Irrigation area water regime multi-source data real-time processing method based on edge computing

Abstract

This invention relates to the field of irrigation district water situation data processing technology, specifically a real-time processing method for multi-source irrigation district water situation data based on edge computing. The method includes: acquiring a time-series data set based on the irrigation district water situation monitoring network; extracting the temporal coupling characteristics between water level and flow rate, the response characteristics between water level and gate opening, and the lag correlation characteristics between flow rate and soil moisture through cross-source data correlation analysis; constructing a multi-dimensional correlation feature set and building a dynamic water situation evolution map with monitoring point nodes, associated directed edges, and relational weights; inputting the map into a pre-trained water situation fusion model of the edge computing nodes for fusion calculation; generating a full-domain fused water situation status map containing calculated water level, flow rate, water level change trend, and estimated water consumption; quantifying the correlation degree of water situation points; reducing data transmission and processing latency; achieving deep fusion of multi-source water situation data; and fully presenting the overall water situation operation status of the irrigation district.

Description

Technical Field

[0001] This invention relates to the field of irrigation district water information data processing technology, and in particular to a method for real-time processing of multi-source irrigation district water information data based on edge computing. Background Technology

[0002] Current irrigation district water situation monitoring mainly collects single-dimensional time-series data, and data processing only involves simple numerical statistics. Cross-source correlation analysis is not carried out for multi-source water situation data, the correlation characteristics of water situation-related elements are not specifically extracted, the correlation between monitoring points is only simply labeled, and a map structure with quantitative weights is not constructed. Water situation data processing relies heavily on cloud server computing, and dedicated models are not deployed on edge computing nodes to carry out water situation fusion and inference. In the end, only scattered water situation monitoring values ​​can be output, which cannot form a comprehensive water situation status presentation of the entire area.

[0003] The inherent correlations between multi-source hydrological data cannot be accurately mined; the correlation features between water level and flow rate, water level and gate opening, and flow rate and soil moisture cannot be fully extracted; the flow correlations at monitoring points cannot be represented by weighted values; and the logic of hydrological evolution lacks a concrete expression. Cloud processing suffers from data transmission delays, hydrological estimation is limited to a single dimension, and it cannot simultaneously generate multiple types of hydrological estimation indicators and a comprehensive status map, thus failing to fully present the overall hydrological status of the irrigation area. Currently, it is necessary to achieve cross-source correlation feature extraction from multi-source hydrological data and the construction of weighted dynamic hydrological evolution maps, while also enabling edge-end models to perform hydrological fusion estimation based on the maps and generate comprehensive multi-dimensional hydrological status maps. Summary of the Invention

[0004] The purpose of this invention is to address the shortcomings of existing technologies by proposing a real-time processing method for multi-source irrigation data based on edge computing.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: a real-time processing method for multi-source irrigation district water situation data based on edge computing, comprising:

[0006] Based on the irrigation district water monitoring network, obtain a set of water situation time series data;

[0007] Cross-source data correlation analysis was performed on the aforementioned hydrological time series data set to extract the temporal coupling characteristics between water level and flow rate, the response characteristics between water level and gate opening, and the hysteretic correlation characteristics between flow rate and soil moisture, thus forming a multidimensional correlation feature set.

[0008] Based on the multidimensional association feature set, a dynamic hydrological evolution map is constructed. The dynamic hydrological evolution map uses nodes to represent hydrological monitoring points and directed edges to represent the water flow direction and association characteristics between monitoring points. Each directed edge is assigned a relation weight calculated from the multidimensional association feature set.

[0009] The pre-trained hydrological fusion model deployed on the edge computing node is invoked, and the dynamic hydrological evolution map is used as input to perform hydrological status fusion and estimation processing to generate a fused hydrological status map of the entire irrigation area. The fused hydrological status map includes the estimated water level, estimated flow rate, water level change trend and estimated regional water consumption of each monitoring point.

[0010] As a further aspect of the present invention, the acquisition of a time-series water situation data set based on the irrigation district water situation monitoring network includes:

[0011] At each edge computing node in the irrigation district water monitoring network, multi-source water data from various heterogeneous sensors are collected and aggregated synchronously. The multi-source water data includes water level time series data, flow time series data, gate opening status data, and soil moisture time series data.

[0012] The collected water level time series data, flow time series data, gate opening status data and soil moisture time series data are processed for data alignment and timestamp synchronization to generate a timestamp-aligned formatted multi-source hydrological data stream.

[0013] Outlier detection and cleaning are performed on each time series data in the formatted multi-source hydrological data stream to remove outlier noise points and form a hydrological time series data set.

[0014] The process of aligning and synchronizing the collected water level time-series data, flow rate time-series data, gate opening status data, and soil moisture time-series data to generate a time-stamp-aligned formatted multi-source hydrological data stream specifically includes:

[0015] The raw time-series data streams of each sensor are timestamped using a unified time reference, which is provided by a network time protocol service deployed on edge computing nodes;

[0016] Detect and complete the missing timestamps in each time-series data stream due to different sensor sampling frequencies or communication delays, forming a complete sequence with continuous timestamps;

[0017] The complete sequence is resampled according to a preset high-frequency time grid to ensure that all types of data have corresponding values ​​at the same time point, thereby generating a formatted multi-source hydrological data stream with timestamp alignment.

[0018] As a further aspect of the present invention, outlier detection and cleaning processing are performed on each time-series data item in the formatted multi-source hydrological data stream, including:

[0019] For each type of hydrological time-series data, the dynamic change threshold range is calculated based on its historical data;

[0020] The real-time data value at the current moment is compared with the dynamic change threshold range to identify outlier points that exceed the dynamic change threshold range;

[0021] For the identified outlier points, linear interpolation or spline interpolation is performed based on the normal data at the adjacent time points to replace the outlier points, thus completing the cleaning process.

[0022] As a further aspect of the present invention, cross-source data correlation analysis is performed on the aforementioned hydrological time-series data set to extract the temporal coupling characteristics between water level and flow rate, including:

[0023] Within a specified time window, cross-correlation analysis is performed on the water level time series data and flow time series data in the hydrological time series data set to calculate the phase difference and correlation coefficient between water level changes and flow changes.

[0024] A water level-flow joint change model was constructed, and the flow change caused by a unit change in water level was calculated through regression analysis, which was used as the static response coefficient.

[0025] The interaction between the water level time series data and the flow rate time series data in the high-frequency and low-frequency components is analyzed, and the delay coupling feature reflecting the dynamic characteristics of water flow propagation is extracted. Together with the phase difference, correlation coefficient and static response coefficient, it constitutes the time series coupling feature between the water level and the flow rate.

[0026] As a further aspect of the present invention, cross-source data correlation analysis is performed on the aforementioned hydrological time-series data set to extract the response features between water level and gate opening, including:

[0027] Based on the water level time series data at the location of the gate and the corresponding gate opening status data, analyze the amplitude, rate and stabilization time of the upstream and downstream water level changes after the gate opening change event occurs.

[0028] Calculate the nonlinear mapping relationship between the rate of change of water level and the amount of change of gate opening within different gate opening ranges;

[0029] The timeliness and intensity of the effect of gate opening adjustment on water level are quantified to form the response characteristics between water level and gate opening, which include response amplitude, response rate, nonlinear mapping relationship and timeliness.

[0030] As a further aspect of the present invention, the construction of the dynamic hydrological evolution map includes:

[0031] Each water level monitoring point within the irrigation area is used as a node in the dynamic water level evolution map, and each node contains the real-time water level and flow rate attributes of the water level monitoring point.

[0032] Based on the actual topology of the irrigation area's water system, directed edges are established between monitoring points that have upstream and downstream or water flow connectivity, and the direction of the directed edges represents the direction of water flow.

[0033] Each directed edge is assigned an attribute, which includes association feature values ​​extracted from the multidimensional association feature set and related to the two monitoring points connected by the directed edge. The association feature values ​​are then normalized and used as the relationship weight.

[0034] As a further aspect of the present invention, the step of calling the pre-trained hydrological fusion model deployed on the edge computing node, taking the dynamic hydrological evolution map as input, and performing hydrological state fusion and estimation processing to generate a fused hydrological state map of the entire irrigation district includes:

[0035] The dynamic hydrological evolution map is input into the graph neural network encoder of the pre-trained hydrological fusion model to encode each node and its connected edges, generating the node's embedding vector representation.

[0036] In the graph propagation layer of the pre-trained hydrological fusion model, based on the relation weights, the embedding vector representation of each node is updated through multiple rounds of information propagation and aggregation, so that the state information of the nodes is transmitted and fused along the graph structure.

[0037] In the decoding output layer of the pre-trained hydrological fusion model, the updated embedding vector representation of each node is decoded into specific state values. The specific state values ​​include at least the estimated water level, estimated flow rate, and water level change trend. By accumulating the estimated flow rate values ​​on the relevant paths, the estimated water consumption value of the region is obtained, and finally the fused hydrological state map is formed.

[0038] As a further aspect of the present invention, it also includes:

[0039] Based on the estimated regional water consumption and the water level change trend at each monitoring point in the integrated water situation map, and combined with the preset irrigation district water use plan, the water demand gap analysis and water allocation strategy generation process are performed to obtain the gate control suggestion scheme for the current scheduling cycle.

[0040] The proposed gate control scheme is converted into a specific gate control command sequence, and the gate control command sequence is sent to the corresponding gate actuators in the irrigation area.

[0041] Based on the estimated regional water consumption and water level change trends at each monitoring point in the integrated water situation map, and in conjunction with the preset irrigation district water use plan, a water demand gap analysis and water allocation strategy generation process are performed to obtain a gate control suggestion scheme for the current scheduling cycle, specifically including:

[0042] From the preset irrigation district water use plan, obtain the planned water use in different sub-irrigation districts within the current scheduling cycle;

[0043] The estimated regional water consumption of each sub-irrigation area calculated from the integrated water situation map is compared with the corresponding planned water consumption to calculate the water demand gap.

[0044] Analyze the water level change trends at each monitoring point, predict the water level trend in the near future, and identify risk areas where the water level is below the warning value.

[0045] Based on the combined results of the water demand gap and water level risk prediction, and under the condition of meeting the channel safety water level constraints, with the goal of minimizing the overall water demand gap, the recommended opening adjustment amount and timing of each gate are determined, thus forming the gate control recommendation scheme.

[0046] As a further aspect of the present invention, the gate control suggestion scheme is converted into a specific gate control command sequence, including:

[0047] The gate control suggestion scheme is analyzed to extract the suggested opening adjustment amount, suggested adjustment start time, and suggested adjustment duration for each target gate;

[0048] Based on the mechanical characteristics of the target gate actuator, the suggested opening adjustment amount is converted into the number of motor rotation pulses or the number of valve stroke steps;

[0049] The suggested adjustment start time, suggested adjustment duration, and converted motor rotation pulse count or valve stroke step count are encapsulated into an instruction data packet that can be directly recognized and executed by the gate controller, according to a preset gate control instruction protocol.

[0050] Following the suggested adjustment of the start time order, all instruction data packets are sorted to form the gate control instruction sequence.

[0051] As a further aspect of the present invention, the step of sorting all instruction data packets according to the suggested order of adjusting the start times to form the gate control instruction sequence further includes:

[0052] Check whether there is any overlap or conflict in the suggested adjustment start time of different gates. If there is a conflict, make a fine adjustment to the suggested adjustment start time of some gates according to the water flow propagation time and the priority of gate adjustment to ensure a smooth transition of the overall water flow.

[0053] After fine-tuning is completed, an ordered list of instruction data packets is regenerated according to the final adjustment time sequence, which serves as the final sequence of gate control instructions to be issued.

[0054] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0055] Cross-source data correlation analysis was conducted on the hydrological time series data set to extract the temporal coupling characteristics between water level and flow rate, the response characteristics between water level and gate opening, and the lag correlation characteristics between flow rate and soil moisture, forming a multi-dimensional correlation feature set. Based on this, a dynamic hydrological evolution map was constructed, with nodes representing hydrological monitoring points and directed edges representing the water flow direction and correlation characteristics between monitoring points. Each directed edge was assigned a relation weight calculated from the multi-dimensional correlation feature set. The linkage characteristics between various hydrological elements were fully decomposed and presented. The water flow transmission and data correlation between monitoring points were intuitively displayed through the map structure. The assignment of relation weights quantified the degree of correlation between points. The internal logic of multi-source hydrological data was systematically sorted out, and the laws of hydrological evolution were visualized through the map structure. The integration of various correlation features made the depiction of hydrological status closely match the actual hydrological operation status of the irrigation area.

[0056] By calling a pre-trained hydrological fusion model deployed on edge computing nodes and taking the dynamic hydrological evolution map as input, hydrological status fusion and extrapolation processing is performed to generate a fusion hydrological status map of the entire irrigation district, which includes extrapolated water levels, extrapolated flow rates, water level change trends, and estimated regional water consumption for each monitoring point. The model deployment at the edge reduces the process of remote data transmission, and the map input provides the model with complete hydrological correlation basis. The fusion extrapolation process can simultaneously complete the calculation and output of multiple types of hydrological indicators. The hydrological status of the entire irrigation district is integrated and presented through a unified status map. Extrapolated values, trend information, and estimated water consumption are output simultaneously. The hydrological data processing flow is adapted to the needs of real-time monitoring scenarios, and the presentation of the entire hydrological status is more complete. The real-time nature of data processing and the richness of results are mutually compatible. Attached Figure Description

[0057] Figure 1 This is a flowchart of the real-time processing method for multi-source irrigation water situation data based on edge computing described in this invention;

[0058] Figure 2 A flowchart for extracting the temporal coupling characteristics between water level and flow rate;

[0059] Figure 3 A time-series monitoring data chart of multi-source water conditions in the irrigation area;

[0060] Figure 4 A diagram illustrating the cross-source water situation correlation characteristics of the irrigation area;

[0061] Figure 5 A bar chart showing the grouping of gate control commands for the irrigation area. Detailed Implementation

[0062] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0063] In the description of this invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, in the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0064] See Figure 1 This invention provides a real-time processing method for multi-source irrigation district water situation data based on edge computing. The method includes: firstly, acquiring a set of time-series water situation data continuously collected and reported by these sensors, based on a deployed irrigation district water situation monitoring network, which includes sensing devices and edge computing nodes distributed at various key monitoring points. Next, conducting in-depth cross-source data correlation analysis on this data set, extracting dynamic temporal coupling characteristics between water level and flow rate, response characteristics of water level to gate opening changes, and hysteretic correlation characteristics between flow rate changes and soil moisture changes from the multi-source heterogeneous data. These characteristics together constitute a multi-dimensional correlation feature set reflecting the inherent correlation of irrigation district water situation. Subsequently, using each water situation monitoring point in the irrigation district as a node, directed edges are established between nodes with water flow relationships according to the actual water system topology, thereby constructing a dynamic water situation evolution map. The direction of the directed edges in this map represents the direction of water flow, and each directed edge is assigned a relationship weight calculated from the aforementioned multi-dimensional correlation feature set. Finally, the pre-trained water situation fusion model deployed on the edge computing node is invoked. The constructed dynamic water situation evolution map is input into the model. The model integrates and extrapolates the water situation status of the entire area through the information propagation and aggregation mechanism on the graph structure, and outputs a fusion water situation status map of the entire irrigation area. This map includes the extrapolated water level, extrapolated flow, water level change trend of each monitoring point, and the estimated regional water consumption based on the flow information.

[0065] In one embodiment of the present invention, at each edge computing node in the irrigation district water monitoring network, multi-source water data from various heterogeneous sensors are synchronously collected and aggregated. This multi-source water data includes time-series data of water level, flow rate, gate opening status, and soil moisture. The aggregated time-series data of water level, flow rate, gate opening status, and soil moisture are processed for data alignment and timestamp synchronization to generate a timestamp-aligned formatted multi-source water data stream. Specifically, this process involves: timestamp calibration of the original time-series data streams from each sensor using a unified time reference provided by the network time protocol service deployed on the edge computing nodes; detecting and supplementing timestamp missing points in each time-series data stream caused by different sensor sampling frequencies or communication delays to form a complete sequence with continuous timestamps; and resampling the complete sequence according to a preset high-frequency time grid to ensure that all types of data have corresponding values ​​at the same time points, thereby generating a timestamp-aligned formatted multi-source water data stream. Outlier detection and cleaning are performed on each time-series data item in the formatted multi-source hydrological data stream. For each type of hydrological time-series data, a dynamic change threshold range is calculated based on its historical data. The real-time data value at the current moment is compared with the dynamic change threshold range to identify outliers that exceed the threshold. For the identified outliers, linear interpolation or spline interpolation is performed based on the normal data at the adjacent moments to replace the outliers, completing the cleaning process and ultimately forming a hydrological time-series data set for subsequent analysis.

[0066] In its implementation, the irrigation district's water monitoring network deployed an edge computing node along the "North Main Canal" section. This edge computing node synchronously collects and aggregates multi-source water information from four types of heterogeneous sensors: pressure level sensors installed on the canal wall generate time-series water level data, ultrasonic flow meters installed at the canal cross-section generate time-series flow rate data, gate position gauges installed on the control gate generate gate opening status data, and dielectric constant soil moisture sensors installed in the farmland beside the canal generate soil moisture time-series data. These sensors operate at different sampling frequencies: the water level sensor uploads a data packet every 5 seconds, the flow sensor uploads a data packet every 10 seconds, the gate opening status data is actively reported when the status changes, and the soil moisture sensor uploads a data packet every minute. The edge computing node continuously receives and caches these raw data streams.

[0067] The collected time-series data on water level, flow rate, gate opening status, and soil moisture are aligned and time-stamped to generate a formatted, time-stamp-aligned multi-source hydrological data stream. In practice, edge computing nodes have a built-in Network Time Protocol (NTP) client that obtains accurate Coordinated Universal Time (UTC) from the NTP server as a unified time reference. The edge computing nodes use this unified time reference to timestamp-calibrate each received raw data packet, replacing any potentially inaccurate local timestamps from the sensors themselves. For example, a data packet sent by a flow sensor at its local timestamp "08:00:05" will have its timestamp corrected to UTC "2026-03-2300:00:05.120" after NTP calibration. It is understandable that delays and packet loss exist in sensor network transmission. If an edge computing node detects that there is no data for the "flow time series data" at the time point "2026-03-2300:00:15" in UTC, it will use the flow values ​​at the preceding and following times "2026-03-2300:00:10" and "2026-03-2300:00:20" to calculate an estimated value through linear interpolation to complete the sequence, thus forming a complete sequence with continuous timestamps. Optionally, the edge computing node will then resample the complete sequence according to a preset high-frequency time grid of once per second (1Hz). For data such as "soil moisture time series data" with a native sampling frequency lower than 1Hz, the data will be filled in at non-original sampling time points by maintaining the previous valid value, ultimately generating a formatted multi-source hydrological data stream with timestamps aligned for all data at every second.

[0068] In some embodiments, outlier detection and cleaning are performed on each time-series data item in the formatted multi-source hydrological data stream. For hydrological time-series data of the type "water level time-series data," edge computing nodes calculate the dynamic change threshold range based on their historical data from the past 24 hours. The calculation follows the formula:

[0069]

[0070]

[0071] in: The arithmetic mean of historical water level time series data. This represents the standard deviation of historical water level time-series data. For example, the historical average water level. It is 1.85 meters, with a standard deviation of 1.85 meters. If the value is 0.05 meters, then the dynamic change threshold range is [1.70 meters, 2.00 meters]. The edge computing node compares the current real-time water level data value "2.15 meters" with the dynamic change threshold range [1.70 meters, 2.00 meters]. Since "2.15 meters" exceeds the upper limit of the threshold "2.00 meters", the edge computing node identifies this as an outlier. For the identified outlier, the edge computing node cleans it based on the normal data of its immediate preceding and following moments. It performs linear interpolation between the previous water level value "1.88 meters" and the next water level value "1.86 meters" to calculate the interpolated value "1.87 meters" and replaces the original outlier value "2.15 meters", thus completing the cleaning of the "water level time series data".

[0072] In one embodiment of the present invention, see [reference] Figure 2 Within a specified time window, cross-correlation analysis is performed on water level and flow rate time series data in the hydrological time series dataset to calculate the phase difference and correlation coefficient between water level and flow rate changes. A joint water level-flow rate change model is constructed, and the flow rate change caused by a unit change in water level is calculated through regression analysis as the static response coefficient. The interaction relationship between water level and flow rate time series data on high-frequency and low-frequency components is analyzed to extract the delay coupling characteristics reflecting the dynamics of water flow propagation. These characteristics, along with the phase difference, correlation coefficient, and static response coefficient, constitute the temporal coupling characteristics between water level and flow rate. Cross-source data correlation analysis is performed on the hydrological time series dataset to extract the response characteristics between water level and gate opening. Based on the water level time series data at the gate location and the corresponding gate opening status data, the amplitude, rate, and stabilization time of upstream and downstream water level changes after a gate opening change event are analyzed. The nonlinear mapping relationship between the rate of change of water level and the amount of change of gate opening is calculated within different gate opening intervals. The timeliness and intensity of the effect of gate opening adjustment on water level are quantified, forming the response characteristics between water level and gate opening that include response amplitude, response rate, nonlinear mapping relationship and timeliness.

[0073] Cross-source data correlation analysis was performed on the hydrological time-series data set to extract the temporal coupling characteristics between water level and flow rate. In the specific implementation, an analysis pair consisting of the "upstream monitoring point of the North Main Canal" and the "downstream monitoring point of the North Main Canal" was selected. Within a specified time window of 30 minutes, cross-correlation analysis was performed on the "water level time-series data from the upstream monitoring point of the North Main Canal" and the "flow time-series data from the downstream monitoring point of the North Main Canal" in the hydrological time-series data set. The calculated phase difference between the upstream water level change and the downstream flow rate change was 85 seconds, and the correlation coefficient was 0.92. A water level-flow rate joint change model was constructed, and the change in downstream flow rate caused by a unit change in upstream water level was calculated using least squares linear regression analysis as the static response coefficient. The calculation formula is as follows:

[0074]

[0075] in: This represents the change in downstream flow. This represents the calculated static response coefficient. This represents the change in upstream water level. The static response coefficient is calculated based on a set of historical data samples. The value is 5.6. It can be understood that analyzing the interaction between the high-frequency and low-frequency components of the "water level time series data from the upstream monitoring point of the North Main Canal" and the "flow time series data from the downstream monitoring point of the North Main Canal," and performing cross-spectral analysis on the high-frequency detail components after wavelet decomposition, allows for the extraction of delay coupling features reflecting the dynamic characteristics of water flow propagation. This delay coupling feature manifests as a frequency-dependent phase delay distribution in the flow response to water level changes during propagation. Together with the previously calculated 85-second phase difference, 0.92 correlation coefficient, and 5.6 static response coefficient, this constitutes the temporal coupling feature between water level and flow between the "upstream monitoring point of the North Main Canal" and the "downstream monitoring point of the North Main Canal."

[0076] In some embodiments, cross-source data correlation analysis is performed on the hydrological time series data set to extract the response characteristics between water level and gate opening. Taking the "main canal control gate" as an example, based on the "main canal gate upstream water level time series data" and "main canal control gate gate opening status data" on the upstream side of the control gate, a specific gate opening change event is analyzed. The opening status data linearly increases from 30% to 60%. After the event occurs, the analysis of the upstream "main canal gate upstream water level time series data" shows that the water level rises from 1.50 meters to 1.65 meters, an increase of 0.15 meters. The water level reaches its peak within 120 seconds and stabilizes between 1.64 meters and 1.66 meters in the following 300 seconds. The nonlinear mapping relationship within different gate opening intervals is calculated. For example, within the opening interval [30%, 50%], the average ratio between the rate of change in water level and the change in opening is approximately 0.005 meters per percentage opening, while within the opening interval [50%, 60%], the ratio decreases to 0.003 meters per percentage opening. The timeliness and intensity of the effect of gate opening adjustment on water level are quantified. The timeliness is reflected in the delay time from the start of the opening change to the start of a significant change in water level, which is approximately 20 seconds. The intensity is reflected in the steady-state change in water level (0.15 meters) caused by a unit change in opening (10%). This forms the response characteristics between water level and gate opening of the "main canal control gate", including a response amplitude of 0.15 meters, a response rate that reaches its peak within 120 seconds, a piecewise linear nonlinear mapping relationship, the timeliness of the 20-second delay, and the intensity of the effect of 0.015 meters per percentage opening. Optionally, for the water level downstream of the "main canal control gate," a similar analysis can be performed based on the "time series data of water level downstream of the main canal gate" to form the response characteristics of the gate opening to the downstream water level. It is understandable that for different gates in the irrigation area, their structural parameters and hydraulic conditions are different, and the specific numerical values ​​of the response characteristics between the water level and the gate opening obtained from the analysis will differ.

[0077] In one embodiment of the present invention, a dynamic hydrological evolution map is constructed, with each hydrological monitoring point within the irrigation district serving as a node in the dynamic hydrological evolution map. Each node contains the real-time water level and flow attributes of the monitoring point. Based on the actual topology of the irrigation district's water system, directed edges are established between monitoring points with upstream / downstream or water flow connectivity, with the direction of the directed edges representing the direction of water flow. Each directed edge is assigned attributes, including association feature values ​​extracted from a multi-dimensional association feature set related to the two monitoring points connected by the directed edge. These association feature values ​​are then normalized and used as relation weights. A pre-trained hydrological fusion model deployed on edge computing nodes is invoked, using the dynamic hydrological evolution map as input, to perform hydrological state fusion and estimation processing, generating a fused hydrological state map of the entire irrigation district. The dynamic hydrological evolution map is then input into the graph neural network encoder of the pre-trained hydrological fusion model to encode each node and its connected edges, generating an embedding vector representation of the node. In the graph propagation layer of the pre-trained hydrological fusion model, based on relation weights, the embedding vector representation of each node is updated through multiple rounds of information propagation and aggregation, enabling the state information of the nodes to be transmitted and fused along the graph structure. In the decoding output layer of the pre-trained hydrological fusion model, the updated embedding vector representation of each node is decoded into specific state values. These specific state values ​​include at least the estimated water level, estimated flow rate, and water level change trend. By accumulating the estimated flow rates on relevant paths, the estimated regional water consumption is obtained, ultimately forming the fused hydrological state map.

[0078] A dynamic hydrological evolution map was constructed, using four hydrological monitoring points in the irrigation district's "North Main Canal-East Branch Canal" area: the "Upstream Inlet Monitoring Point," the "North Main Canal Control Gate Monitoring Point," the "East Branch Canal Diversion Point Monitoring Point," and the "East Branch Canal Field Monitoring Point." The "Upstream Inlet Monitoring Point" node includes real-time water level attributes of "1.98 meters" and real-time flow attributes of "2.5 cubic meters per second," while the "North Main Canal Control Gate Monitoring Point" node includes real-time water level attributes of "1.65 meters" and real-time flow attributes of "2.3 cubic meters per second." Based on the actual topology of the irrigation district's water system, the "Upstream Inlet Monitoring Point" is located upstream of the "North Main Canal Control Gate Monitoring Point," with the water flow direction from upstream to downstream. Therefore, a directed edge is established between the "Upstream Inlet Monitoring Point" and the "North Main Canal Control Gate Monitoring Point," with the direction of the directed edge pointing from the "Upstream Inlet Monitoring Point" to the "North Main Canal Control Gate Monitoring Point." It is understandable that a directed edge is established between the "North Main Canal Sluice Gate Monitoring Point" and the "East Branch Canal Diversion Point Monitoring Point," with the direction pointing from the "North Main Canal Sluice Gate Monitoring Point" to the "East Branch Canal Diversion Point Monitoring Point." Each directed edge is assigned attributes. For the edge pointing from the "Upstream Inlet Monitoring Point" to the "North Main Canal Sluice Gate Monitoring Point," its attributes include correlation feature values ​​extracted from the multidimensional correlation feature set related to these two monitoring points. Examples include the phase difference "85 seconds" and correlation coefficient "0.92" in the water level-flow time-series coupling feature, and the response rate "120 seconds" in the water level-gate opening response feature. These correlation feature values ​​are normalized, and the calculated relation weight value "0.76" is used as the relation weight of this directed edge. In some embodiments, the normalization calculation formula for the relation weight is as follows:

[0079]

[0080] in: Represents the node Pointing to node The relation weights of directed edges; This represents the number of association features selected from the multidimensional association feature set; Representing the A related feature in the node With nodes The specific characteristic value between them has been min-max normalized to the [0,1] interval before being input into the formula; Representing the Preset weight coefficients for each associated feature, all The sum is 1. Based on the above formula, using three characteristics—phase difference, correlation coefficient, and response rate—and setting their values... The values ​​are 0.4, 0.4, and 0.2 respectively, and the calculations are as follows: It is 0.76.

[0081] The pre-trained hydrological fusion model deployed on edge computing nodes is invoked. A constructed dynamic hydrological evolution map, containing the aforementioned four nodes and three directed edges, is used as input for hydrological state fusion and estimation processing to generate a fused hydrological state map of the "North Main Canal - East Branch Canal" area of ​​the irrigation district. In specific implementation, the dynamic hydrological evolution map is input into the graph neural network encoder of the pre-trained hydrological fusion model. The encoder encodes each node in the map and its connected edges, transforming the node attributes (water level, flow rate) and the attributes of connected edges (relationship weights) into a high-dimensional numerical vector, generating an embedded vector representation of the "East Branch Canal Diversion Point Monitoring Point" node. In the graph propagation layer of the pre-trained hydrological data fusion model, based on the relation weights of directed edges, the embedding vector representation of each node is updated through multiple rounds of information propagation and aggregation. Taking the "East Branch Canal Field Monitoring Point" node as an example, its initial embedding vector representation will receive information from its only upstream node, "East Branch Canal Diversion Point Monitoring Point," along the directed edge in the first round of graph propagation. The amount of information received is modulated by the relation weights of the directed edge. Afterward, the "East Branch Canal Field Monitoring Point" node aggregates its own information with the received upstream information and updates its own embedding vector representation. In subsequent propagation rounds, information will continue to be propagated and fused along the graph structure, so that the state of downstream nodes can be integrated with the information of upstream nodes. In the decoding output layer of the pre-trained hydrological fusion model, the final embedded vector representation of each node after multiple rounds of updates is decoded into specific state values. For the "East Branch Canal Field Monitoring Point" node, the decoded values ​​are "0.45 meters", "0.12 cubic meters per second", and "slowly rising in the next 15 minutes". By accumulating the estimated flow values ​​of each side along the path from "Upstream Inlet Monitoring Point" to "East Branch Canal Field Monitoring Point", the estimated regional water consumption value of the "East Branch Canal" area is obtained as "150 cubic meters per hour". Finally, a fusion hydrological state map containing the estimated states of all nodes and the estimated regional water consumption value is formed.

[0082] See Figure 3This is a time-series monitoring data chart of multi-source water conditions in the irrigation area, visually displaying the changing trends of three core water condition indicators over 0-7 hours. The gate opening remains stable between 98% and 100%, with minimal overall fluctuation, indicating that the gates are in a stable, almost fully open operating state without significant adjustments. The flow rate is extremely low and almost unchanged, remaining stable at approximately 2-3 m³ / s, highly consistent with the stable gate opening. The water level also remains stable, close to 0 m, with no significant rise or fall, reflecting the current stable low-flow and low-water-level conditions in the irrigation area. The stable time-series data indicates that there are no sudden hydrological events in the irrigation area, and the data quality is good, suitable for direct use in feature extraction and state fusion. The low water level and low flow rate can be combined with subsequent correlation feature analysis to further assess the matching degree between the irrigation area's water supply capacity and crop water requirements.

[0083] In one embodiment of the present invention, based on the estimated regional water consumption and the water level change trends at each monitoring point in the integrated hydrological status map, and combined with the preset irrigation district water use plan, a water demand gap analysis and water allocation strategy generation process are performed to obtain a gate control suggestion scheme for the current scheduling cycle. The planned water consumption in different sub-irrigation districts within the current scheduling cycle is obtained from the preset irrigation district water use plan. The estimated regional water consumption of each sub-irrigation district calculated from the integrated hydrological status map is compared with the corresponding planned water consumption to calculate the water demand gap. The water level change trends at each monitoring point are analyzed to predict the water level trend in the short term and identify risk areas where the water level is below the warning value. Combining the water demand gap and water level risk prediction results, and under the condition of meeting the channel safety water level constraint, with the goal of minimizing the overall water demand gap, the suggested opening adjustment amount and adjustment timing of each gate are solved to form a gate control suggestion scheme. The gate control suggestion scheme is converted into a specific gate control command sequence, and the gate control command sequence is issued to the corresponding gate actuators in the irrigation district.

[0084] Based on the estimated regional water consumption and water level trends at various monitoring points in the integrated hydrological status map, and combined with the pre-set irrigation district water use plan, a water demand gap analysis and water allocation strategy generation process are performed to obtain a gate control proposal for the current scheduling cycle. In specific implementation, the current scheduling cycle is set as the next 2 hours. The planned water consumption for the three sub-irrigation districts of "East Branch Canal Irrigation District," "West Branch Canal Irrigation District," and "South Dou Canal Irrigation District" within the current scheduling cycle is obtained from the pre-set irrigation district water use plan database. The planned water consumption is 300 cubic meters, 450 cubic meters, and 280 cubic meters, respectively. From the newly generated integrated hydrological status map, the calculated estimated regional water consumption values ​​for "East Branch Canal Irrigation District," "West Branch Canal Irrigation District," and "South Dou Canal Irrigation District" are extracted. These three estimated values ​​are 270 cubic meters, 480 cubic meters, and 260 cubic meters, respectively. The estimated regional water consumption of each sub-irrigation area calculated from the integrated water situation map is compared with the corresponding planned water consumption to calculate the water demand gap. The formula for calculating the water demand gap is defined as follows:

[0085]

[0086] in: This represents the water demand shortfall. This represents the planned water consumption of a sub-irrigation district, obtained from the district's water use plan. This represents the estimated water consumption of the sub-irrigation area obtained from the integrated water situation map. The water demand gap for the "East Branch Canal Irrigation Area" is calculated based on this formula. +30 cubic meters (indicating a water shortage of 30 cubic meters), the water demand gap for the "West Branch Canal Irrigation Area" The water shortage in the "Nandouqu Irrigation District" is -30 cubic meters (indicating a water surplus of 30 cubic meters). +20 cubic meters (indicating a water shortage of 20 cubic meters). This is understandable; please refer to Table 1 for a comparison of the planned water consumption, estimated regional water consumption, and calculated water demand gap for each sub-irrigation area.

[0087] Table 1: Analysis of Water Demand Gap in Sub-Irrigation Districts during the Current Scheduling Cycle

[0088] Sub-irrigation district name Planned water consumption Estimated water consumption Water demand gap East Branch Canal Irrigation Area 300 270 +30 West Branch Canal Irrigation Area 450 480 -30 Nandouqu Irrigation District 280 260 +20

[0089] The water level change trends at each monitoring point in the integrated hydrological status map are analyzed. For example, the water level change trend at the "East Branch Canal Head Monitoring Point" shows a decrease of approximately 0.05 meters within the next 30 minutes, while the water level change trend at the "West Branch Canal Middle Section Monitoring Point" shows an increase of approximately 0.03 meters within the next 30 minutes. Based on these water level change trends, the short-term water level trend is predicted, and risk areas with water levels below the warning value are identified. The prediction model determines that the water level at the "East Branch Canal End Monitoring Point" may drop below the minimum operating water level of 0.40 meters within the next hour, thus identifying the "East Branch Canal Irrigation Area" as a water level risk area. In some embodiments, by comprehensively calculating the water demand gap and the water level risk prediction results, and under the condition of meeting the canal safety water level constraints, the recommended opening adjustment amount and adjustment timing of each gate are determined with the objective of minimizing the overall water demand gap. Specifically, the canal safety water level constraints include "the water level at the West Branch Canal Head must not exceed 1.80 meters" and "the water level at the East Branch Canal Head must not be lower than 0.45 meters." The optimization process is performed at edge computing nodes. The goal is to minimize the absolute value of the difference between the total water shortage in the "East Branch Canal Irrigation Area" and the "South Dou Canal Irrigation Area" and the total excess in the "West Branch Canal Irrigation Area," while preventing the "East Branch Canal Headwater Level" from falling below 0.45 meters. The solver calculates a scheme: lowering the suggested opening of the "West Branch Canal Diversion Gate" by 8%, increasing the suggested opening of the "East Branch Canal Intake Gate" by 5%, and slightly adjusting the suggested opening of the "South Main Canal Control Gate" by 2%. The adjustment timing is set to start execution after 10 minutes, forming a gate control suggestion scheme that includes the above gate, opening adjustment amount, and adjustment timing.

[0090] See Figure 4 This is a diagram analyzing the cross-source water situation correlation characteristics in an irrigation district, visually presenting the correlation coefficients and phase differences of three core cross-source water situation correlation characteristics. The water level-flow coupling characteristic shows a high correlation coefficient, indicating a strong linear positive correlation between water level and flow, consistent with the basic physical laws of open channel flow. The 15-minute phase difference represents a lag of approximately 15 minutes between flow changes and water level changes, reflecting the propagation delay of water flow in the channel. The water level-gate response characteristic shows a correlation coefficient of ~0.7, indicating that gate opening adjustments have a significant effect on water level control, but due to factors such as channel topology and flow inertia, the correlation is slightly lower than that of the water level-flow characteristic. The 8-minute phase difference is the shortest of the three characteristics, indicating that gate control responds fastest to water level, consistent with the direct control relationship between "gate and water level". The flow-soil moisture lag characteristic shows a correlation coefficient of ~0.75, indicating a moderate correlation between flow and soil moisture, reflecting the replenishment effect of irrigation water supply on soil moisture. The phase difference of 25 minutes is the longest among the three types of features.

[0091] In one embodiment of the present invention, the gate control suggestion scheme is converted into a specific gate control command sequence. The gate control suggestion scheme is parsed to extract the suggested opening adjustment amount, suggested adjustment start time, and suggested adjustment duration for each target gate. Based on the mechanical characteristics of the target gate actuator, the suggested opening adjustment amount is converted into the number of motor rotation pulses or valve stroke steps. The suggested adjustment start time, suggested adjustment duration, and the converted number of motor rotation pulses or valve stroke steps are encapsulated into a command data packet that can be directly recognized and executed by the gate controller according to a preset gate control command protocol. All command data packets are sorted according to the order of suggested adjustment start times to form a gate control command sequence. The suggested adjustment start times of different gates are checked for overlap or conflict. If a conflict exists, the suggested adjustment start times of some gates are fine-tuned according to the water flow propagation time and the priority of gate adjustment to ensure a smooth transition of the overall water flow. After fine-tuning, an ordered list of command data packets is regenerated according to the final adjustment time order as the final gate control command sequence.

[0092] The received gate control proposals are analyzed, and the suggested opening adjustment amount, suggested start time, and suggested duration for each target gate are extracted. For example, the proposed opening adjustment amount for the "West Branch Canal Diversion Gate" is "decreased by 8%", the suggested start time is "2026-03-23 ​​10:00:00", and the suggested duration is "300 seconds"; the proposed opening adjustment amount for the "East Branch Canal Intake Gate" is "increased by 5%", the suggested start time is "2026-03-23 ​​10:00:00", and the suggested duration is "180 seconds". Based on the mechanical characteristics of the target gate actuator, the suggested opening adjustment amount is converted into specific control parameters. For the "West Branch Canal Diversion Gate", whose actuator is a stepper motor, the edge computing node, based on the calibration relationship between the gate opening and the motor rotation angle, converts the "decreased by 8%" suggested opening adjustment amount into the specific number of pulses required for the motor to rotate. The formula for calculating the number of motor rotation pulses is as follows:

[0093]

[0094] in: This represents the calculated total number of pulses required by the motor. This represents the suggested opening adjustment amount extracted from the gate control proposal; here it is 8, and the unit is a percentage. This represents the change in gate opening per revolution of the motor. According to the mechanical calibration of the "West Branch Canal Diversion Gate", this value is 2, and the unit is a percentage per revolution. This represents the number of pulses required for one revolution of the motor; this motor requires 1600 pulses per revolution. Substituting this into the calculation yields... One pulse.

[0095] The suggested start time, suggested duration, and converted motor rotation pulse count or valve stroke steps are encapsulated into a command data packet that can be directly recognized and executed by the gate controller, according to a preset gate control command protocol. In specific implementation, the command data packet generated for the "West Branch Canal Diversion Gate" includes the following fields: command header "0xAA55", target gate address code "0x03", command type "absolute positioning mode", parameter field (including start time "2026-03-23 ​​10:00:00", duration "300 seconds", target pulse count "6400"), and cyclic redundancy check code. Optionally, for the "East Branch Canal Intake Gate", the command type in its command data packet is "speed mode", and the parameter field includes the target opening change rate. Following the suggestion to adjust the order of start times, all instruction data packets were sorted to initially form a gate control instruction sequence. In the initial sequence, the instruction data packets of "West Branch Canal Diversion Gate" and "East Branch Canal Intake Gate" were placed side by side because they had the same start time "2026-03-23 ​​10:00:00". Then came the instruction data packet of "South Main Canal Control Gate".

[0096] The edge computing node checked for overlaps or conflicts in the suggested adjustment start times of different gates. It detected a complete overlap in the execution times of the instructions for the "West Branch Canal Diversion Gate" and the "East Branch Canal Intake Gate," which could cause simultaneous changes in flow in both branches, resulting in hydraulic impact on the downstream main canal. Based on the water flow propagation time and gate adjustment priority, the suggested adjustment start times of some gates were fine-tuned. The time for water to flow from the "West Branch Canal Diversion Gate" to its confluence with the "East Branch Canal" is approximately 200 seconds, while the adjustment priority of the "East Branch Canal Intake Gate" is higher than that of the "West Branch Canal Diversion Gate" in the water supply security sequence. Understandably, to ensure a smooth transition of the overall water flow, the edge computing node fine-tuned the suggested adjustment start time of the "West Branch Canal Diversion Gate," delaying it by 200 seconds to "2026-03-23 ​​10:03:20," thus staggering the time it takes for the flow change to propagate to the confluence point from the impact of the "East Branch Canal Intake Gate" adjustment. After the fine-tuning is completed, the edge computing nodes regenerate an ordered list of instruction data packets according to the final adjustment time sequence. The first instruction data packet is "East Branch Canal Intake Gate" (start time 10:00:00), the second instruction data packet is "South Main Canal Control Gate", and the third instruction data packet is "West Branch Canal Diversion Gate" (start time 10:03:20). This list serves as the final sequence of gate control instructions issued.

[0097] See Figure 5This is a bar chart showing the grouped control commands for the irrigation district's gates, visually presenting the opening adjustments and durations of the three core gates. All gates are slightly closed, with adjustments ranging from 2% to 5%, representing refined control to avoid drastic impacts on the irrigation district's water flow. The West Branch Canal's diversion gate has the longest duration, corresponding to the largest opening adjustment, reflecting a "large adjustment, long duration" control strategy. The East Branch Canal's intake gate has the shortest duration, corresponding to the smallest opening adjustment, reflecting a "small adjustment, short duration" control strategy. The South Main Canal's control gate has a moderate duration, matching a moderate adjustment range. Overall, the gate regulation is primarily gentle and gradual, complying with the irrigation district's safe water conveyance constraints and preventing sudden changes in water level / flow rate from affecting downstream water use and canal safety. This provides foundational data for subsequent conflict detection and timing optimization.

[0098] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A method for real-time processing of multi-source irrigation district water situation data based on edge computing, characterized in that, The method includes: Based on the irrigation district water monitoring network, obtain a set of water situation time series data; Cross-source data correlation analysis was performed on the aforementioned hydrological time series data set to extract the temporal coupling characteristics between water level and flow rate, the response characteristics between water level and gate opening, and the hysteretic correlation characteristics between flow rate and soil moisture, thus forming a multidimensional correlation feature set. Based on the multidimensional association feature set, a dynamic hydrological evolution map is constructed. The dynamic hydrological evolution map uses nodes to represent hydrological monitoring points and directed edges to represent the water flow direction and association characteristics between monitoring points. Each directed edge is assigned a relation weight calculated from the multidimensional association feature set. The pre-trained hydrological fusion model deployed on the edge computing node is invoked, and the dynamic hydrological evolution map is used as input to perform hydrological status fusion and estimation processing to generate a fused hydrological status map of the entire irrigation area. The fused hydrological status map includes the estimated water level, estimated flow rate, water level change trend and estimated regional water consumption of each monitoring point.

2. The method for real-time processing of multi-source irrigation district water situation data based on edge computing according to claim 1, characterized in that, The set of time-series water situation data obtained based on the irrigation district water situation monitoring network includes: At each edge computing node in the irrigation district water monitoring network, multi-source water data from various heterogeneous sensors are collected and aggregated synchronously. The multi-source water data includes water level time series data, flow time series data, gate opening status data, and soil moisture time series data. The collected water level time series data, flow time series data, gate opening status data and soil moisture time series data are processed for data alignment and timestamp synchronization to generate a timestamp-aligned formatted multi-source hydrological data stream. Outlier detection and cleaning are performed on each time series data in the formatted multi-source hydrological data stream to remove outlier noise points and form a hydrological time series data set. The process of aligning and synchronizing the collected water level time-series data, flow rate time-series data, gate opening status data, and soil moisture time-series data to generate a time-stamp-aligned formatted multi-source hydrological data stream specifically includes: The raw time-series data streams of each sensor are timestamped using a unified time reference, which is provided by a network time protocol service deployed on edge computing nodes; Detect and complete the missing timestamps in each time-series data stream due to different sensor sampling frequencies or communication delays, forming a complete sequence with continuous timestamps; The complete sequence is resampled according to a preset high-frequency time grid to ensure that all types of data have corresponding values ​​at the same time point, thereby generating a formatted multi-source hydrological data stream with timestamp alignment.

3. The method for real-time processing of multi-source irrigation district water situation data based on edge computing according to claim 2, characterized in that, Outlier detection and cleaning are performed on each time-series data item in the formatted multi-source hydrological data stream, including: For each type of hydrological time-series data, the dynamic change threshold range is calculated based on its historical data; The real-time data value at the current moment is compared with the dynamic change threshold range to identify outlier points that exceed the dynamic change threshold range; For the identified outlier points, linear interpolation or spline interpolation is performed based on the normal data at the adjacent time points to replace the outlier points, thus completing the cleaning process.

4. The method for real-time processing of multi-source irrigation district water situation data based on edge computing according to claim 3, characterized in that, Cross-source data correlation analysis was performed on the aforementioned hydrological time series data set to extract the temporal coupling characteristics between water level and flow rate, including: Within a specified time window, cross-correlation analysis is performed on the water level time series data and flow time series data in the hydrological time series data set to calculate the phase difference and correlation coefficient between water level changes and flow changes. A water level-flow joint change model was constructed, and the flow change caused by a unit change in water level was calculated through regression analysis, which was used as the static response coefficient. The interaction between the water level time series data and the flow rate time series data in the high-frequency and low-frequency components is analyzed, and the delay coupling feature reflecting the dynamic characteristics of water flow propagation is extracted. Together with the phase difference, correlation coefficient and static response coefficient, it constitutes the time series coupling feature between the water level and the flow rate.

5. The method for real-time processing of multi-source irrigation district water situation data based on edge computing according to claim 4, characterized in that, Cross-source data correlation analysis was performed on the aforementioned hydrological time series data set to extract the response features between water level and gate opening, including: Based on the water level time series data at the location of the gate and the corresponding gate opening status data, analyze the amplitude, rate and stabilization time of the upstream and downstream water level changes after the gate opening change event occurs. Calculate the nonlinear mapping relationship between the rate of change of water level and the amount of change of gate opening within different gate opening ranges; The timeliness and intensity of the effect of gate opening adjustment on water level are quantified to form the response characteristics between water level and gate opening, which include response amplitude, response rate, nonlinear mapping relationship and timeliness.

6. The method for real-time processing of multi-source irrigation district water situation data based on edge computing according to claim 5, characterized in that, The construction of the dynamic hydrological evolution map includes: Each water level monitoring point within the irrigation area is used as a node in the dynamic water level evolution map, and each node contains the real-time water level and flow rate attributes of the water level monitoring point. Based on the actual topology of the irrigation area's water system, directed edges are established between monitoring points that have upstream and downstream or water flow connectivity, and the direction of the directed edges represents the direction of water flow. Each directed edge is assigned an attribute, which includes association feature values ​​extracted from the multidimensional association feature set and related to the two monitoring points connected by the directed edge. The association feature values ​​are then normalized and used as the relationship weight.

7. The method for real-time processing of multi-source irrigation district water situation data based on edge computing according to claim 6, characterized in that, The process involves calling a pre-trained hydrological fusion model deployed on edge computing nodes, taking the dynamic hydrological evolution map as input, performing hydrological state fusion and estimation processing, and generating a fused hydrological state map of the entire irrigation district, including: The dynamic hydrological evolution map is input into the graph neural network encoder of the pre-trained hydrological fusion model to encode each node and its connected edges, generating the node's embedding vector representation. In the graph propagation layer of the pre-trained hydrological fusion model, based on the relation weights, the embedding vector representation of each node is updated through multiple rounds of information propagation and aggregation, so that the state information of the nodes is transmitted and fused along the graph structure. In the decoding output layer of the pre-trained hydrological fusion model, the updated embedding vector representation of each node is decoded into specific state values. The specific state values ​​include at least the estimated water level, estimated flow rate, and water level change trend. By accumulating the estimated flow rate values ​​on the relevant paths, the estimated water consumption value of the region is obtained, and finally the fused hydrological state map is formed.

8. The method for real-time processing of multi-source irrigation district water situation data based on edge computing according to claim 7, characterized in that, Also includes: Based on the estimated regional water consumption and the water level change trend at each monitoring point in the integrated water situation map, and combined with the preset irrigation district water use plan, the water demand gap analysis and water allocation strategy generation process are performed to obtain the gate control suggestion scheme for the current scheduling cycle. The proposed gate control scheme is converted into a specific gate control command sequence, and the gate control command sequence is sent to the corresponding gate actuators in the irrigation area. Based on the estimated regional water consumption and water level change trends at each monitoring point in the integrated water situation map, and in conjunction with the preset irrigation district water use plan, a water demand gap analysis and water allocation strategy generation process are performed to obtain a gate control suggestion scheme for the current scheduling cycle, specifically including: From the preset irrigation district water use plan, obtain the planned water use in different sub-irrigation districts within the current scheduling cycle; The estimated regional water consumption of each sub-irrigation area calculated from the integrated water situation map is compared with the corresponding planned water consumption to calculate the water demand gap. Analyze the water level change trends at each monitoring point, predict the water level trend in the near future, and identify risk areas where the water level is below the warning value. Based on the combined results of the water demand gap and water level risk prediction, and under the condition of meeting the channel safety water level constraints, with the goal of minimizing the overall water demand gap, the recommended opening adjustment amount and timing of each gate are determined, thus forming the gate control recommendation scheme.

9. The method for real-time processing of multi-source irrigation data based on edge computing in irrigation districts according to claim 8, characterized in that, The proposed gate control scheme is converted into a specific sequence of gate control commands, including: The gate control suggestion scheme is analyzed to extract the suggested opening adjustment amount, suggested adjustment start time, and suggested adjustment duration for each target gate; Based on the mechanical characteristics of the target gate actuator, the suggested opening adjustment amount is converted into the number of motor rotation pulses or the number of valve stroke steps; The suggested adjustment start time, suggested adjustment duration, and converted motor rotation pulse count or valve stroke step count are encapsulated into an instruction data packet that can be directly recognized and executed by the gate controller, according to a preset gate control instruction protocol. Following the suggested adjustment of the start time order, all instruction data packets are sorted to form the gate control instruction sequence.

10. The method for real-time processing of multi-source irrigation district water situation data based on edge computing according to claim 9, characterized in that, The step of sorting all instruction data packets according to the suggested order of start times to form the gate control instruction sequence also includes: Check whether there is any overlap or conflict in the suggested adjustment start time of different gates. If there is a conflict, make a fine adjustment to the suggested adjustment start time of some gates according to the water flow propagation time and the priority of gate adjustment to ensure a smooth transition of the overall water flow. After fine-tuning is completed, an ordered list of instruction data packets is regenerated according to the final adjustment time sequence, which serves as the final sequence of gate control instructions to be issued.

Images