Water conservancy digital twin data fusion method, system and equipment based on multi-source heterogeneous data dynamic alignment and medium

The water conservancy digital twin data fusion method, which uses dynamic alignment of multi-source heterogeneous data, solves the problems of low data fusion accuracy and poor real-time performance in water conservancy information systems, achieves high-precision and stable data fusion results, and improves the system's dynamic adaptability and intelligence level.

CN121479632APending Publication Date: 2026-02-06NANJING HEHAI NANZI HYDROPOWER AUTOMATION
View PDF 0 Cites 1 Cited by

Patent Information

Application Number
CN202511312929.2
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2025-09-15
Publication Date
2026-02-06

AI Technical Summary

Technical Problem

In existing water conservancy information systems, the integration of multi-source heterogeneous data faces challenges in terms of dynamic adaptability and intelligence, resulting in low data integration accuracy, poor real-time performance, difficulty in tracing the source, low value, high risk, and difficulty in integration.

Method used

A water conservancy digital twin data fusion method with dynamic alignment of multi-source heterogeneous data is adopted. Through data cleaning, format standardization, timestamp alignment, sliding window partitioning and spatial grid numbering, combined with dynamic weighted averaging and Kalman filtering, the data is fused under a unified spatiotemporal framework.

Benefits of technology

When faced with complex and ever-changing water conservancy scenarios, it ensures high accuracy, stability and reliability of data, improves the real-time performance and intelligence level of data, and solves the alignment problem caused by differences in data source sampling frequency and irregular spatial distribution.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN121479632A_ABST
    Figure CN121479632A_ABST
Patent Text Reader

Abstract

The invention discloses a water conservancy digital twin data fusion method, system and device based on multi-source heterogeneous data dynamic alignment and a medium, and belongs to the technical field of data fusion. Performing cleaning processing, format standardization processing and timestamp alignment processing based on a time window on the data to obtain preprocessed data; performing sliding window division processing on the preprocessed data, performing regional grid numbering processing according to a space division rule, and enabling each data point to correspond to a divided time window and a space grid to obtain time-space mapping data; and performing fusion processing on the data subjected to space-time mapping according to the data type. According to the method, the fusion strategy is designed by distinguishing the data types in the fusion stage, and it is guaranteed that the fusion result can still keep high reliability and stability when facing data of different qualities and different sources.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of data fusion technology, specifically to a method, system, device, and medium for water conservancy digital twin data fusion based on dynamic alignment of multi-source heterogeneous data. Background Technology

[0002] In existing water conservancy information systems, the fusion of multi-source heterogeneous data (such as sensor data, satellite remote sensing data, and meteorological data) faces numerous challenges. Traditional methods typically employ static alignment, which cannot adapt to real-time changing data environments, resulting in low data fusion accuracy and poor real-time performance. Furthermore, problems such as difficulty in data traceability, low value, high risk, integration difficulties, and poor quality are also prevalent. For example, while existing data fusion platforms achieve end-to-end data exchange and transaction processing mechanisms to ensure data consistency, their dynamic adaptability and intelligence level still need improvement when facing complex and ever-changing water conservancy scenarios. Summary of the Invention

[0003] In view of the above-mentioned problems, the present invention is proposed.

[0004] Therefore, the technical problem solved by this invention is: how to address the issue that although existing methods achieve end-to-end data exchange and transaction processing mechanisms to ensure data consistency, their dynamic adaptability and intelligence level still need to be improved when facing complex and ever-changing water conservancy scenarios.

[0005] To address the aforementioned technical problems, this invention provides the following technical solution: a water conservancy digital twin data fusion method based on dynamic alignment of multi-source heterogeneous data, comprising: collecting water conservancy data from different sensor devices, remote sensing platforms, and meteorological models; performing data cleaning, format standardization, and timestamp alignment based on time windows to obtain preprocessed data; performing sliding window partitioning on the preprocessed data and performing regional grid numbering according to spatial partitioning rules, mapping each data point to the partitioned time window and spatial grid to obtain spatiotemporally mapped data; and performing fusion processing on the spatiotemporally mapped data according to data type, wherein real-time monitoring data and historical data are subjected to weighted average processing based on dynamically set weights, and predicted data are subjected to Kalman filtering processing based on dynamically calculated Kalman gain.

[0006] As a preferred embodiment of the water conservancy digital twin data fusion method based on dynamic alignment of multi-source heterogeneous data described in this invention, the steps of performing data cleaning, format standardization, and timestamp alignment based on time windows include: importing multi-source data from sensor devices, remote sensing platforms, and meteorological models into a database; deleting data records with null values ​​using data processing tools; performing a unified format conversion on information fields representing time, ensuring that all data records use the same time expression method; and performing window alignment processing on timestamps from different sources according to a set time window strategy to generate a data set with a unified time benchmark.

[0007] As a preferred embodiment of the water conservancy digital twin data fusion method based on dynamic alignment of multi-source heterogeneous data described in this invention, the step of performing sliding window partitioning on the preprocessed data and performing regional grid numbering according to spatial partitioning rules includes: performing continuous window partitioning on the data time axis based on a set time step and identifying the time period corresponding to each data record; generating multiple spatial units for the coverage area according to the set spatial partitioning rules and assigning a uniquely identifiable number to each spatial unit; extracting spatial location information for each data record and associating it with the corresponding numbered spatial unit; and classifying the data records according to the time period and spatial number to form a standardized spatiotemporal index structure.

[0008] As a preferred embodiment of the water conservancy digital twin data fusion method based on dynamic alignment of multi-source heterogeneous data described in this invention, the following steps are performed: For the data that has completed spatiotemporal mapping, fusion processing is performed according to the data type. Specifically, a weighted average processing based on dynamically set weights is performed on real-time monitoring data and historical data, and a Kalman filter processing based on dynamically calculated Kalman gain is performed on the predicted data. This includes: setting the data fusion weights required for weighted averaging according to the type and attributes of the data source, and allocating the required data fusion weights to real-time monitoring data and historical data; performing a weighted average operation on monitoring data from different sources within the same time period and spatial unit; obtaining the difference between the state estimate and the observed value for the predicted data, and calculating the Kalman gain required for fusion; updating the predicted state value based on the calculated Kalman gain, thus completing the fusion processing of the predicted data.

[0009] As a preferred embodiment of the water conservancy digital twin data fusion method based on dynamic alignment of multi-source heterogeneous data described in this invention, the step of performing window alignment processing on the timestamps of data from different sources includes: setting sliding time window parameters and determining the window granularity based on real-time requirements; obtaining the timestamps of the original records in each data source and aligning them to the reference time point of the window according to their window positions; and aggregating records from multiple data sources within the same window and uniformly labeling them as the window's timestamps.

[0010] This preferred solution sets a sliding time window and aggregates and labels the timestamps of data from different sources, enabling multi-source data to be uniformly identified within the same time window. This ensures the alignment of data records in the time dimension even when data frequencies are inconsistent, avoiding data fusion errors caused by time misalignment.

[0011] As a preferred embodiment of the water conservancy digital twin data fusion method based on dynamic alignment of multi-source heterogeneous data described in this invention, the step of extracting spatial location information from each data record and associating it with a corresponding numbered spatial unit includes: reading the latitude and longitude coordinate fields contained in the data record and obtaining the grid number according to a preset spatial grid rule; the spatial grid rule is an equidistant rectangular division rule constructed based on a two-dimensional geographic coordinate system, ensuring that the grid number is unique and indexable within the entire region; and writing the obtained grid number as a spatial location label into the original data record to form a binding relationship between data and space.

[0012] This preferred solution establishes a binding relationship between data and spatial units by reading the latitude and longitude coordinates in the data records and generating unique numbers according to the grid rules of the two-dimensional coordinate system. This enables data from different sources to be integrated within a unified spatial framework, solving the problem of data indexing difficulties caused by inconsistent spatial division in traditional methods.

[0013] As a preferred embodiment of the water conservancy digital twin data fusion method based on dynamic alignment of multi-source heterogeneous data described in this invention, the step of obtaining the difference between the state estimate and the observed value of the predicted data and calculating the Kalman gain required for fusion includes: before performing fusion processing, using the predicted state estimate result of the previous period as the current initial prediction value; collecting sensor observation data corresponding to the predicted state and performing mean processing on the observation data within a time window to reduce the impact of single-point fluctuations on the estimation process; constructing a weight factor group for calculating the Kalman gain based on the difference between the predicted value and the observed value, combining the current state covariance and the observation error variance; dynamically calculating the Kalman gain and using the Kalman gain to correct the predicted state, forming a fused state value adjusted by the observed data, which serves as the state estimation input for the next period.

[0014] This preferred scheme suppresses single-point observation fluctuations by introducing mean processing within a time window before fusion, and simultaneously considers the prediction covariance and observation error variance when calculating the Kalman gain, so that the predicted state can be dynamically corrected during the fusion process. Thus, even when the uncertainty of multi-source prediction data is large, it can still generate stable and reliable state estimation results.

[0015] This invention provides a water conservancy digital twin data fusion system based on dynamic alignment of multi-source heterogeneous data.

[0016] To address the aforementioned technical problems, this invention provides the following technical solution: a water conservancy digital twin data fusion system based on dynamic alignment of multi-source heterogeneous data, comprising: a preprocessing module, a dynamic alignment module, and a data fusion module; the preprocessing module is used to collect data from different sensor devices, remote sensing platforms, and meteorological models, and to perform data cleaning, format standardization, and timestamp alignment based on time windows to obtain preprocessed data; the dynamic alignment module is used to perform sliding window partitioning on the preprocessed data, and to perform regional grid numbering according to spatial partitioning rules, mapping each data point to the partitioned time window and spatial grid to obtain data with completed spatiotemporal mapping; the data fusion module is used to perform fusion processing on the spatiotemporally mapped data according to data type, wherein real-time monitoring data and historical data are subjected to weighted average processing based on dynamically set weights, and predicted data are subjected to Kalman filtering processing based on dynamically calculated Kalman gain.

[0017] The present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, characterized in that the processor executes the computer program to implement the steps of the water conservancy digital twin data fusion method based on dynamic alignment of multi-source heterogeneous data.

[0018] The present invention provides a computer-readable storage medium having a computer program stored thereon, characterized in that, when the computer program is executed by a processor, it implements the steps of the water conservancy digital twin data fusion method based on dynamic alignment of multi-source heterogeneous data.

[0019] The beneficial effects of this invention are as follows: By performing data cleaning, unifying time formats, and aligning timestamps during the preprocessing stage, this invention ensures the consistency of multi-source data from sensors, remote sensing platforms, and meteorological models in the time dimension; by introducing a sliding time window and spatial grid partitioning mechanism during the alignment stage, it achieves the mapping of heterogeneous data at a unified time and spatial scale, fundamentally solving the alignment problem caused by differences in data source sampling frequencies and irregular spatial distribution; by designing fusion strategies that differentiate data types during the fusion stage, using dynamic weighted averaging for real-time and historical data, and employing filtering processing that combines uncertainty-based dynamic calculation of Kalman gain for predicted data, this invention ensures that the fusion results maintain high reliability and stability when facing data of different quality and from different sources. Attached Figure Description

[0020] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 The present invention provides an overall flowchart of a water conservancy digital twin data fusion method based on dynamic alignment of multi-source heterogeneous data, which is an embodiment of the present invention.

[0022] Figure 2 The flowchart illustrates a water conservancy digital twin data fusion method based on dynamic alignment of multi-source heterogeneous data, as provided in one embodiment of the present invention.

[0023] Figure 3 An adaptive fusion strategy diagram of a water conservancy digital twin data fusion method based on dynamic alignment of multi-source heterogeneous data is provided as an embodiment of the present invention. Detailed Implementation

[0024] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0025] Example 1, referring to Figure 1 This is one embodiment of the present invention, which provides a method for fusing water conservancy digital twin data based on dynamic alignment of multi-source heterogeneous data, including:

[0026] S1. Collect water conservancy data from different sensor devices, remote sensing platforms and meteorological models, and perform data cleaning, format standardization and time window-based timestamp alignment to obtain preprocessed data.

[0027] S2. Perform sliding window partitioning on the preprocessed data and perform regional grid numbering according to the spatial partitioning rules, mapping each data point to the partitioned time window and spatial grid to obtain the data with completed spatiotemporal mapping.

[0028] S3. For the data that has completed the spatiotemporal mapping, perform fusion processing according to the data type. Specifically, perform weighted averaging processing based on dynamically set weights on real-time monitoring data and historical data, and perform Kalman filtering processing based on dynamically calculated Kalman gain on predicted data.

[0029] It should be noted that in the operation of water conservancy systems, measured data from sensors, remote sensing data, and meteorological model prediction data are characterized by inconsistent sampling frequencies, irregular spatial distribution, and significant differences in accuracy. Direct fusion often leads to data misalignment and estimation bias. Through steps S1-S3, a unified time window can be established in the time dimension, a unified grid coordinate system can be constructed in the spatial dimension, and dynamic weights and Kalman filtering strategies can be introduced in the fusion process. This enables the fusion of multi-source heterogeneous data within a unified spatiotemporal framework, providing a reliable data foundation for the construction of water conservancy digital twin models.

[0030] Example 2, refer to Figure 2 and Figure 3 As an embodiment of the present invention, based on the previous embodiment, a method for fusing water conservancy digital twin data based on dynamic alignment of multi-source heterogeneous data is provided, comprising:

[0031] In this embodiment of the application, the timestamp alignment process based on the time window in step S1 can set the sliding time window parameter as the time alignment benchmark, obtain the original timestamps from the data records of different sensor devices, remote sensing platforms and meteorological models, align the timestamps to the reference time point of the corresponding window according to their window position, and uniformly label the data in the same time window as the time label of that window, thereby forming a time-consistent data set.

[0032] In one alternative implementation, the timestamp alignment process based on time windows can also divide the time axis into static equal-length time periods, such as dividing time intervals into fixed intervals at the minute or hour level, and directly merge the timestamps of all data records to the starting time point of their respective time periods for standardization and consolidation.

[0033] In another alternative implementation, the timestamp alignment processing based on time windows can also use key hydrological events as the basis for dividing time windows. After detecting specific events such as sudden changes in water level or rainfall exceeding the threshold, time windows are dynamically generated, and all data records within the relevant time period are uniformly time-marked.

[0034] This invention achieves consistent calibration of data from multiple sources, heterogeneous data, and data with different sampling frequencies in the time dimension through a sliding time window alignment method, avoiding data fusion deviations caused by time misalignment and establishing a unified time benchmark system for subsequent sliding window analysis and fusion calculation.

[0035] Furthermore, in step S1, water conservancy data from different sensor devices, remote sensing platforms, and meteorological models are collected, and the data is cleaned, standardized, and aligned with timestamps based on time windows to obtain preprocessed data, including the following steps A1-A3:

[0036] A1. Import multi-source data from sensor devices, remote sensing platforms, and meteorological models into the database, and use data processing tools to delete data records with null values.

[0037] A2. Perform a uniform format conversion on the information fields representing time, and use the same time expression method for all data records.

[0038] A3. Based on the set time window strategy, perform window alignment processing on the timestamps of data from different sources to generate a data set with a unified time base.

[0039] In the embodiments of this application, the import of multi-source data into the database in step A1 can be achieved by using ETL tools to extract raw data from sensor devices, remote sensing platforms and meteorological models, and writing various types of data into a unified data management platform according to a preset field structure. A relational database system is used to organize and store the data, and a field-level metadata index is established to ensure that the multi-source data has a consistent processing foundation in terms of structure, format and time series.

[0040] In one alternative implementation, importing multi-source data into the database can also be achieved by writing data acquisition scripts to connect to various data interfaces in real time, and using APIs or data push mechanisms to directly write the acquired data streams into a high-performance time series database for real-time monitoring of scene data storage and retrieval.

[0041] In another alternative implementation, importing multi-source data into the database can also be achieved by configuring data middleware to enable unified access to multiple data sources. Structured data, semi-structured data, and unstructured data are then standardized and transformed before being written into a multi-model database system to support more complex data fusion needs.

[0042] This invention achieves standardized management of data from different sources, in different formats, and with different time precision by uniformly importing multi-source heterogeneous data into a structured database environment. This improves the compatibility and efficiency of data cleaning, alignment, and subsequent fusion processing, and builds a unified data infrastructure platform for the entire fusion process.

[0043] Specifically, in step A1, use an SQL script (such as DELETE FROM table WHERE columnIS NULL) to delete null or abnormal values.

[0044] ETL tools are used to collect heterogeneous data from multiple sources and store it in a database. The data is stored uniformly in a MySQL database. Stored procedures are created by writing SQL statements to clean invalid or erroneous data in batches, such as null or outlier values ​​in some key fields.

[0045] Specifically, in step A2, the date is standardized to YYYY-MM-DD HH:MM:SS, for example, 2025 / 08 / 2011:14 is converted to 2025-08-2011:14:00.

[0046] Specifically, in step A3, window alignment is performed on the timestamps of data from different sources, including the following steps A31-A33:

[0047] A31. Set the sliding time window parameters and determine the window granularity based on real-time requirements.

[0048] A32. Obtain the timestamp of the original record in each data source and align it to the reference time point of the window based on its position in the window.

[0049] A33. Collect records from multiple data sources within the same window and uniformly label them with the window's time stamp.

[0050] Timestamps from different data sources are aligned using a time window algorithm. Preprocessed data (such as unified timestamps and spatial coordinates) serves as input for dynamic alignment.

[0051] The time window algorithm aligns timestamps from different data sources to ensure data consistency over time. The time window algorithm is as follows:

[0052]

[0053] Among them, T aligned It is the calibrated timestamp, T i It is the first The timestamp of the data source Here is the index of the data source, and n is the total number of data sources.

[0054] In this embodiment, step S2, which involves performing regional grid numbering according to spatial division rules, divides the target geographic area into multiple regular spatial units. A gridding method based on a two-dimensional geographic coordinate system is used to construct equally spaced rectangular grids, and each grid unit is assigned a unique number. By extracting the latitude and longitude coordinates from each data record, the spatial unit it belongs to is determined, and the corresponding grid number is written into the data record as a spatial location label, forming a unified spatial structure index.

[0055] In one alternative implementation, the regional grid numbering process can also be carried out based on administrative boundaries or hydrological zones, and different data records can be assigned to their respective natural watersheds, administrative areas or reservoir control units, and spatial correspondence can be established through regional coding.

[0056] In another alternative implementation, the regional grid numbering process according to the spatial division rules can also be combined with the remote sensing image resolution or sensor deployment density to dynamically generate spatial units in irregular shapes, and use different grid scales for different regions to adapt to the differences in data acquisition accuracy and actual hydrological topography.

[0057] This invention constructs a unified spatial indexing mechanism by numbering and mapping multi-source data in a unified gridded coordinate system. This solves the normalization problem for different data sources with inconsistent spatial reference systems and large resolution differences, providing a stable and reliable spatial organization foundation for subsequent spatiotemporal fusion and spatial aggregation analysis.

[0058] Furthermore, in step S2, a sliding window partitioning process is performed on the preprocessed data, and a region grid numbering process is performed according to the spatial partitioning rules, mapping each data point to the partitioned time window and spatial grid to obtain data with completed spatiotemporal mapping, including the following steps B1-B4:

[0059] B1. Perform continuous window division on the data time axis based on the set time step, and identify the time period corresponding to each data record.

[0060] B2. Generate multiple spatial units for the coverage area according to the set spatial division rules, and assign a unique identification number to each spatial unit.

[0061] B3. Extract spatial location information for each data record and associate it with the corresponding numbered spatial unit.

[0062] B4. Classify data records according to time period and spatial number to form a standardized spatiotemporal index structure.

[0063] Specifically, in step B1, the time window is divided into multiple fixed-length time windows based on real-time requirements (e.g., a time window of 1 minute).

[0064] Specifically, in step B2, the geographic region is divided into multiple grid cells, each with unique spatial coordinates. For example, the geographic region is divided into 500m × 500m grids, and each grid is assigned a unique number (such as G_32N_120E).

[0065] Furthermore, in step B3, spatial location information is extracted for each data record and associated with the corresponding numbered spatial unit, including the following steps B31-B33:

[0066] B31. Read the latitude and longitude coordinate fields contained in the data record and obtain the grid number according to the preset spatial grid rules.

[0067] B32. The spatial grid rule is a equidistant rectangular division rule based on a two-dimensional geographic coordinate system, which ensures that the grid number is unique and indexable throughout the entire area.

[0068] B33. Write the obtained grid number as a spatial location label into the original data record to form a binding relationship between data and space.

[0069] Specifically, in step B4, data points (such as water level data from sensor A) are mapped to corresponding time windows and spatial grids. The aligned data is then distributed according to type using different fusion strategies.

[0070] Each data point is mapped to a corresponding time window and spatial grid to form a unified data structure. Different fusion strategies are employed based on the data type (e.g., real-time monitoring data, historical data, predictive data) to generate a high-precision comprehensive dataset.

[0071] Furthermore, in step S3, the spatiotemporal mapped data is fused according to its data type. This includes performing a weighted average based on dynamically set weights on real-time monitoring data and historical data, and performing Kalman filtering based on dynamically calculated Kalman gain on the predicted data. This includes the following steps C1-C4:

[0072] C1. Set the data fusion weights required for weighted averaging according to the type and attributes of the data source, and allocate the data fusion weights required for weighted averaging to the real-time monitoring data and historical data.

[0073] C2. Within the same time period and spatial unit, perform a weighted average operation on monitoring data from different sources.

[0074] C3. Obtain the difference between the state estimate and the observed values ​​for the predicted data, and calculate the Kalman gain required for fusion.

[0075] C4. Update the predicted state values ​​based on the calculated Kalman gain to complete the fusion processing of the predicted data.

[0076] The flowchart of this invention is as follows Figure 2 As shown, the adaptive fusion strategy is as follows: Figure 3 As shown.

[0077] In this embodiment, the data fusion weights set in step C1 can be dynamically adjusted by assigning weight coefficients to sensor data, remote sensing data, and historical monitoring data based on the reliability, accuracy level, and acquisition timeliness of different data sources. Weights are set with reference to attributes such as sensor type, installation location, and fault records. For example, high-precision radar level gauges are assigned higher weights, while remote sensing data affected by clouds are assigned lower weights. These weight parameters are then incorporated into the weighted calculation logic along with the data before fusion.

[0078] In one alternative implementation, the data fusion weights can also be set as fixed weights according to the institution or system to which the data source belongs. For example, data calibrated by the water conservancy department is fixed at 1.0, data from ordinary stations is set at 0.6, and remote sensing data is uniformly set at 0.3. These weights are not dynamically adjusted over time and are used for fixed strategy scenarios that can be implemented quickly.

[0079] In another alternative implementation, the data fusion weights can also be set based on the historical fusion results to build a feedback model, perform reverse learning on the error of the fusion results, and automatically adjust the weight factors of various data sources to achieve a weight configuration method based on adaptive optimization of fusion accuracy.

[0080] This invention dynamically sets weighted fusion parameters based on data source attributes and credibility, ensuring that the contribution of different types of data in the fusion process is controllable and differentiated. This reduces the risk of low-quality data interfering with the results, enhances the stability and credibility of the fusion results, and provides a basis for weight control in refined fusion algorithms.

[0081] Specifically, in step C2, a weighted average method is used to fuse real-time monitoring data and historical data. The weights can be adjusted based on the reliability and timeliness of the data. The weighted average formula is:

[0082]

[0083] Among them, D fused It is the merged data. It is the first Data from one data source, yes The weight of each data source. It is not a fixed value, but a dynamic one, which takes into account the reliability and timeliness of the data source and assigns weights accordingly, including high-precision sensors (such as radar level gauges and ultrasonic flow meters). Higher, for example, Common sensors (such as float-type water level gauges), Satellite remote sensing data (affected by cloud cover and resolution) Depending on the specific quality; meteorological model forecast data It depends on the model's accuracy.

[0084] Furthermore, in step C3, the difference between the state estimate and the observed values ​​is obtained for the predicted data, and the Kalman gain required for fusion is calculated, including the following steps C31-C34:

[0085] C31. Before performing fusion processing, the estimated state of the previous cycle is used as the initial value for the current prediction.

[0086] C32. Collect sensor observation data corresponding to the predicted state, and perform mean processing on the observation data within a time window to reduce the impact of single-point fluctuations on the estimation process.

[0087] C33. Based on the difference between the predicted and observed values, construct a set of weighting factors for calculating the Kalman gain by combining the current state covariance and the observation error variance.

[0088] C34. Dynamically calculate the Kalman gain and use it to correct the predicted state, forming a fused state value adjusted by the observed data, which serves as the state estimation input for the next cycle.

[0089] Specifically, in step C3, the predicted data is fused using a Kalman filter algorithm to reduce prediction errors. The state update equation for the Kalman filter is:

[0090]

[0091] in, It is the state estimate at the current moment (the obtained fused data value). It is the state prediction from the previous moment (the existing prediction value), K k It is the Kalman gain (which determines the correction weight of the observation to the prediction).

[0092] Kalman gain K k It is not a fixed constant, but a time-varying matrix dynamically calculated based on system noise, observation noise, and current estimation uncertainty. Its core function is to automatically balance the confidence levels of predicted and observed values, thereby achieving optimal state estimation. The calculation formula is:

[0093] K k =P k|k-1 H T HP k|k-1 H T +R) -1

[0094] Among them, P k|k-1 To predict the covariance matrix, H is the observation matrix, T is the transpose of the matrix, and R is the observation noise covariance.

[0095] Taking the actual calculation of predicted reservoir water level as an example:

[0096] The state vector is:

[0097]

[0098] Predicting the covariance matrix:

[0099]

[0100] Observation matrix (water level only):

[0101] H =

[10]

[0102] Observation noise covariance (sensor error):

[0103] R = [0.01]

[0104] That is, the standard deviation is 0.1m.

[0105] Calculate HP k|k-1 H T :

[0106]

[0107] Calculate HP k|k-1 H T +R:

[0108] 0.04 + 0.01 = 0.05

[0109] Find the inverse:

[0110] 0.05 -1 =20

[0111] Calculate K k :

[0112]

[0113] Kalman gain

[0114] H is the current observation value, and H is the observation matrix. If only the water level sensor is used, then: H =

[10] .

[0115] This refers to the numerical value obtained after fusion. The purpose is to predict the current state (such as water level and flow velocity) based on historical data. The predicted value is then corrected using actual observations (such as satellite remote sensing data or sensor data) to obtain a more accurate estimate.

[0116] After data fusion, the fused data can be input into a water conservancy digital twin model to simulate and predict dynamic changes in hydrology, water quality, etc.

[0117] Example 3 is an embodiment of the present invention. This embodiment provides a water conservancy digital twin data fusion system based on dynamic alignment of multi-source heterogeneous data, including a preprocessing module, a dynamic alignment module and a data fusion module.

[0118] The preprocessing module is used to collect data from different sensor devices, remote sensing platforms and meteorological models, and to perform data cleaning, format standardization and time window-based timestamp alignment to obtain preprocessed data.

[0119] The dynamic alignment module is used to perform sliding window partitioning on the preprocessed data and perform regional grid numbering according to the spatial partitioning rules, mapping each data point to the partitioned time window and spatial grid to obtain data with complete spatiotemporal mapping.

[0120] The data fusion module is used to perform fusion processing on the data that has completed spatiotemporal mapping, according to the data type. Specifically, it performs weighted averaging processing on real-time monitoring data and historical data based on dynamically set weights, and performs Kalman filtering processing on predicted data based on dynamically calculated Kalman gain.

[0121] This embodiment also provides an electronic device applicable to a water conservancy digital twin data fusion method based on dynamic alignment of multi-source heterogeneous data, comprising: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the water conservancy digital twin data fusion method based on dynamic alignment of multi-source heterogeneous data as proposed in the above embodiment.

[0122] This embodiment also provides a storage medium storing a computer program that, when executed by a processor, implements a water conservancy digital twin data fusion method based on dynamic alignment of multi-source heterogeneous data as proposed in the above embodiment.

[0123] The storage medium proposed in this embodiment and the water conservancy digital twin data fusion method based on dynamic alignment of multi-source heterogeneous data proposed in the above embodiments belong to the same inventive concept. Technical details not described in detail in this embodiment can be found in the above embodiments, and this embodiment has the same beneficial effects as the above embodiments.

[0124] Based on the above description of the implementation methods, those skilled in the art can clearly understand that the present invention can be implemented using software and necessary general-purpose hardware, and of course, it can also be implemented using hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk, or optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of the various embodiments of the present invention.

[0125] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for fusing water conservancy digital twin data based on dynamic alignment of multi-source heterogeneous data, characterized in that: include, Water conservancy data were collected from different sensor devices, remote sensing platforms and meteorological models, and the data were cleaned, formatted, and timestamp aligned based on time windows to obtain preprocessed data. The preprocessed data is divided into sliding windows and then divided into regional grids according to spatial division rules. Each data point is mapped to the divided time window and spatial grid to obtain the data with complete spatiotemporal mapping. For the data that has completed spatiotemporal mapping, fusion processing is performed according to the data type. Specifically, real-time monitoring data and historical data are subjected to weighted averaging based on dynamically set weights, and predicted data are subjected to Kalman filtering based on dynamically calculated Kalman gain.

2. The water conservancy digital twin data fusion method based on dynamic alignment of multi-source heterogeneous data as described in claim 1, characterized in that: The data cleaning, format standardization, and timestamp alignment based on time windows are performed, including: Multi-source data from sensor devices, remote sensing platforms, and meteorological models are imported into the database, and data processing tools are used to delete data records with null values. Perform a uniform format conversion on the information field representing time, so that all data records use the same time expression method; Based on the set time window strategy, window alignment is performed on the timestamps of data from different sources to generate a dataset with a unified time base.

3. The water conservancy digital twin data fusion method based on dynamic alignment of multi-source heterogeneous data as described in claim 2, characterized in that: The process of performing sliding window partitioning on the preprocessed data and performing region grid numbering according to spatial partitioning rules includes, The data time axis is divided into continuous windows based on the set time step, and the time period corresponding to each data record is identified. Based on the set spatial division rules, multiple spatial units are generated in the coverage area, and each spatial unit is assigned a uniquely identifiable number; Extract spatial location information from each data record and associate it with the corresponding numbered spatial unit; Data records are categorized according to time period and spatial number to form a standardized spatiotemporal index structure.

4. The water conservancy digital twin data fusion method based on dynamic alignment of multi-source heterogeneous data as described in claim 3, characterized in that: The data that has undergone spatiotemporal mapping is then fused according to its data type. This includes performing a weighted average process based on dynamically set weights on real-time monitoring data and historical data, and performing Kalman filtering on predicted data based on dynamically calculated Kalman gain. Set the data fusion weights required for weighted averaging based on the type and attributes of the data source, and allocate the data fusion weights required for weighted averaging to the real-time monitoring data and historical data; Within the same time period and spatial unit, a weighted average operation is performed on monitoring data from different sources; For predictive data, obtain the difference between the state estimate and the observed values, and calculate the Kalman gain required for fusion; The predicted state values ​​are updated based on the calculated Kalman gain, thus completing the fusion processing of the predicted data.

5. The water conservancy digital twin data fusion method based on dynamic alignment of multi-source heterogeneous data as described in claim 4, characterized in that: The process of performing window alignment on timestamps from data from different sources is described. include, Set the sliding time window parameters and determine the window granularity based on real-time requirements; Get the timestamp of the original record in each data source and align it to the reference time point of the window based on its position in the window; Records from multiple data sources within the same window are aggregated and uniformly labeled with the window's timestamp.

6. The water conservancy digital twin data fusion method based on dynamic alignment of multi-source heterogeneous data as described in claim 5, characterized in that: The step of extracting spatial location information for each data record and associating it with a corresponding numbered spatial unit includes: Read the latitude and longitude coordinate fields contained in the data record and obtain the grid number according to the preset spatial grid rules; The spatial grid rule is a equidistant rectangular division rule based on a two-dimensional geographic coordinate system, which ensures that the grid number is unique and indexable throughout the entire area; The obtained grid number is written into the original data record as a spatial location label, forming a binding relationship between data and space.

7. The water conservancy digital twin data fusion method based on dynamic alignment of multi-source heterogeneous data as described in claim 6, characterized in that: The process of obtaining the difference between the state estimate and the observed values ​​of the predicted data, and calculating the Kalman gain required for fusion, includes: Before performing the fusion process, the predicted state estimation results of the previous cycle are used as the initial values ​​for the current prediction; Collect sensor observation data corresponding to the predicted state, and perform mean processing on the observation data within a time window to reduce the impact of single-point fluctuations on the estimation process; Based on the difference between the predicted and observed values, a set of weighting factors for calculating the Kalman gain is constructed by combining the current state covariance and the observation error variance. The Kalman gain is dynamically calculated and used to correct the predicted state, forming a fused state value adjusted by the observed data, which serves as the state estimation input for the next cycle.

8. A water conservancy digital twin data fusion system based on dynamic alignment of multi-source heterogeneous data, employing the water conservancy digital twin data fusion method based on dynamic alignment of multi-source heterogeneous data as described in any one of claims 1 to 7, characterized in that, include: Preprocessing module, dynamic alignment module, and data fusion module; The preprocessing module is used to collect data from different sensor devices, remote sensing platforms and meteorological models, and to perform data cleaning, format standardization and time window-based timestamp alignment to obtain preprocessed data. The dynamic alignment module is used to perform sliding window partitioning on the preprocessed data and perform regional grid numbering according to the spatial partitioning rules, so as to map each data point to the partitioned time window and spatial grid, and obtain data with completed spatiotemporal mapping. The data fusion module is used to perform fusion processing on the data that has completed spatiotemporal mapping according to the data type. Specifically, it performs weighted averaging processing on real-time monitoring data and historical data based on dynamically set weights, and performs Kalman filtering processing on predicted data based on dynamically calculated Kalman gain.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the water conservancy digital twin data fusion method based on dynamic alignment of multi-source heterogeneous data as described in any one of claims 1 to 7.

10. 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 steps of the water conservancy digital twin data fusion method based on dynamic alignment of multi-source heterogeneous data as described in any one of claims 1 to 7.

Citation Information

Cited By

  • Roadside holographic sensing platform data synchronization method and system based on big data

    CN122093408A