial correlation factor is set to a preset maximum value. The strongest spatial correlation occurs when the geographical locations of the operation item's target and the abnormal parameter monitoring point completely overlap. Similarly, to avoid an infinitely large value, the spatial correlation factor is set to the preset maximum value at this point.

[0209] Step S535: Multiply the time correlation factor and the spatial correlation factor to obtain the two-dimensional correlation coefficient. The correlation coefficient comprehensively reflects the correlation strength between the operation item and the abnormal change node in time and space.

[0210] Multiplying the temporal correlation factor and the spatial correlation factor yields a two-dimensional correlation coefficient, which comprehensively considers the temporal and spatial relationships between the operation item and the anomalous change node. The temporal correlation factor reflects the degree of correlation in the temporal dimension, while the spatial correlation factor reflects the degree of correlation in the spatial dimension. Multiplying the two provides a comprehensive index measuring the strength of the correlation between the operation item and the anomalous change node. The correlation coefficient comprehensively reflects the strength of the correlation between the operation item and the anomalous change node in both time and space. The larger the correlation coefficient, the stronger the correlation between the operation item and the anomalous change node in both time and space, and the more likely the operation item is to be the cause of the anomalous change. Using the two-dimensional correlation coefficient, operations in the anomalous correlation operation pool can be sorted and filtered to identify the operations with the strongest correlation to the anomalous change node.

[0211] Step S536: Normalize the two-dimensional correlation coefficients of all operation items in the abnormal correlation operation pool so that the coefficient values ​​are within a preset range. The normalized coefficient values ​​are used for subsequent operation item sorting.

[0212] The two-dimensional correlation coefficients of all operation items in the abnormal correlation operation pool are normalized. Since the two-dimensional correlation coefficients of different operation items may have different value ranges, these coefficients need to be normalized for easy comparison and sorting. Normalization maps the coefficient values ​​to a preset interval, usually the interval [0, 1].

[0213] By normalizing the coefficient values ​​to a preset range, the two-dimensional correlation coefficients are adjusted to the preset range, thus eliminating the scale difference between the coefficient values ​​of different operation items and making them comparable.

[0214] The normalized coefficient values ​​are used for subsequent operation item ranking. The normalized two-dimensional correlation coefficients can more accurately reflect the correlation strength between operation items and nodes with abnormal changes. Operation items in the abnormal correlation operation pool are ranked according to the normalized coefficient values, with larger coefficient values ​​ranked higher. These are the operation items most likely to cause abnormal changes, providing a basis for subsequently identifying the primary influencing operation items.

[0215] Step S540: Sort the operation items in the abnormal correlation operation pool based on the two-dimensional correlation coefficient, select the operation item with the highest correlation coefficient as the primary influencing operation item, and analyze the parameter chain change path caused by the execution of the operation item. By comparing the change order and magnitude of each parameter in the path, determine whether there is a situation where multiple operations are coordinated to cause anomalies.

[0216] The operations in the anomaly correlation operation pool are sorted based on a two-dimensional correlation coefficient. The normalized two-dimensional correlation coefficient is used to rank all operations in the pool from largest to smallest. The purpose of this sorting is to identify the operations with the highest correlation to the nodes experiencing anomalies. The operation with the highest correlation coefficient is selected as the primary influencing operation, as it has the strongest temporal and spatial correlation with the nodes experiencing anomalies and is most likely the main cause of the anomalies. This operation is thus identified as the primary influencing operation and will be the focus of subsequent correction and analysis.

[0217] Simultaneously, analyze the parameter chain change path triggered by the execution of this operation. The execution of the primary operation may trigger a series of parameter chain changes, forming a parameter chain change path. For example, the flood discharge operation of a reservoir may cause the downstream water level to rise, which may in turn affect the operation of the downstream sluice gate, and thus affect other related parameters. By analyzing this chain change path, understand the order and degree of the operation's impact on different parameters. By comparing the change order and magnitude of each parameter in the path, determine whether there is a situation where multiple operations are working together to cause anomalies. In the parameter chain change path, observe whether the change order and magnitude of each parameter conform to the influence pattern of a single operation. If it is found that the changes of some parameters cannot be explained by a single operation, or the change magnitude exceeds expectations, there may be a situation where multiple operations are working together to cause abnormal changes.

[0218] Step S550: Determine the parameter correction direction based on the direction of the difference in the rate of change of the abnormal change nodes. If the rate of change exceeds the upper limit of the normal range, reduce the parameter adjustment amount of the primary affected operation item. If the rate of change is lower than the lower limit of the normal range, increase the parameter adjustment amount of the primary affected operation item. Combine the product of the absolute value of the difference in the rate of change and the two-dimensional correlation coefficient to determine the initial correction magnitude.

[0219] The direction of parameter correction is determined by the direction of the difference in the rate of change of nodes with abnormal changes. This direction reflects whether the parameter change exceeds or falls below the normal range. When the rate of change exceeds the upper limit of the normal range, it indicates that the parameter is changing too rapidly, and measures need to be taken to reduce the rate of change. Therefore, the adjustment amount of the parameter that primarily affects the operation item should be reduced. When the rate of change is below the lower limit of the normal range, it indicates that the parameter is changing too slowly, and the rate of change needs to be increased. Therefore, the adjustment amount of the parameter that primarily affects the operation item should be increased.

[0220] The initial correction magnitude is determined by multiplying the absolute value of the rate of change difference with the two-dimensional correlation coefficient. The absolute value of the rate of change difference reflects the severity of the anomalous change, while the two-dimensional correlation coefficient reflects the strength of the correlation between the primary influencing operation and the anomalous change node. Multiplying the two allows for a comprehensive consideration of both the severity of the anomalous change and the correlation strength of the operation, thus determining an initial correction magnitude. For example, a larger absolute value of the rate of change difference indicates a more severe anomalous change, requiring a larger correction magnitude; a larger two-dimensional correlation coefficient indicates a stronger correlation between the operation and the anomalous change, suggesting that adjusting the parameters of that operation may be more effective in eliminating the anomalous change.

[0221] Step S560: Adjust the parameter adjustment amount of the primary influencing operation item according to the correction direction and the initial correction magnitude. If there are operation items with synergistic effects, adjust the parameters of the synergistic operation items synchronously according to the magnitude attenuation ratio of the chain change path. Generate a corrected scheduling instruction candidate set, input it again into the digital twin watershed model to perform simulation operation calculation, repeat the anomaly identification and parameter correction process until there are no abnormal change nodes in the simulation operation results, and output the water conservancy scheduling scheme.

[0222] Adjust the parameters that primarily affect the operational item according to the correction direction and initial correction magnitude. Based on the previously determined parameter correction direction and initial correction magnitude, adjust the parameter adjustments accordingly. If the correction direction is to decrease, subtract the initial correction magnitude from the original parameter adjustment; if the correction direction is to increase, add the initial correction magnitude to the original parameter adjustment. Through this adjustment, the parameters of the primary operational item are made to better conform to the requirements of the normal variation range.

[0223] If there are synergistic effects between operation items, their parameters should be adjusted synchronously according to the amplitude decay ratio of the cascading change path. When the analysis reveals anomalies caused by multiple synergistic operations, the synergistically affected operation items need to be adjusted synchronously. The amplitude decay ratio of the cascading change path reflects the degree of decay of the operation item's influence on different parameters. Based on this ratio, the parameters of the synergistically affected operation items should be adjusted accordingly to ensure that the synergistic effects between operation items are more reasonable and to avoid the recurrence of anomalous changes.

[0224] A revised set of scheduling instruction candidates is generated. The parameters of the adjusted primary and secondary impact operations are updated in the original set of scheduling instruction candidates, forming the revised set of scheduling instruction candidates. This set contains the revised scheduling operation instructions, which are more likely to enable the water conservancy project to reach a normal operating state in the simulation.

[0225] The simulation calculations are then performed again using the digital twin watershed model. The revised set of scheduling instruction candidates is re-inputted into the digital twin watershed model for a new round of simulation. Through the simulation, it is observed whether the parameter change trend has improved and whether there are still any abnormal change nodes.

[0226] The process of anomaly identification and parameter correction is repeated until no abnormal changes occur in the simulation results. Since a single correction may not completely eliminate abnormal changes, the process of anomaly identification and parameter correction needs to be repeated continuously. After each simulation, the results are analyzed to identify new abnormal change nodes, and the operation items causing the anomalies are located and corrected again before the simulation is run again. This continues until no abnormal change nodes with parameter change rates exceeding the preset range appear in the simulation results, indicating that the scheduling scheme has reached a relatively ideal state.

[0227] The above description is merely an embodiment of this application and is not intended to limit the scope of protection of this application. Any modifications, equivalent substitutions, and improvements made within the spirit and scope of this application are included within the scope of protection of this application.

Claims

1. A water conservancy scheduling and processing method based on digital twins, characterized in that, The method includes: Multi-source sensing data corresponding to historical extreme climate events are acquired. Based on the multi-source sensing data, the continuous change patterns of the events on the time scale and the coverage characteristics on the spatial scale are sorted out. The continuous change patterns and spatial coverage characteristics are associated with the corresponding water conservancy scheduling operation records to construct a historical event resource database containing the event's time change patterns, spatial coverage characteristics and associated scheduling operation records. The multi-source sensing data includes meteorological sensing data, hydrological sensing data and water conservancy project operation sensing data. Acquire real-time water conservancy monitoring data, match the real-time water conservancy monitoring data with the event time change pattern and spatial coverage characteristics in the historical event resource database, calculate the matching degree between the real-time water conservancy monitoring data and each historical event by comparing the time dimension and the spatial dimension, filter similar historical events that meet the preset requirements, extract the scheduling operation records associated with the similar historical events and organize them into a reference scheduling operation sequence in chronological order; Obtain current water conservancy project operation data, which includes current operation status parameters and operable constraint range of water conservancy project facilities. Map and adapt the reference scheduling operation sequence to the current water conservancy project operation data. Adjust the parameter values ​​of each operation item in the reference scheduling operation sequence according to the current operation status parameters and operable constraint range to generate a candidate set of scheduling instructions that meet the current project operation conditions. The candidate set of scheduling instructions is input into the digital twin watershed model, and simulation operation calculations are performed to simulate the entire process of water conservancy project operation under the guidance of the candidate set of scheduling instructions. The data on the change of water conservancy project operation status and the dynamic change of watershed hydrology output by the digital twin watershed model are recorded in real time to obtain simulation operation results including parameter change trends. The simulation results are analyzed to identify abnormal change nodes where the parameter change rate exceeds the preset range. The scheduling operation items that cause the abnormal change nodes are located and their parameter values ​​are corrected. The corrected scheduling instruction candidate set is then input into the digital twin watershed model for simulation calculation. The above correction and simulation process is repeated until there are no abnormal change nodes in the simulation results. Finally, the water conservancy scheduling scheme is output.

2. The method according to claim 1, characterized in that, The process involves acquiring multi-source sensing data corresponding to historical extreme climate events, analyzing the continuous variation patterns of these events over time and their spatial coverage characteristics based on this data, and associating these continuous variation patterns and spatial coverage characteristics with corresponding water conservancy dispatch operation records. This results in the construction of a historical event resource database containing the event's temporal variation patterns, spatial coverage characteristics, and associated dispatch operation records. The multi-source sensing data is fused in a hierarchical spatiotemporal manner. Time series alignment and outlier suppression are performed within the respective modalities of meteorological sensing data, hydrological sensing data, and water conservancy project operation sensing data. The correlation weights of different modal data are calculated through a cross-modal attention mechanism to generate a fused data cube with spatiotemporal consistency. Each data unit of the fused data cube contains multimodal sensing values ​​and correlation confidence. Temporal feature mining is performed on the fused data cube to extract pattern sequences of data changes at multiple time granularities. The phase transition characteristics of event development are identified by sequence pattern matching. Combining the duration and magnitude of change of the transition characteristics, an event time phase division model including the initiation phase, development phase, peak phase and decline phase is constructed to sort out the continuous change pattern of the event. Spatial topology features are extracted from the fused data cube to generate a state classification map of the watershed monitoring unit. The morphological parameters and spatial adjacency relationships of the abnormal state regions in the classification map are calculated. A spatial topology network of the abnormal regions is constructed, where network nodes represent abnormal sub-regions and edge weights represent the connectivity strength between sub-regions. The overall structural features of the topology network are used as the spatial coverage features of the event. The water conservancy scheduling operation records corresponding to the historical extreme climate events are analyzed, the engineering status change data before and after the operation is extracted, the direct correlation between the operation and the status change is identified, and an operation impact map including the operation object, operation intensity, impact time and chain reaction path is constructed. The chain reaction path is represented by the transmission delay and amplitude attenuation rate of the status change. The event time phase segmentation model is dynamically matched with the timestamps of the operation impact map. At the same time, the spatial correlation between each path and the abnormal sub-region in the operation impact map is calculated based on the node connectivity strength of the spatial topology network, generating a two-dimensional correlation matrix containing time dependency weights and spatial correlation weights. Based on the aforementioned two-dimensional association matrix, a dynamic index structure for the historical event resource database is constructed. The index key includes the event time stage identifier, the spatial topology network node ID, and the operation impact path identifier.

3. The method according to claim 1, characterized in that, The process involves acquiring real-time water conservancy monitoring data, matching the real-time water conservancy monitoring data with the event time change patterns and spatial coverage characteristics in the historical event resource database, calculating the matching degree between the real-time water conservancy monitoring data and each historical event through comparison of time and spatial dimensions, filtering similar historical events with matching degrees meeting preset requirements, extracting the scheduling operation records associated with the similar historical events, and organizing them into a reference scheduling operation sequence in chronological order, including: The real-time water conservancy monitoring data is processed in multiple dimensions. Meteorological, hydrological and engineering operation data are separated according to data type. The trend analysis of adjacent time intervals of each type of data is performed. Isolated data points that deviate from the overall trend are identified and removed. For the missing time intervals in the data sequence, the trend extension is filled according to the slope of the change of the preceding and following data segments to generate a real-time monitoring data sequence. The monitoring data of the most recent observation period is extracted from the continuous and complete real-time monitoring data sequence. The unit time change sequence, extreme value location distribution and cumulative consistency of the continuous change direction of the data within the most recent observation period are calculated. A real-time data feature set containing temporal dynamic features and spatial distribution features is constructed. The cumulative consistency of the continuous change direction is represented by the number of continuous data points with the same change sign. Traverse the historical events in the historical event resource library, and for each historical event, extract the trend characteristics of each stage in the event time change pattern and the abnormal area morphological characteristics in the spatial coverage characteristics, and construct a historical event feature template containing time trend sequence and spatial morphological parameters. Each historical event feature template corresponds to a unique historical event identifier. In the time dimension, the unit time change sequence of the real-time data feature set is compared with the time trend sequence of the historical event feature template segment by segment. The number of times the change direction is consistent and the degree of closeness of the absolute value of the change are calculated within the corresponding time period to generate a time matching score, wherein the degree of closeness is represented by the reciprocal of the difference in change. In the spatial dimension, the current spatial distribution pattern of the continuous and complete real-time monitoring data sequence is compared with the spatial morphological parameters of the historical event feature template to calculate the regional similarity. The number and area ratio of sub-regions that conform to the morphological characteristics of historical abnormal regions in the current monitoring region are counted to generate a spatial matching score. The time matching score and spatial matching score are combined proportionally to generate a comprehensive event matching degree. Historical events with a comprehensive matching degree that meet the set conditions are selected as similar historical events. The scheduling operation records associated with similar historical events are extracted, and the time validity of the operation records is verified. The operation records executed within the corresponding stage of the event time change pattern are retained, and a reference scheduling operation sequence is generated by organizing the operation execution time in order.

4. The method according to claim 3, characterized in that, The process involves extracting monitoring data from the most recent observation period from a continuous and complete real-time monitoring data sequence, calculating the cumulative consistency of the unit-time change sequence, the distribution of extreme value locations, and the direction of continuous change within that most recent observation period, and constructing a real-time data feature set that includes both temporal dynamic features and spatial distribution features. The monitoring data of the most recent observation period is extracted from the continuous and complete real-time monitoring data sequence, and the period is divided into multiple time segments of equal length according to the data acquisition time interval. Each time segment contains multiple continuous monitoring data points. Linear fitting is performed on the monitoring data points within each time segment, and the slope of the fitted line is calculated as the unit time change of that segment. The unit time change of all time segments is arranged sequentially to obtain the unit time change sequence. Traverse the monitoring data throughout the entire observation period, identify the maximum and minimum points in the data sequence, record the specific time location and value of each extreme point within the observation period, count the number of extreme points distributed in each time segment, and generate the distribution characteristics of extreme value occurrence locations. Analyze the direction of change of continuous data points. When the value of the next point of an adjacent data point is greater than the value of the previous point, it is marked as a positive change; when it is less than the previous point, it is marked as a negative change; when it is equal to the previous point, it is marked as a zero change. Calculate the maximum length and total length of consecutive positive or negative changes, and generate a cumulative consistency index of the direction of continuous change. The spatial distribution data of the continuous and complete real-time monitoring data sequence is gridded, the difference between the value of each grid cell and the mean value of the surrounding adjacent grid cells is calculated, the grid cells are divided into different state levels according to the size of the difference, the number of grid cells and spatial clustering of each state level are counted, and spatial distribution characteristics are generated. The sequence of changes per unit time, the distribution characteristics of extreme value occurrence locations, the cumulative consistency index of continuous change direction, and the spatial distribution characteristics are combined in the order of temporal dynamic characteristics first and spatial distribution characteristics last to construct a real-time data feature set containing multi-dimensional sub-features.

5. The method according to claim 1, characterized in that, The process involves acquiring current water conservancy project operation data, which includes current operating status parameters and operable constraints of the water conservancy project facilities. This includes mapping and adapting the reference scheduling operation sequence to the current water conservancy project operation data, adjusting the parameter values ​​of each operation item in the reference scheduling operation sequence according to the current operating status parameters and operable constraints, and generating a candidate set of scheduling instructions that meets the current project operation conditions. This includes: The reference scheduling operation sequence is parsed, and the execution time stamp, target identifier, and original operation parameters of each operation item are extracted. The operation items are grouped according to the target identifier, the time interval distribution of the operation items in the same group is calculated, and continuous operation items with decreasing time intervals are identified and marked as operation sequences with potential associations. A basic list of operation items containing potential association markers is generated. Extract the current operating status parameters corresponding to the target identification from the current water conservancy project operation data, including real-time equipment operation indicators, structural fatigue cumulative values ​​and surrounding hydrological environment parameters. Calculate the synergistic change coefficient between indicators through cross-analysis of the status parameters. The synergistic change coefficient is represented by the consistency of the change direction of the two parameters and the ratio of the change magnitude, and generate the target status correlation matrix. Obtain the operable constraint range from the current water conservancy project operation data, organize the upper and lower limit thresholds, single adjustment boundary and operation interval threshold of the operation parameters according to the object identification, and adjust the boundary value of the constraint range by combining the cooperative change coefficient in the object status correlation matrix to generate a dynamic constraint parameter table. For each operation item, its original operation parameters are compared with the upper and lower limits of the dynamic constraint parameter table to calculate the parameter deviation. At the same time, the historical execution effect data of the preceding operation items in the potential related operation sequence are traced. The residual impact value of the preceding operation on the current operation is calculated through effect propagation analysis. The residual impact value is calculated by the change amplitude and decay rate of the state parameters after the preceding operation. Based on the co-change coefficient of the state correlation matrix of the target object, the influence weight of each state parameter on the operating parameter is determined. The higher the co-change coefficient of the state parameter and the operating parameter, the greater the influence weight. Combined with the parameter deviation and considering the system state offset trend indicated by the residual influence value, the adjustment direction and initial adjustment magnitude of the operating parameter are calculated. Based on the single adjustment boundary and operation interval threshold of the dynamic constraint parameter table, the initial adjustment range is corrected so that the adjusted operation parameters do not exceed the upper and lower thresholds and the adjustment amount is within the single adjustment boundary. All corrected operation items are sorted according to the execution time mark to generate a scheduling instruction candidate set.

6. The method according to claim 5, characterized in that, The process involves parsing the reference scheduling operation sequence, extracting the execution time stamp, target identifier, and original operation parameters for each operation item, grouping the operation items by target identifier, calculating the time interval distribution of operation items within the same group, identifying consecutive operation items with decreasing time intervals, marking them as operation sequences with potential associations, and generating a basic list of operation items containing potential association markers, including: Iterate through each operation item in the reference scheduling operation sequence, extract the execution time stamp and convert it into a time offset relative to the start time of the sequence, extract the target identifier and original operation parameters, assign a unique identifier to each operation item, and generate an operation item information table containing the identifier, time offset, target identifier, and original operation parameters. The operation items in the operation item information table are grouped according to the target object identifier. The operation items in the same group contain the same target object identifier, generating multiple target object operation groups. The operation items in each group are arranged in ascending order according to the time offset. For each group of operations, the time offset difference between adjacent operations is calculated to obtain a time interval sequence. Trend analysis is performed on the time interval sequence to identify subsequences that show a continuous decreasing trend. These subsequences must show the time interval decreasing feature multiple times consecutively and are marked as potentially related subsequences. Analyze the changing trends of the original operation parameters of each operation item in the potentially related subsequence, calculate the ratio of the parameter changes of adjacent operation items, and if the ratio shows an increasing or decreasing characteristic, the subsequence is confirmed to be an operation sequence with correlation; otherwise, the marking is canceled. For the confirmed sequence of associated operations, the operation items are marked in order of their position in the sequence. The first operation item is marked as the starting operation, and subsequent operation items are marked as associated subsequent operations. The correspondence between the identifiers of the starting operation and the associated subsequent operation is recorded. Add an association marker field to the operation item information table, mark the operation item in the associated operation sequence with the associated sequence number and the position number in the sequence, and generate a basic list of operation items containing potential association markers. The association marker field of non-associated operation items is empty.

7. The method according to claim 5, characterized in that, Based on the cooperative change coefficients of the state correlation matrix of the affected object, the influence weight of each state parameter on the operating parameter is determined. The higher the cooperative change coefficient between the state parameter and the operating parameter, the greater the influence weight. Combining the parameter deviation and considering the system state offset trend indicated by the residual influence value, the adjustment direction and initial adjustment magnitude of the operating parameters are calculated, including: Extract the cooperative change coefficients of the operation parameters and each state parameter from the state correlation matrix of the object of action. A positive cooperative change coefficient indicates that the change direction is consistent, and a negative coefficient indicates that the change direction is opposite. The absolute value indicates the degree of correlation of the change magnitude. Calculate the relative distance between the current value and the critical value of each state parameter. The relative distance is represented by the ratio obtained by comparing the difference between the current value and the critical value with the critical value. The smaller the relative distance, the closer the state parameter is to the critical state. Multiplying the absolute value of the cooperative change coefficient by the relative distance yields the comprehensive influence value of the state parameter on the operating parameter. The larger the comprehensive influence value, the more significant the influence of the state parameter on the operating parameter. The overall influence value of all state parameters is normalized to obtain the influence weight of each state parameter. The sum of the influence weights is 1. The state parameter with the larger overall influence value has a larger influence weight. The positive and negative directions of the parameter deviation are analyzed. If the original operating parameter is higher than the upper limit threshold, the adjustment direction is to decrease; if it is lower than the lower limit threshold, the adjustment direction is to increase. The adjustment component contributed by each state parameter is calculated in combination with the influence weight. The adjustment component is obtained by multiplying the parameter deviation by the influence weight. Add up the adjustment components of all state parameters to obtain the initial adjustment magnitude based on the parameter deviation; analyze the sign of the residual influence value. If the residual influence value is positive, it means that the historical operation has caused the system state to shift in the same direction as the adjustment direction, so the initial adjustment magnitude is reduced. If it is negative, it means that it has shifted in the opposite direction, so the initial adjustment magnitude is increased. Generate the final operation parameter adjustment direction and initial adjustment magnitude.

8. The method according to claim 1, characterized in that, The process involves inputting the candidate set of scheduling instructions into a digital twin watershed model, performing simulation calculations, simulating the entire operation process of the water conservancy project under the guidance of the candidate set of scheduling instructions, and recording in real time the water conservancy project operation status change data and watershed hydrological dynamic change data output by the digital twin watershed model to obtain simulation operation results including parameter change trends, including: The scheduling instruction candidate set is parsed, and the time offset, target identifier, and correction operation parameters of each operation item are extracted. For operation items with the same target identifier, the correspondence between the time offset and the parameter adjustment amount is analyzed. The continuous pattern of parameter adjustment over time is identified by the ratio of parameter adjustment amounts of adjacent operations, and a model input instruction stream containing pattern features is generated. The current operating status parameters in the current water conservancy project operation data are mapped to the digital twin watershed model. At the same time, watershed topographic data and initial hydrological observation data are loaded, and related parameters are transmitted through the data interface, so that the project operation data and hydrological environment data form a dynamic response relationship. Operation commands are initiated sequentially according to the time offset of the input command stream of the model. Before triggering each command, the real-time running data of the target object is obtained. If the real-time running data meets the environmental conditions for operation execution, the operation is executed and the execution time is recorded. If it does not meet the environmental conditions, the current state is maintained and the next time interval is waited for re-evaluation. During the simulation operation, data on the changes in the operating status of water conservancy projects are collected at set time intervals, including the status parameters of the affected objects after operation, response delay data, and status linkage data of related facilities. Simultaneously, dynamic changes in watershed hydrology are collected, including hydrological observation values ​​in different regions and hydrological transfer data between regions, generating a multi-source simulation data set. By establishing the spatial correspondence between the geographical location of the target and the spatial relationship between the hydrological observation area, the spatial correspondence between engineering data and hydrological data is established. At the same time, the collection times of different types of data are aligned according to the time interval to generate spatiotemporal correlation data pairs. The direction of change and the amount of change per unit time of each parameter in the spatiotemporal correlation data are analyzed. The direction of change is determined by comparing the parameter values ​​of adjacent time intervals, and the amount of change per unit time is calculated by the ratio of the parameter difference to the time interval. Based on the direction of change and the amount of change per unit time, the trend curves of each parameter are generated. By combining all the trend curves and the corresponding spatiotemporal correlations, the simulation results containing the parameter change trends are obtained.

9. The method according to claim 8, characterized in that, The process involves parsing the candidate set of scheduling instructions, extracting the time offset, target identifier, and correction parameters for each operation item, analyzing the correspondence between the time offset and parameter adjustment for operations with the same target identifier, identifying continuous patterns of parameter adjustment over time by comparing the parameter adjustment ratios of adjacent operations, and generating a model input instruction stream containing pattern features, including: Traverse each operation item in the candidate set of scheduling instructions, extract the time offset, the target identifier and the correction operation parameters, assign a unique instruction identifier to each operation item and record it in the instruction basic information table; The operation items in the instruction basic information table are grouped according to the target identifier, so that each group of operation items has the same target identifier. The operation items within the group are arranged in natural order of time offset to form a target operation sequence. For each operation sequence of the target object, calculate the ratio of parameter adjustment amounts of adjacent operation items, and identify the continuous pattern of parameter adjustment changes over time by the changing pattern of the ratio; Operations with the same continuous pattern are divided into one operation phase. The starting instruction identifier and the number of operations contained in each operation phase are recorded to generate the phase division result. Within each operational phase, the mean and fluctuation range of the parameter adjustment ratio are calculated, and the mean and fluctuation range are used as the pattern characteristics of that phase and added to the phase division results. By combining the operation sequence of the target object with the stage division results, a model input command stream is generated, which includes command identifier, time offset, correction operation parameters, target object identifier, and the mode characteristics of the stage to which it belongs.

10. A water conservancy dispatching system, characterized in that, include: Memory is used to store executable instructions or computer programs. The processor, when executing computer-executable instructions or computer programs stored in the memory, implements the water conservancy scheduling processing method based on digital twins as described in any one of claims 1 to 9.

Images