Irrigation district digital twinborn simulation method based on multi-source remote sensing data fusion

Abstract

The invention provides an irrigation district digital twinborn simulation method based on multi-source remote sensing data fusion. The method comprises the steps that multi-source remote sensing data are acquired and preprocessed; constructing a time sequence sample based on a sliding window, learning a spatio-temporal evolution law of the multi-source remote sensing data by using an LSTM model, and generating a daily scale continuous time sequence; fusing the multi-source remote sensing data based on a weighted fusion algorithm to generate a global multi-parameter data field; and constructing an irrigation area dynamic twinborn model based on the daily scale continuous time sequence, the global space continuous data and the basic geographic data of the irrigation area, and outputting a simulation result.

Description

Technical Field

[0001] This invention relates to the field of digital twin simulation technology, and more specifically, to a digital twin simulation method for irrigation districts based on multi-source remote sensing data fusion. Background Technology

[0002] Existing remote sensing technology, with its advantages of wide-area coverage and periodic observation, has become an indispensable means of macro-monitoring in irrigation areas. However, all types of remote sensing data have inherent limitations: while optical remote sensing data can intuitively reflect surface vegetation and water body information, it is easily affected by cloud and rain interference, leading to discontinuous observation sequences; although synthetic aperture radar (SAR) data has the ability to penetrate clouds and rain in all weather conditions, its spatial resolution is relatively limited, and the interpretation of surface parameters is subject to multiple interpretations; meteorological remote sensing data can provide large-scale meteorological driving fields, but the inversion accuracy and spatiotemporal resolution of key parameters (such as evapotranspiration) are still insufficient to meet the needs of refined management in irrigation areas. These shortcomings of single data sources restrict the in-depth application of remote sensing technology in irrigation decision-making, from monitoring to prediction and regulation.

[0003] Meanwhile, digital twin technology, as a cutting-edge paradigm for realizing dynamic interaction between the physical world and cyberspace, provides a revolutionary framework for the simulation, prediction, and optimization of the entire process of irrigation water resources. By constructing virtual models that couple water cycle, crop growth, and water conservancy project regulation, it is expected to achieve real-time mapping of irrigation district conditions and projection of future scenarios. However, the construction and operation of high-fidelity digital twins are highly dependent on high-quality, spatiotemporally consistent, and physically meaningful driving data. Currently, most research focuses either on improving inversion algorithms for single data sets or on developing mechanistic models for specific processes, failing to systematically address the core bottleneck of how to transform flawed multi-source remote sensing data into reliable, spatiotemporally continuous data products that can directly drive and correct multi-process digital twin models.

[0004] Therefore, existing technologies lack a technical chain that integrates the intelligent fusion and reconstruction of multi-source remote sensing data, the coupling and assimilation of multi-process mechanism models, and simulation and deduction for irrigation decision-making. This leads to digital twin models often facing the dilemma of lacking data or input data of low quality, making it difficult to achieve the leap from proof of concept to operational use in real irrigation district management, thus limiting their ability to solve practical water resource optimization scheduling and precision irrigation decision-making problems.

[0005] In order to solve the above problems, people have been seeking an ideal technological solution. Summary of the Invention

[0006] Therefore, it is necessary to provide a digital twin simulation method for irrigation districts based on multi-source remote sensing data fusion to address the aforementioned technical problems, combining the wide-area nature of multi-source remote sensing data with the dynamic nature of digital twins.

[0007] To achieve the above objectives, the first aspect of the present invention provides a digital twin simulation method for irrigation districts based on multi-source remote sensing data fusion, comprising the following steps: Acquire multi-source remote sensing data and auxiliary data, and perform spatiotemporal benchmark unification preprocessing; the multi-source remote sensing data includes optical remote sensing data, meteorological remote sensing data, synthetic aperture radar (SAR) data, interferometric synthetic aperture radar (InSAR) data, and auxiliary data including field monitoring data, basic geographic data attributes, and prior knowledge data. The on-site monitoring data includes soil moisture, channel flow, and water level data; The basic geographic data includes a digital elevation model (DEM) and vector maps of irrigation district boundaries and irrigation and drainage canal systems. The attribute and prior knowledge data include crop planting structure data; Based on the sliding window method, the Long Short-Term Memory (LSTM) model is used to reconstruct the daily-scale time series of the optical remote sensing data, meteorological remote sensing data, SAR radar data and the field monitoring data, respectively, to generate the corresponding daily-scale continuous time series. Based on the weighted fusion algorithm, the time series reconstructed in the previous step are fused together to generate a preliminary, spatiotemporally continuous global multi-parameter data field. The multi-parameter data field includes at least soil moisture and crop growth parameters obtained from optical and SAR data inversion. Based on the daily-scale continuous time series, the global multi-parameter data field, and the auxiliary data, a dynamic digital twin model of the irrigation district is constructed and run. The construction and operation of a dynamic digital twin model of an irrigation district specifically includes: Based on the aforementioned basic geographic data, a three-dimensional geographic base model including topography and digital irrigation and drainage facilities is constructed; On the aforementioned three-dimensional geographic base model, the following mechanistic sub-models are coupled and run: A distributed water cycle model is configured to simulate dynamic changes in soil moisture content; a crop growth and water requirement model is configured to simulate crop growth and predict water requirements based on crop growth parameters and meteorological data in the corrected data field. The water conservancy facility operation response model is configured to quantitatively simulate the correlation between water conservancy facility regulation and channel hydraulic state; By sharing soil moisture content, crop water requirements, and facility control commands as coupling variables, the dynamic two-way coupled simulation of the water cycle model, crop growth model, and facility operation model is realized, and the results of the irrigation area system state evolution are output.

[0008] The aforementioned scheme connects a deep learning model for temporal reconstruction with a weighted algorithm for spatial fusion, forming a collaborative processing core. This core specifically addresses the inherent limitations of remote sensing data such as optical, meteorological, and SAR data: it utilizes an LSTM model to learn long-term series patterns, intelligently filling in observational gaps caused by clouds and rain, and reconstructing low-frequency data into daily-scale series; subsequently, a weighted fusion algorithm integrates the spatial information advantages of data from different sources, suppresses noise, and outputs a spatially continuous standardized product. This step systematically solves the three major challenges of spatiotemporal resolution mismatch, observational gaps, and noise interference, directly producing a high-quality driving data field that can be directly used with the mechanistic model.

[0009] Subsequently, these highly processed, standardized data products, along with basic geographic data and attribute knowledge, were precisely injected into physical mechanism-based models of water cycle, crop growth, and facility operation. This process achieved precise alignment and efficient coupling of the data field and the model field at spatiotemporal scales, ensuring that the physical model operates under realistic and consistent driving forces.

[0010] The advantage of this approach is that by using a unified and intelligent upstream data preprocessing process, the simulation accuracy and reliability of complex downstream mechanism models are greatly improved. This completely frees up the limited computing power of the models from the heavy workload of data quality control, interpolation, and format conversion, allowing them to focus on the mechanism calculation itself. This provides an efficient, reliable, and scalable complete technical path for rapid, large-scale simulation of all elements of irrigation districts and multi-scenario decision analysis.

[0011] To achieve the above objectives, a second aspect of the present invention provides a digital twin simulation device for irrigation districts based on multi-source remote sensing data fusion, comprising: The preprocessing module is used to acquire multi-source remote sensing data and auxiliary data, and perform spatiotemporal benchmark unification preprocessing. The multi-source remote sensing data includes optical remote sensing data, meteorological remote sensing data, synthetic aperture radar (SAR) data, interferometric synthetic aperture radar (InSAR) data, and auxiliary data containing field monitoring data, basic geographic data attributes, and prior knowledge data. The field monitoring data includes soil moisture, canal flow, and water level data. The basic geographic data includes digital elevation model (DEM), irrigation district boundaries, and irrigation and drainage canal system vector maps. The attribute and prior knowledge data includes crop planting structure data. The time fusion module is used to perform daily-scale time series reconstruction on the optical remote sensing data, meteorological remote sensing data, SAR radar data and the field monitoring data based on the sliding window method and using the long short-term memory LSTM model, respectively, to generate corresponding daily-scale continuous time series sequences. The spatial fusion module is used to collaboratively fuse the time series obtained by the temporal fusion module based on a weighted fusion algorithm to generate a preliminary, spatiotemporally continuous global multi-parameter data field. The multi-parameter data field includes at least soil moisture and crop growth parameters obtained based on optical and SAR data inversion. The model building module is used to build and run a dynamic digital twin model of the irrigation district based on the daily-scale continuous time series, the global multi-parameter data field, and the auxiliary data. The construction and operation of the dynamic digital twin model of the irrigation district specifically includes: Based on the aforementioned basic geographic data, a three-dimensional geographic base model including topography and digital irrigation and drainage facilities is constructed; On the aforementioned three-dimensional geographic base model, the following mechanistic sub-models are coupled and run: A distributed water cycle model is configured to simulate dynamic changes in soil moisture content; a crop growth and water requirement model is configured to simulate crop growth and predict water requirements based on crop growth parameters and meteorological data in the corrected data field. The water conservancy facility operation response model is configured to quantitatively simulate the correlation between water conservancy facility regulation and channel hydraulic state; By sharing soil moisture content, crop water requirements, and facility control commands as coupling variables, the dynamic two-way coupled simulation of the water cycle model, crop growth model, and facility operation model is realized, and the results of the irrigation area system state evolution are output.

[0012] The aforementioned solution does not simply pile up data and models; instead, it innovatively links a deep learning model for time-series reconstruction with a weighted algorithm for spatial fusion, constructing a standardized data refining process. This systematically addresses the characteristics of multi-source remote sensing data (optical, meteorological, SAR, etc.) by resolving issues such as spatiotemporal resolution mismatch, missing observations, and noise interference, directly producing standardized, diurnal-scale, spatially continuous data usable by the model. Subsequently, these deeply processed high-dimensional data products, along with basic geographic and attribute knowledge, are precisely injected into physical mechanism-based models of water cycle, crop growth, and facility operation, achieving precise alignment and efficient coupling of the data field and model field at spatiotemporal scales. Furthermore, the solution, through unified and intelligent upstream data preprocessing, significantly improves the simulation accuracy and reliability of downstream complex mechanism models, freeing limited model computing power from arduous data quality control and interpolation work, allowing the model to focus on the mechanism calculation itself. This provides an efficient and scalable technical path for rapid, large-scale irrigation area state simulation and scenario analysis.

[0013] To achieve the above objectives, a third aspect of the present invention provides a method for generating irrigation strategies for irrigation districts based on multi-source remote sensing data fusion, characterized by comprising the following steps: Based on the irrigation district digital twin simulation method based on multi-source remote sensing data fusion described in the first aspect, the simulation results are output; Based on the user-defined irrigation strategy, output an irrigation water distribution plan.

[0014] The method proposed in this invention forms a complete technical closed loop with multi-source remote sensing data fusion, modeling, and application as its core architecture. At the data layer, it integrates heterogeneous data from multiple sources, including optical remote sensing, SAR radar, and meteorological remote sensing. Through a spatiotemporal collaborative fusion algorithm, it overcomes the limitations of single data sources, effectively improving data accuracy, continuity, and resistance to cloud and rain interference. At the application layer, through experimental verification and multi-dimensional benefit analysis, it forms a scalable smart irrigation district precision management solution, providing scientific decision-making support for water conservation and efficiency improvement. This technology chain systematically solves the complete technical gap between multi-source remote sensing data spatiotemporal fusion, twin modeling, and decision-making applications for the first time. It has significant methodological innovation and closely aligns with the actual needs of irrigation district management, demonstrating the dual necessity of theory and practice.

[0015] To achieve the above objectives, a fourth aspect of the present invention provides a computer device, including a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus. Memory, used to store computer programs; When the processor executes the program stored in the memory, it implements the irrigation district digital twin simulation method based on multi-source remote sensing data fusion as described in the first aspect.

[0016] To achieve the above objectives, a fifth aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the irrigation district digital twin simulation method based on multi-source remote sensing data fusion as described in the first aspect.

[0017] To achieve the above objectives, a sixth aspect of the present invention provides a computer program product, including a computer program that, when executed by a processor, implements the irrigation district digital twin simulation method based on multi-source remote sensing data fusion as described in the first aspect.

[0018] The beneficial effects of this invention are as follows: The method proposed in this invention connects a deep learning model for time-series reconstruction with a weighted algorithm for spatial fusion, constructing a standardized data refining process. This systematically addresses the characteristics of multi-source remote sensing data (optical, meteorological, SAR, etc.) by resolving issues such as spatiotemporal resolution mismatch, missing observations, and noise interference, directly producing standardized, daily-scale, spatially continuous data usable by the model. Subsequently, these deeply processed high-dimensional data products, along with basic geographic and attribute knowledge, are precisely injected into physical mechanism-based models of water cycle, crop growth, and facility operation, achieving precise alignment and efficient coupling of the data field and model field at the spatiotemporal scale. Furthermore, the above scheme, through unified and intelligent upstream data preprocessing, significantly improves the simulation accuracy and reliability of downstream complex mechanism models, freeing limited model computing power from arduous data quality control and interpolation work, allowing the focus to concentrate on the mechanism calculation itself. This provides an efficient and scalable technical path for rapid, large-scale irrigation area state simulation and scenario analysis. Finally, through experimental verification and multi-dimensional benefit analysis, a scalable smart irrigation area precision management solution is formed, providing scientific decision support for water conservation and efficiency improvement.

[0019] This technology chain is the first to systematically solve the complete technical gap between spatiotemporal fusion of multi-source remote sensing data, multi-process twin modeling, and decision application. It has significant innovation at the methodological level and closely meets the actual needs of irrigation district management, reflecting the dual necessity of theory and practice. Attached Figure Description

[0020] Figure 1 This is a flowchart illustrating the digital twin simulation method for irrigation districts based on multi-source remote sensing data fusion according to the present invention. Detailed Implementation

[0021] The technical solution of the present invention will be further described in detail below through specific embodiments.

[0022] To facilitate understanding, the interactive parties and / or terms and / or custom terms involved in this invention will first be explained in conjunction with the technical solution of this invention: InSAR data: Utilizing microwave synthetic aperture radar images, this method calculates the phase difference by interferometrically analyzing two images of the same area to obtain information on surface deformation. Data from the European Space Agency's Sentinel-1 radar is typically used.

[0023] Radar data: High-resolution images of the Earth's surface acquired using synthetic aperture radar technology.

[0024] Example 1 This embodiment provides a digital twin simulation method for irrigation districts based on multi-source remote sensing data fusion, including the following steps: Acquire and preprocess multi-source remote sensing data; the multi-source remote sensing data includes optical remote sensing data, meteorological remote sensing data, synthetic aperture radar (SAR) data, interferometric synthetic aperture radar (InSAR) data, and field auxiliary data containing field monitoring data, basic geographic data attributes, and prior knowledge data. The field monitoring data includes soil moisture, channel flow and water level data; among which, soil moisture refers to the soil volumetric water content or water potential measured at different depths (usually the root zone), which is directly assimilated into the subsequent coupled water cycle model, corrects the soil water state simulated by the model in real time, keeps the digital twin synchronized with reality, and serves as a trigger indicator for real-time irrigation. The channel flow refers to the water velocity / flow rate and water body elevation measured at key cross-sections of the channel (such as the headworks and branch outlets). It is used to calibrate and verify the hydraulic calculation formulas or parameters in the facility operation response model; as the upstream boundary conditions or real-time driving input for the water cycle model and facility model; and to calculate the actual water supply. It is the core of water management operations.

[0025] The basic geographic data includes vector maps of irrigation district boundaries and irrigation and drainage canal systems, and digital elevation models (DEMs); the vector map of the irrigation and drainage canal system distribution includes parameters of key equipment such as pumps and valves.

[0026] The digital elevation model is a dataset that records surface elevations in the form of a regular grid. It is the physical basis for calculating slope, aspect, and runoff path, and is also the unified base for simulating surface runoff in the distributed water cycle model. It is also the unified base for constructing three-dimensional visualization scenes and all spatial computing grids.

[0027] The irrigation and drainage canal distribution vector map defines the spatial polygon of the irrigation district management area, as well as the various levels of channels, drainage ditches and their topological connections represented by point and line elements; and simultaneously provides the parameters of various key equipment such as pumps and valves.

[0028] The attribute and prior knowledge data includes crop planting structure data. This crop planting structure data describes information such as "what to plant, on which type, and when to plant," and typically includes crop type (e.g., rice, corn), variety, spatial distribution map, and sowing / transplanting dates. It can clearly tell the crop growth model: on which grid, from what time, which crop simulation program to run, and also facilitates the development of differentiated irrigation regimes for different crops.

[0029] Specifically, the raw optical remote sensing data is Landsat data from the Gaofen series, such as Gaofen-6 Landsat-9 data, used for monitoring crop growth and water area; the SAR radar data is Sentinel-1 data, used to acquire soil moisture and topographic deformation; the InSAR data is used to monitor surface subsidence in irrigation areas; and the meteorological remote sensing data is MODIS data, used to acquire precipitation and temperature. Soil moisture is acquired using a soil moisture sensor, flow rate and water level are acquired using a flow meter, and crop planting structure data, such as crop cover, is acquired using a crop growth survey form.

[0030] Specifically, the raw optical remote sensing data is Gaofen series Landsat data, such as Gaofen-6 Landsat-9 data, used for monitoring crop growth and water area; the SAR radar data is Sentinel-1 data, used to obtain soil moisture and topographic deformation; the InSAR data is used to monitor surface subsidence in irrigation areas; and the meteorological remote sensing data is MODIS data, used to obtain precipitation and temperature.

[0031] It is understandable that after acquiring multi-source remote sensing data, it is necessary to preprocess the multi-source remote sensing data first.

[0032] Specifically, the FLAASH model was used to perform atmospheric correction on the optical data; Gamma software was used to perform radiometric calibration, topographic correction, and geocoding on the SAR data. Using optical remote sensing data as a reference, other data are registered to the WGS84 coordinate system, with the registration error controlled within 1 pixel; Based on the data clipping from the study area boundary, all data were uniformly resampled to a spatial resolution of 10m to ensure spatial consistency of the data. Median filtering was used to process optical remote sensing data to remove cloud and snow interference; Lee filtering was used to process SAR radar data to eliminate speckle noise. Soil moisture, flow rate, and water level are spatially interpolated to generate a continuous field for subsequent fusion.

[0033] After preprocessing, standardized multi-source remote sensing datasets and field auxiliary data are generated, which then proceed to the multi-source remote sensing data fusion stage: Temporal fusion: Based on the sliding window method, using the Long Short-Term Memory (LSTM) model, daily-scale time series reconstruction is performed on the optical remote sensing data, meteorological remote sensing data, SAR radar data, and the field monitoring data to generate corresponding daily-scale continuous time series sequences. For example, filling in the time interval gaps in optical remote sensing data, Landsat data that is updated every 16 days is supplemented into daily data, and dynamic datasets of key parameters of irrigation areas are generated daily.

[0034] Spatial fusion: Based on the weighted fusion algorithm, the time series reconstructed in the previous step are fused together to generate a preliminary, spatiotemporally continuous global multi-parameter data field. The multi-parameter data field includes at least soil moisture and crop growth parameters obtained from optical and SAR data inversion.

[0035] Specifically, the weights in the weighted fusion algorithm are obtained using the following formula: , in, ω i For the first i Weights of class data H i For the first i Information entropy of data is a class of data, where the smaller the entropy value, the higher the information content of the data and the greater the weight.

[0036] For example, optical data has low information entropy in crop growth monitoring, so its weight is set to 0.4; SAR radar data has the second lowest information entropy in soil moisture monitoring, so its weight is set to 0.3; meteorological data and field monitoring data have weights of 0.2 and 0.1, respectively. By weighted summation, a spatially continuous irrigation district parameter dataset (such as a 10m resolution soil moisture map) is generated.

[0037] Finally, based on the daily-scale continuous time series, the global spatial continuous data, and the auxiliary data, a dynamic digital twin model of the irrigation district is constructed and run.

[0038] Specifically, the field monitoring data includes soil moisture, channel flow, and water level data; The basic geographic data includes a digital elevation model (DEM) and vector maps of irrigation district boundaries and irrigation and drainage canal systems. The attribute and prior knowledge data include crop planting structure data; The construction and operation of a dynamic digital twin model of an irrigation district specifically includes: Based on the aforementioned basic geographic data, a three-dimensional geographic base model including topography and digital irrigation and drainage facilities is constructed; On the aforementioned three-dimensional geographic base model, the following mechanistic sub-models are coupled and run: A distributed water cycle model is configured to simulate dynamic changes in soil moisture content; a crop growth and water requirement model is configured to simulate crop growth and predict water requirements based on crop growth parameters and meteorological data in the corrected data field. The water conservancy facility operation response model is configured to quantitatively simulate the correlation between water conservancy facility regulation and channel hydraulic state; By sharing soil moisture content, crop water requirements, and facility control commands as coupling variables, the dynamic two-way coupled simulation of the water cycle model, crop growth model, and facility operation model is realized, and the results of the irrigation area system state evolution are output.

[0039] Specifically, the channel flow and water level changes calculated by the water conservancy facility operation response model according to the control instructions are input into the distributed water cycle model as new boundary conditions to drive the redistribution of soil moisture content.

[0040] In one embodiment, the crop growth model adopts the WOFOST model. The facility operation model simulates the relationship between gate opening and water flow rate, for example, a 50% opening corresponds to a flow rate of 5 m³ / s.

[0041] The crop growth and water requirement model is a mechanistic crop model, which uses data assimilation to take the time series of crop growth parameters in the corrected data field as observation values ​​and continuously corrects the growth state simulated by the model.

[0042] Furthermore, the linkage simulation supports two modes: real-time synchronization and scenario simulation. In scenario simulation mode, it can simulate and predict the spatiotemporal evolution of soil moisture and crop water shortage index in the future period based on the preset irrigation scheduling scheme.

[0043] Compared with traditional irrigation districts that rely on scattered on-site monitoring points, resulting in coverage blind spots and data incompatibility, this embodiment achieves 100% full coverage of the irrigation district by integrating multi-source remote sensing data such as optical, SAR, and meteorological data, combined with spatiotemporal collaborative algorithms. It can also generate daily continuous data with 10m resolution (such as soil moisture and crop growth), improving the data sensing range and timeliness by 3-5 times compared with traditional methods.

[0044] Compared with traditional irrigation districts that require the construction and maintenance of a large number of on-site monitoring points, this embodiment can replace 30-40% of the on-site monitoring points through remote sensing data coverage. At the same time, the twin model automatically outputs monitoring reports, reducing the workload of manual data analysis and lowering labor costs by 25-30%.

[0045] This embodiment does not depend on specific irrigation area terrain or crop type, and can be adapted to different landforms such as plains and hills, as well as irrigation areas for different crops such as rice, wheat, and corn; it has strong scalability and can be replicated to create large-scale value.

[0046] It should be understood that although the steps in the flowcharts of the above embodiments are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the above embodiments may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0047] Example 2 The difference between this embodiment and Embodiment 1 is that: The field monitoring data was divided into a fusion correction subset and an independent verification subset; Based on the weighted fusion algorithm, the optical remote sensing data, meteorological remote sensing data, and SAR data reconstructed in the previous step are spatially fused with the fusion correction subset to generate domain spatial continuous data. After generating the global multi-parameter data field, the generated global multi-parameter data field is further validated using the independent validation subset. The root mean square error (RMSE) between the generated global multi-parameter data field and the sensor's field monitoring data is calculated. If the RMSE meets a set threshold, a dynamic digital twin model of the irrigation area is constructed based on the daily continuous time series, the global spatial continuous data, and the auxiliary data, and is assimilated and run. If the RMSE does not meet the set threshold, the parameters of the LSTM model and / or the weighted fusion algorithm are iteratively corrected based on the validation error until a corrected global multi-parameter data field that meets the RMSE requirement is output.

[0048] It is understood that incremental training and Bayesian updates are used to iteratively optimize the parameters of the LSTM model and / or the weighted fusion algorithm.

[0049] Specifically, the division of the calibration subset and the independent validation subset is integrated, and a spatial stratified random sampling method is adopted to ensure that the two types of datasets are representative in terms of spatial distribution and statistical characteristics.

[0050] It is understandable that combining on-site monitoring data from irrigation areas (such as water level gauges and soil moisture sensors) can correct errors in remote sensing data and improve the reliability of fused data.

[0051] Taking soil moisture as an example, the root mean square error (RMSE) of the fused data and the field monitoring data from sensors is calculated as follows: .

[0052] It is understood that after the above data fusion, a high spatiotemporal resolution fusion dataset of the irrigation area is obtained. The high spatiotemporal resolution fusion dataset of the irrigation area is daily and 10m resolution, and specifically includes parameters such as soil moisture, crop coverage, and water area.

[0053] It is understandable that this embodiment, by embedding a portion of field data as calibration anchors into the fusion algorithm, fundamentally corrects the spatial systematic biases commonly found in remote sensing data, generating a preliminary product that is closer to the real world in both spatial pattern and absolute values. This step ensures that the data is constrained by high-precision standards during the production process. The verification using reserved, unprocessed field data is not simply a report on accuracy, but rather the ultimate test of the data product's generalization ability and robustness. Only when the verification metrics (such as RMSE, R²) meet the preset stringent thresholds are the data allowed to be input into the downstream digital twin model.

[0054] This closed-loop process has a dual effect: at the technical level, it produces a spatiotemporally continuous dataset with a clear accuracy certificate, whose error is quantified and controllable, greatly reducing the risk of simulation distortion caused by the uncertainty of input data; at the system level, it establishes a complete standard from data generation and quality control to qualified application, enabling the entire digital twin system to be built on a solid and reliable data foundation, and its output simulation and prediction results thus have higher decision-making reference value, realizing a fundamental leap from "having data available" to "having reliable data to be confident in".

[0055] Example 3 Based on the same inventive concept, this application also provides a device for implementing the above-mentioned method for digital twin simulation of irrigation areas based on multi-source remote sensing data fusion. The solution provided by this device is similar to the solution described in the above method. Therefore, the specific limitations of one or more embodiments of the device for digital twin simulation of irrigation areas based on multi-source remote sensing data fusion provided below can be found in the limitations of the method for digital twin simulation of irrigation areas based on multi-source remote sensing data fusion described above, and will not be repeated here.

[0056] To achieve the above objectives, a second aspect of the present invention provides a digital twin simulation device for irrigation districts based on multi-source remote sensing data fusion, comprising: The data acquisition module is used to acquire and preprocess optical remote sensing data, meteorological remote sensing data, SAR radar data, InSAR data, and field auxiliary data. The real-time auxiliary data includes field monitoring data, basic geographic data attributes, and prior knowledge data. The field monitoring data includes soil moisture, canal flow, and water level data. The basic geographic data includes irrigation district boundaries and irrigation / drainage canal vector maps, attribute parameters of key facilities, digital elevation models (DEMs), and soil type distribution maps. The attribute and prior knowledge data includes crop planting structure data. The time fusion module is used to construct sequence samples from preprocessed multi-source remote sensing data based on a sliding window, and to train the sequence samples using a long short-term memory (LSTM) model to reconstruct the daily-scale continuous time series of the optical remote sensing data, meteorological remote sensing data, SAR data, and InSAR data. The spatial fusion module is used to spatially fuse the optical remote sensing data, meteorological remote sensing data, and SAR data reconstructed in the previous step based on a weighted fusion algorithm to generate a global multi-parameter data field. The model building module is used to build and run a dynamic digital twin model of the irrigation district based on the daily-scale continuous time series, the global spatial continuous data, and the auxiliary data. The dynamic digital twin model of the irrigation district includes: Three-dimensional geographic base model; On the aforementioned three-dimensional geographic base model, the following mechanistic sub-models are coupled and run: A distributed water cycle model is configured to couple precipitation, evapotranspiration, infiltration and runoff processes and output gridded dynamic changes in soil moisture content. A crop growth and water requirement model is configured to simulate crop growth processes and predict water requirements at different growth stages based on crop growth parameters and meteorological data in the fused data. The water conservancy facility operation response model is configured to quantify the correlation between the control actions of facilities such as gates and pumping stations and the resulting channel flow and water level responses. By sharing the soil moisture content, crop water requirements, and facility control commands as coupling variables, the water cycle model, crop growth model, and facility operation model are linked for simulation, so as to output the overall state evolution of the irrigation area system.

[0057] It is understandable that through the real-time coupled computation of the above models, the system can output simulation results such as soil moisture forecasts, crop water shortage status, and channel water flow evolution under different future scenarios, thereby supporting quantitative decision-making for precise irrigation scheduling, drought early warning, and optimal allocation of water resources.

[0058] Example 4 Based on the same inventive concept, this application also provides a method for generating irrigation strategies for irrigation districts based on multi-source remote sensing data fusion, including the following steps: Based on the digital twin simulation method for irrigation districts based on multi-source remote sensing data fusion described in Example 1, the simulation results are output. Based on the user-defined irrigation strategy, output an irrigation water distribution plan.

[0059] Example 5 To achieve the above objectives, this embodiment provides a computer device, including a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus; Memory, used to store computer programs; When the processor executes the program stored in the memory, it implements the irrigation district digital twin simulation method based on multi-source remote sensing data fusion as described in Embodiment 1 or 2.

[0060] Example 6 To achieve the above objectives, this embodiment provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the irrigation district digital twin simulation method based on multi-source remote sensing data fusion as described in Embodiment 1 or 2.

[0061] Example 7 To achieve the above objectives, this embodiment provides a computer program product, including a computer program that, when executed by a processor, implements the irrigation district digital twin simulation method based on multi-source remote sensing data fusion as described in Embodiment 1 or 2.

[0062] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them; although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications can still be made to the specific implementation of the present invention or equivalent substitutions can be made to some technical features without departing from the spirit of the technical solutions of the present invention, and all such modifications and substitutions should be covered within the scope of the technical solutions claimed in the present invention.

Claims

1. A digital twin simulation method for irrigation districts based on multi-source remote sensing data fusion, characterized in that, Includes the following steps: Acquire multi-source remote sensing data and auxiliary data, and perform spatiotemporal benchmark unification preprocessing; the multi-source remote sensing data includes optical remote sensing data, meteorological remote sensing data, synthetic aperture radar (SAR) data, interferometric synthetic aperture radar (InSAR) data, and auxiliary data including field monitoring data, basic geographic data attributes, and prior knowledge data. The on-site monitoring data includes soil moisture, channel flow, and water level data; The basic geographic data includes a digital elevation model (DEM) and vector maps of irrigation district boundaries and irrigation and drainage canal systems. The attribute and prior knowledge data include crop planting structure data; Based on the sliding window method, the Long Short-Term Memory (LSTM) model is used to reconstruct the daily-scale time series of the optical remote sensing data, meteorological remote sensing data, SAR radar data and the field monitoring data, respectively, to generate the corresponding daily-scale continuous time series. Based on the weighted fusion algorithm, the time series reconstructed in the previous step are fused together to generate a preliminary, spatiotemporally continuous global multi-parameter data field. The multi-parameter data field includes at least soil moisture and crop growth parameters obtained from optical and SAR data inversion. Based on the daily-scale continuous time series, the global multi-parameter data field, and the auxiliary data, a dynamic digital twin model of the irrigation district is constructed and run. The construction and operation of a dynamic digital twin model of an irrigation district specifically includes: Based on the aforementioned basic geographic data, a three-dimensional geographic base model including topography and digital irrigation and drainage facilities is constructed; On the aforementioned three-dimensional geographic base model, the following mechanistic sub-models are coupled and run: A distributed water cycle model is configured to simulate dynamic changes in soil moisture content; a crop growth and water requirement model is configured to simulate crop growth and predict water requirements based on crop growth parameters and meteorological data in the corrected data field. The water conservancy facility operation response model is configured to quantitatively simulate the correlation between water conservancy facility regulation and channel hydraulic state; By sharing soil moisture content, crop water requirements, and facility control commands as coupling variables, the dynamic two-way coupled simulation of the water cycle model, crop growth model, and facility operation model is realized, and the results of the irrigation area system state evolution are output.

2. The method for digital twin simulation of irrigation districts based on multi-source remote sensing data fusion according to claim 1, characterized in that, The field monitoring data was divided into a fusion correction subset and an independent verification subset; Based on the weighted fusion algorithm, the optical remote sensing data, meteorological remote sensing data, SAR data reconstructed in the previous step are spatially fused with the fusion correction subset to generate a global multi-parameter data field. After generating the global multi-parameter data field, the generated global multi-parameter data field is further validated using the independent validation subset. The root mean square error (RMSE) between the generated global multi-parameter data field and the sensor's field monitoring data is calculated. If the RMSE meets a set threshold, a dynamic digital twin model of the irrigation area is constructed based on the daily continuous time series, the global spatial continuous data, and the auxiliary data, and is assimilated and run. If the RMSE does not meet the set threshold, the parameters of the LSTM model and / or the weighted fusion algorithm are iteratively corrected based on the validation error until a corrected global multi-parameter data field that meets the RMSE requirement is output.

3. The method for digital twin simulation of irrigation districts based on multi-source remote sensing data fusion according to claim 2, characterized in that, The fusion weights in the weighted fusion algorithm are obtained using the following formula: ， in, ω i For the first i Weights of class data H i For the first i Information entropy of class data.

4. The method for digital twin simulation of irrigation districts based on multi-source remote sensing data fusion according to claim 3, characterized in that, Acquire optical remote sensing data, meteorological remote sensing data, SAR radar data, InSAR data, and field auxiliary data, and perform preprocessing, including: Acquire raw optical remote sensing data from the Landsat-9 satellite of the Gaofen-6 remote sensing satellite; Acquire IW mode SAR data from Sentinel-1 satellite; Acquire InSAR data from the ALOS-2 radar satellite; Acquire meteorological remote sensing data collected by MODIS sensors; Soil moisture was obtained using a moisture sensor, flow rate and water level were obtained using a flow meter, and crop coverage was obtained using a crop growth survey form. The FLAASH model was used to perform atmospheric correction on the optical data; Gamma software was used to perform radiometric calibration, topographic correction, and geocoding on the SAR data. Using optical remote sensing data as a reference, other data are registered to the WGS84 coordinate system, with the registration error controlled within 1 pixel; Based on the data clipping from the study area boundary, all data were uniformly resampled to a spatial resolution of 10m to ensure spatial consistency of the data. Median filtering was used to process optical remote sensing data to remove cloud and snow interference; Lee filtering was used to process SAR radar data to eliminate speckle noise. Soil moisture, flow rate, and water level are spatially interpolated to generate a continuous field for subsequent fusion and verification.

5. A digital twin simulation device for irrigation districts based on multi-source remote sensing data fusion, characterized in that: The preprocessing module is used to acquire multi-source remote sensing data and auxiliary data, and perform spatiotemporal benchmark unification preprocessing. The multi-source remote sensing data includes optical remote sensing data, meteorological remote sensing data, synthetic aperture radar (SAR) data, interferometric synthetic aperture radar (InSAR) data, and auxiliary data containing field monitoring data, basic geographic data attributes, and prior knowledge data. The field monitoring data includes soil moisture, canal flow, and water level data. The basic geographic data includes digital elevation model (DEM), irrigation district boundaries, and irrigation and drainage canal system vector maps. The attribute and prior knowledge data includes crop planting structure data. The time fusion module is used to perform daily-scale time series reconstruction on the optical remote sensing data, meteorological remote sensing data, SAR radar data and the field monitoring data based on the sliding window method and using the long short-term memory LSTM model, respectively, to generate corresponding daily-scale continuous time series sequences. The spatial fusion module is used to collaboratively fuse the time series obtained by the temporal fusion module based on a weighted fusion algorithm to generate a preliminary, spatiotemporally continuous global multi-parameter data field. The multi-parameter data field includes at least soil moisture and crop growth parameters obtained based on optical and SAR data inversion. The model building module is used to build and run a dynamic digital twin model of the irrigation district based on the daily-scale continuous time series, the global multi-parameter data field, and the auxiliary data. The construction and operation of the dynamic digital twin model of the irrigation district specifically includes: Based on the aforementioned basic geographic data, a three-dimensional geographic base model including topography and digital irrigation and drainage facilities is constructed; On the aforementioned three-dimensional geographic base model, the following mechanistic sub-models are coupled and run: A distributed water cycle model is configured to simulate dynamic changes in soil moisture content; a crop growth and water requirement model is configured to simulate crop growth and predict water requirements based on crop growth parameters and meteorological data in the corrected data field. The water conservancy facility operation response model is configured to quantitatively simulate the correlation between water conservancy facility regulation and channel hydraulic state; By sharing soil moisture content, crop water requirements, and facility control commands as coupling variables, the dynamic two-way coupled simulation of the water cycle model, crop growth model, and facility operation model is realized, and the results of the irrigation area system state evolution are output.

6. A method for generating irrigation strategies for irrigation districts based on multi-source remote sensing data fusion, characterized in that, Includes the following steps: Based on the digital twin simulation method for irrigation districts based on multi-source remote sensing data fusion as described in any one of claims 1-4, output simulation results; Based on the user-defined irrigation strategy, output an irrigation water distribution plan.

7. A computer device, characterized in that: It includes a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus; Memory, used to store computer programs; The processor, when executing a program stored in memory, implements the digital twin simulation method for irrigation districts based on multi-source remote sensing data fusion as described in any one of claims 1 to 4.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the irrigation district digital twin simulation method based on multi-source remote sensing data fusion as described in any one of claims 1 to 4.

9. A computer program product, comprising a computer program, characterized in that, When executed by a processor, the computer program implements the digital twin simulation method for irrigation districts based on multi-source remote sensing data fusion as described in any one of claims 1 to 4.

Images