Multi-dimensional multi-spatio-temporal data base plate construction method and system

Abstract

The invention provides a multi-dimensional multi-spatio-temporal data base plate construction method and system, and the method comprises the steps: carrying out the standardization processing of obtained original multi-source heterogeneous data, and generating a standardized data set; performing multi-level space fusion and dynamic time weighting processing based on the standardized data set to generate a multi-level fusion spatio-temporal data set; performing space-time grid aggregation and adaptive node screening on the multi-level fusion space-time data set to generate an optimized space-time node set; based on the optimized space-time node set, dynamic association is constructed, and an association model is generated and comprises event rules, topological relations and dynamic weights; and generating a multi-dimensional multi-spatiotemporal data base plate based on the optimized spatiotemporal node set and the association model. By adopting the method, the data base plate which can cover global features and has real-time response capability can be constructed, and the decision-making precision and efficiency of scenes such as water conservancy disaster prevention and control and environment monitoring can be improved.

Description

Technical Field

[0001] This invention belongs to the field of data processing, and in particular relates to a method and system for constructing a multidimensional, multi-temporal data base. Background Technology

[0002] The construction of a multidimensional, multi-temporal data platform is a key technology in the fields of big data analytics and intelligent decision-making. Its core lies in integrating multi-source, heterogeneous data to build an analytical framework capable of reflecting complex spatiotemporal dynamics. This technology is crucial for scenarios such as water resource management, disaster prevention and control, and river basin scheduling, as it provides comprehensive data support for dynamic decision-making. However, existing methods often face the dilemma of balancing data heterogeneity and dynamism when processing multidimensional, spatiotemporal data. Traditional solutions often rely on fixed rules or single features for data filtering, making it difficult to adapt to the diverse data sources (such as remote sensing, sensors, and operational systems) and real-time changes in water resource management scenarios, resulting in analytical results lacking global representativeness or timeliness.

[0003] When constructing a multidimensional, multi-temporal data base, the core challenge lies in effectively selecting spatiotemporal nodes that represent the essential characteristics of the data. Due to the uneven distribution and frequency of change of spatiotemporal data density, data points in some areas may be redundant due to high density (e.g., urban hydrological stations), while in other areas, key information may be lost due to sparseness (e.g., rainfall monitoring in remote mountainous areas). A single selection rule cannot simultaneously capture the characteristics of both. This further leads to the challenge of analyzing correlation strength: the correlation between different nodes is jointly affected by geographical location, time span, and data type (e.g., the lagged response of water level and meteorological data). Traditional methods struggle to dynamically adjust strategies to adapt to complex relationships. For example, in flood disaster early warning, if the selection strategy ignores the spatiotemporal lagged correlation between upstream rainfall and downstream water level, redundant nodes may be mistakenly selected or key stations may be missed, weakening the data base's ability to represent the evolution of disasters. Summary of the Invention

[0004] Based on this, it is necessary to provide a method and system for constructing a multidimensional, multi-temporal data base to address the aforementioned technical problems. This system can construct a data base that can both cover global features and have real-time response capabilities by adaptively selecting spatiotemporal nodes and dynamically associating them with models, taking into account the heterogeneity of data density and the complexity of change frequency. This will improve the decision-making accuracy and efficiency in scenarios such as water conservancy disaster prevention and control and environmental monitoring.

[0005] Firstly, this application provides a method for constructing a multidimensional, multi-temporal data base, including:

[0006] The acquired raw multi-source heterogeneous data is standardized to generate a standardized dataset;

[0007] Multi-level fused spatiotemporal datasets are generated by performing multi-level spatial fusion and dynamic temporal weighting based on standardized datasets.

[0008] Spatiotemporal grid aggregation and adaptive node filtering are performed on multi-level fused spatiotemporal datasets to generate an optimized spatiotemporal node set;

[0009] Dynamic associations are constructed based on optimized spatiotemporal node sets to generate an association model, which includes event rules, topological relationships, and dynamic weights.

[0010] A multidimensional, multi-temporal data base is generated based on an optimized spatiotemporal node set and a correlation model.

[0011] In one embodiment, the acquired raw multi-source heterogeneous data is standardized to generate a standardized dataset, including:

[0012] Spatial benchmark transformation and time series alignment are performed on the original multi-source heterogeneous data to generate a unified benchmark dataset;

[0013] The attribute fields of the benchmark unified dataset are formatted to obtain a formatted dataset. The formatting process includes unit standardization, field type conversion, and null value imputation.

[0014] Anomaly detection and noise cleaning are performed on the formatted dataset to generate a standardized dataset.

[0015] In one embodiment, multi-level spatial fusion and dynamic temporal weighting are performed based on a standardized dataset to generate a multi-level fused spatiotemporal dataset, including:

[0016] Spatial analysis was performed on the standardized dataset, which was divided into topographic-level dataset, meso-level dataset, and micro-level dataset.

[0017] The terrain-level dataset is subjected to basic geographic element fusion processing to generate terrain-level fused data.

[0018] Dynamic monitoring data and geographic entity association processing are performed on the meso-level dataset to generate meso-level fused data;

[0019] The micro-level dataset is integrated with 3D facility models to generate micro-level fused data.

[0020] A time-dimensional sliding window weighting method is applied to topographic-level fused data, meso-level fused data, and micro-level fused data to generate a multi-level fused spatiotemporal dataset.

[0021] In one embodiment, spatiotemporal grid aggregation and adaptive node filtering are performed on the multi-level fused spatiotemporal dataset to generate an optimized spatiotemporal node set, including:

[0022] Spatiotemporal grid cell partitioning is performed on the multi-level fused spatiotemporal dataset to generate a spatiotemporal grid cell set;

[0023] Multidimensional index statistical aggregation processing is performed on the spatiotemporal grid cell set to generate a grid statistical index set;

[0024] High-density grid cell identification is performed based on grid statistical index set to determine redundant grid cell set;

[0025] Spatial clustering center extraction is performed on redundant grid cell sets to generate representative node sets;

[0026] Low-density grid cell identification is performed based on grid statistical index set to determine sparse grid cell set;

[0027] Spatiotemporal interpolation is performed on the sparse grid cell set to generate a supplementary node set;

[0028] The representative node set and the supplementary node set are topologically integrated to generate an optimized spatiotemporal node set.

[0029] In one embodiment, a dynamic association is constructed based on an optimized spatiotemporal node set to generate an association model, including:

[0030] Based on the acquired historical disaster events, the event trigger threshold is defined for the optimized spatiotemporal node set, and an event rule set is generated.

[0031] Perform event mapping processing on the event rule set and the optimized spatiotemporal node set to generate an event response relationship set;

[0032] Based on the pre-constructed spatial topology of hydraulic entities, a connectivity path analysis is performed on the optimized spatiotemporal node set to generate a node connectivity network.

[0033] Spatiotemporal correlation calculations are performed using a node-connected network to generate an initial correlation strength matrix.

[0034] The initial association strength matrix is ​​dynamically weighted and optimized using a pre-trained machine learning model to generate an association model.

[0035] Secondly, this application also provides a multidimensional, multi-temporal, and spatiotemporal data baseboard construction system, comprising:

[0036] The data standardization module is used to standardize the acquired raw multi-source heterogeneous data and generate a standardized dataset.

[0037] The multi-spatiotemporal fusion module is used to perform multi-level spatial fusion and dynamic temporal weighting processing based on standardized datasets to generate multi-level fused spatiotemporal datasets.

[0038] The node optimization module is used to perform spatiotemporal grid aggregation and adaptive node filtering on multi-level fused spatiotemporal datasets to generate an optimized spatiotemporal node set.

[0039] The association modeling module is used to construct dynamic associations based on an optimized spatiotemporal node set and generate an association model, which includes event rules, topological relationships, and dynamic weights.

[0040] The data base generation module is used to generate multidimensional, multi-temporal data bases based on optimized spatiotemporal node sets and correlation models.

[0041] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the above-described multidimensional multi-temporal data baseboard construction method.

[0042] Fourthly, this application also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the above-described method for constructing a multidimensional, multi-temporal data base plate.

[0043] The aforementioned method and system for constructing a multidimensional, multi-temporal data platform standardizes the original multi-source heterogeneous data to eliminate data source differences and form a unified dataset. Through multi-level spatial fusion and dynamic temporal weighting, it integrates spatiotemporal features at different scales and strengthens the weight of timely data, generating a multi-level fused spatiotemporal dataset. Using spatiotemporal grid aggregation and adaptive node selection techniques, it dynamically optimizes node distribution to address data density heterogeneity, generating an optimized spatiotemporal node set covering global features. Based on this node set, it constructs an association model including event rules, topological relationships, and dynamic weights to accurately capture complex associations such as spatiotemporal lag. Finally, it fuses the optimized node set and the dynamic association model to generate a multidimensional, multi-temporal data platform. This solution achieves the goal of balancing data density heterogeneity and the complexity of change frequency, constructing a data platform with both global representation capabilities and real-time response characteristics, thereby effectively improving the decision-making accuracy and efficiency in water conservancy disaster prevention and control and environmental monitoring scenarios. Attached Figure Description

[0044] To more clearly illustrate the technical solutions in the embodiments or related technologies of this application, the accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0045] Figure 1 This is a flowchart illustrating a method for constructing a multidimensional, multi-temporal data base plate according to an embodiment of the present invention.

[0046] Figure 2 This is a flowchart illustrating a multidimensional, multi-temporal data standardization method provided by the present invention.

[0047] Figure 3This is a schematic diagram of a multidimensional, multi-temporal data baseboard construction system provided in an embodiment of the present invention. Detailed Implementation

[0048] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0049] First, a brief introduction to the terms used in the embodiments of this application will be given.

[0050] A multidimensional, multi-temporal data platform refers to a unified data carrier that integrates spatial three-dimensional coordinates (longitude, latitude, and elevation), temporal dimensions (historical / real-time / predictive time series), and correlation dimensions (event rules, topological networks). Its core feature is the collaborative mapping between the macroscopic geographical environment and the microscopic entity state through spatial hierarchical expression and dynamic weighting mechanisms, supporting decision-making and simulation in scenarios such as flooding simulation and resource scheduling. Unlike traditional static platforms, its multidimensionality is reflected in its ability to simultaneously drive the coupled computation of spatial analysis, temporal series simulation, and event response within a unified framework.

[0051] Multi-level spatial fusion involves layering and consistent processing of data in different forms, such as points, lines, surfaces, and grids, across multiple spatial levels from macro to micro, so that information at each level can form a continuous and conflict-free overall spatial expression while maintaining scale characteristics.

[0052] Adaptive node selection is a process of dynamically selecting key spatiotemporal nodes based on data density and frequency of change. Its core mechanism is to identify high-value areas (such as disaster-prone areas) and suppress redundant data points (such as densely populated urban sites). It uses machine learning to evaluate node information entropy to achieve dynamic optimization of the selection strategy, thus solving the problem that traditional fixed rules are difficult to adapt to spatial heterogeneity.

[0053] Based on the above definitions, the implementation environment of the multi-dimensional, multi-temporal data baseboard construction method provided in this application embodiment will be described. Indicatively, the implementation environment includes: a terminal, sensing devices, a processor, and a memory. The terminal is connected to the processor, sensing devices, and memory via a network. Sensing devices include, but are not limited to, radar level gauges, ultrasonic flow meters, multi-parameter water quality monitors, UAV LiDAR, temperature and humidity sensors, anemometers, and image acquisition devices. The processor can be a central processing unit, a graphics processing unit, or an artificial intelligence chip. The memory can be a distributed cloud storage system or a local server cluster, and is not limited here.

[0054] Based on the above explanations of terms and implementation environments, the application scenarios of the embodiments of this application are described. The multidimensional, multi-temporal, and spatiotemporal data baseboard construction method provided in the embodiments of this application can be applied to scenarios including but not limited to the following:

[0055] In the context of water conservancy disaster prevention and control, specifically in flood early warning, this method constructs a unified data base by standardizing and multi-level spatiotemporal fusion of heterogeneous data from multiple sources, including satellite remote sensing, hydrological sensors, and meteorological operational systems. It then dynamically optimizes the distribution of monitoring stations through adaptive node selection technology (e.g., strengthening the weight of sparse rain gauge stations in mountainous areas) and accurately captures the lag response relationship between upstream and downstream water levels based on a dynamic correlation model. The resulting data base can reflect the entire evolution of the disaster in real time, improving the accuracy of early warnings and the efficiency of emergency response.

[0056] In the scenario of watershed water resource allocation, in the face of cross-regional water resource allocation needs, this solution first integrates multi-dimensional data such as reservoir storage, agricultural water use, and ecological flow, and highlights real-time water demand characteristics through dynamic time weighting. Then, it eliminates data density differences through spatiotemporal grid aggregation (such as balancing dense urban monitoring stations and sparse agricultural monitoring points), and analyzes water use conflicts by combining topological relationship modeling. The generated data base can support dynamic water allocation decisions and optimize the efficiency of dry season scheduling and flood retention during wet season.

[0057] In the context of ecological and environmental monitoring, the technical solution addresses the issue of tracing the source of pollution in watersheds by integrating multi-dimensional information such as data from water quality monitoring stations, polluting enterprises, and climate factors. It strengthens the coverage of pollution-sensitive areas through adaptive node screening. By modeling pollution diffusion paths through event rules (such as the scouring and migration of pollutants by rainfall), and by using a dynamic weighting mechanism to respond in real time to sudden pollution events, the data base forms a comprehensive environmental risk profile, providing a continuous spatiotemporal basis for pollution prevention and control decisions.

[0058] This is merely an illustrative example; the dynamic perception method for safety status of coal mine personnel provided in this application embodiment can also be applied to other application scenarios. This is only an example and does not limit the specific application scenarios.

[0059] In one exemplary embodiment, such as Figure 1 As shown, a method for constructing a multidimensional, multi-temporal data base is provided. This embodiment illustrates the application of this method to a terminal in the aforementioned implementation environment. It is understood that this method can also be applied to a server, and further to a system including both a terminal and a server, and is implemented through interaction between the terminal and the server. In this embodiment, the method includes the following steps 101 to 105:

[0060] Step 101: Standardize the obtained raw multi-source heterogeneous data to generate a standardized dataset.

[0061] Specifically, the process involves acquiring raw, heterogeneous data from remote sensing satellites, ground sensors, and hydrological operational systems. Specifically, it can receive data streams with different spatiotemporal resolutions and formats (such as GeoTIFF remote sensing images, CSV format sensor readings, and JSON operational records) through data access interfaces. The raw data undergoes standardization processing, for example: unifying the spatiotemporal reference by converting geographic coordinates and time fields using the CGCS2000 coordinate system and UTC timestamps; and normalizing numerical attributes by mapping continuous variables such as water level and rainfall to the [0,1] interval through deviation standardization, while processing discrete variables (such as station type) through one-hot encoding. For instance, missing data is filled using spatiotemporal Kriging interpolation to generate a standardized dataset with a consistent structure. This step eliminates the heterogeneity of the source data, ensuring that the data foundation for subsequent processing has spatiotemporal comparability and computational compatibility, and avoiding analytical biases caused by data format conflicts.

[0062] Step 102: Perform multi-level spatial fusion and dynamic temporal weighting processing based on the standardized dataset to generate a multi-level fused spatiotemporal dataset.

[0063] Specifically, in the stage of generating a multi-level fused spatiotemporal dataset, the fusion path can be divided into topographic level, meso-level, and micro-level. Topographic level processing overlays DEM data with remote sensing images to generate a macro-elevation-texture base. Meso-level processing uses density clustering algorithms to associate dynamic monitoring points with geographic entity boundaries, such as establishing spatial attribution relationships between water level sensors and river channel vector surfaces. Micro-level processing uses ICP registration algorithms to achieve spatial coupling between BIM models and laser point clouds. Dynamic time-weighted processing sets differentiated strategies based on data frequency: for second-level sensor streams, mean aggregation with adaptive window size (the window shrinks as variance increases) is used; for daily-level remote sensing data, weights are assigned according to an exponential decay model (the weight decay coefficient is 0.8 after 24 hours). This fusion mechanism overcomes the accuracy loss problem when integrating cross-scale data and can generate a multi-level fused spatiotemporal dataset that simultaneously contains micro-details (such as fluctuations in urban drainage outlet flow) and macro-trends (such as monthly changes in watershed water levels), breaking through the limitations of single-scale analysis.

[0064] Step 103: Perform spatiotemporal grid aggregation and adaptive node filtering on the multi-level fused spatiotemporal dataset to generate an optimized spatiotemporal node set.

[0065] Specifically, the core of spatiotemporal grid aggregation and adaptive node selection lies in three-dimensional voxelization processing. For example, the spatial resolution is set to a 50m×50m grid, the elevation layer accuracy is 0.5m, and the time unit granularity is 1 hour. Grid statistical aggregation uses spatial overlay analysis to calculate seven types of indicators (including point density, extreme values, and coefficient of variation) for each unit. Node selection is performed in two paths: in high-density areas, representative nodes are extracted by K-means clustering with contour coefficient optimization (the number of clusters is dynamically determined by the error function); in low-density areas, virtual completion nodes are generated by combining Kriging spatial interpolation and ARIMA time series prediction. The resulting optimized node set achieves a balanced distribution across the entire region, overcoming the blind spot defects caused by uneven deployment of monitoring equipment.

[0066] Step 104: Construct dynamic associations based on the optimized spatiotemporal node set and generate an association model. The association model includes event rules, topological relationships, and dynamic weights.

[0067] Specifically, a third-order association model is constructed based on a node set. The event rule set is generated through historical disaster case mining; for example, "continuous 3-hour rainfall > 50 mm and water level rise rate > 0.2 m / h" is defined as a flood triggering rule. Topological relationship analysis uses graph theory algorithms to construct a node connectivity network and calculate the shortest seepage path and facility dependence strength. Dynamic weight optimization introduces spatiotemporal lag correlation coefficients, such as quantifying the delay time of upstream rainfall events' impact on downstream water levels (0.8 for T+3h, decaying to 0.3 for T+6h), and learns an adaptive weight adjustment mechanism through an LSTM model. The generated association model achieves unified modeling of static topological constraints and dynamic event responses.

[0068] Step 105: Generate a multidimensional spatiotemporal data base based on the optimized spatiotemporal node set and association model.

[0069] Specifically, the data classification of the optimized spatiotemporal node set is performed using data structure parsing. For example, structured data (such as sensor streams) is indexed using an R-tree and stored in a distributed database; unstructured data (such as 3D models) is spatially encoded and stored in an object storage repository. Further, the application of the correlation model includes two-stage computation: hydrodynamic simulation uses the finite volume method to solve the Saint-Venant equation, generating an inundation depth grid; emergency path planning integrates the A* algorithm and topological constraints to output evacuation routes, and uses a WebGL engine to overlay and render the spatial base data, inundation simulation results, and path planning results, generating an interactive, dynamic decision-making platform. Furthermore, this multidimensional, multi-spatiotemporal data platform can be stored in a distributed spatiotemporal database (such as GeoMesa) and has an open RESTful API interface; when new data flows in, an incremental update process is triggered, recalculating node selection and correlation weights only for the affected spatiotemporal grids, ensuring the platform's real-time performance. In the scenario of water conservancy disaster prevention and control, the data platform constructed in this way can simultaneously capture the correlation effect between sudden rainfall in mountainous areas and delayed urban drainage, improving decision-making accuracy.

[0070] The aforementioned method and system for constructing a multidimensional, multi-temporal data platform standardizes the original multi-source heterogeneous data to eliminate data source differences and form a unified dataset. Through multi-level spatial fusion and dynamic temporal weighting, it integrates spatiotemporal features at different scales and strengthens the weight of timely data, generating a multi-level fused spatiotemporal dataset. Using spatiotemporal grid aggregation and adaptive node selection techniques, it dynamically optimizes node distribution to address data density heterogeneity, generating an optimized spatiotemporal node set covering global features. Based on this node set, it constructs an association model including event rules, topological relationships, and dynamic weights to accurately capture complex associations such as spatiotemporal lag. Finally, it fuses the optimized node set and the dynamic association model to generate a multidimensional, multi-temporal data platform. This solution achieves the goal of balancing data density heterogeneity and the complexity of change frequency, constructing a data platform with both global representation capabilities and real-time response characteristics, thereby effectively improving the decision-making accuracy and efficiency in water conservancy disaster prevention and control and environmental monitoring scenarios.

[0071] like Figure 2 As shown, in one embodiment, the acquired raw multi-source heterogeneous data is standardized to generate a standardized dataset, including:

[0072] Step 201: Perform spatial benchmark transformation and time series alignment on the original multi-source heterogeneous data to generate a unified benchmark dataset.

[0073] Specifically, the original multi-source heterogeneous data may be acquired from discrete data sources, including but not limited to aerospace remote sensing image sets, time-series streams acquired by ground sensors, geographic information system vector layers, and / or operational platform records. These data may have different spatial coordinate systems (e.g., CGCS2000, WGS84, and local coordinate systems coexist) and inconsistent timestamp granularities (from second-level to day-level). To address the spatial misalignment caused by the differences in coordinate systems among multi-source data, this method first employs a seven-parameter Bursa model to perform spatial benchmark transformation. For example, it dynamically converts the WGS-84 coordinate point set acquired by GPS devices into spatial points under the CGCS2000 national geodetic coordinate system, eliminating spatial overlay biases across platforms. Simultaneously, to unify the temporal discreteness generated by asynchronous acquisition systems, minute-level hydrological sensor data is resampled using the UTC time axis, aligning it with the daily-updated satellite image timestamps to the hourly time benchmark. This resolves the interference of temporal discontinuities on the fusion calculation. The resulting unified benchmark dataset possesses coordinate system uniformity and temporal series continuity, establishing a spatial-temporal framework for subsequent attribute processing.

[0074] Step 202: Format the attribute fields of the benchmark unified dataset to obtain a formatted dataset. The formatting process includes unit standardization, field type conversion, and null value imputation.

[0075] Specifically, the attribute field formatting process performed on a benchmark unified dataset aims to eliminate numerical dimensional inconsistencies. For example, this method maps millimeter-per-hour rainfall data and cubic meter-per-second river flow data to a standardized interval of [0,1] using a dimensionless function; for text-based land use type fields, one-hot encoding is used to convert them into binary feature vectors; for missing monitoring points, an inverse distance weighted interpolation method is used to generate spatially continuous water quality parameter surfaces for these points, based on the first law of geography. Furthermore, for abnormally negative flow records caused by equipment failure, numerical zeroing correction based on business rules is performed. This process transforms heterogeneous data into a quantifiable and comparable homogeneous dataset, resolving analytical obstacles caused by differences in units of measurement and types, and providing numerical comparability assurance for fusion modeling.

[0076] Step 203: Perform anomaly detection and noise cleaning on the formatted dataset to generate a standardized dataset.

[0077] For example, to address potential anomalies in formatted datasets, this method deploys a two-stage cleaning mechanism. Specifically, the anomaly detection stage integrates statistical principles and domain knowledge: it identifies statistical outliers in water level data based on the three-standard-deviation rule, while simultaneously filtering out irrational data through physical threshold constraints (such as soil moisture not exceeding saturation moisture content). Noise cleaning employs sliding window mid-range filtering to smooth minute-level water level fluctuations, eliminating pulse interference while preserving flood peak characteristics. Spatial noise in remote sensing images is eliminated by morphological opening operations to remove small isolated pixels. The generated standardized dataset achieves three characteristics: spatial topology error is controlled to sub-meter accuracy, the time series meets stationarity requirements, and the attribute value range conforms to physical constraints. This dataset serves as input for multi-level spatiotemporal fusion, ensuring data quality for substrate construction from the source. This embodiment addresses spatial misalignment of multi-source data through dynamic spatial benchmark transformation, overcomes asynchronous acquisition gaps through temporal resampling, establishes a unified calculation benchmark for cross-source data through attribute formatting, and achieves scientific cleaning by fusing statistical and physical constraints through anomaly detection. The entire standardized process transforms raw heterogeneous data into high-precision, highly consistent spatiotemporal data assets, removing technical obstacles for subsequent fusion computing and laying a computable foundation for multidimensional and multi-spatiotemporal data.

[0078] In one embodiment, multi-level spatial fusion and dynamic temporal weighting are performed based on a standardized dataset to generate a multi-level fused spatiotemporal dataset, including:

[0079] Spatial analysis was performed on the standardized dataset, which was divided into topographic-level dataset, meso-level dataset, and micro-level dataset.

[0080] Specifically, spatial analysis is a preprocessing technique that uses both geographic scale and semantic type as dual criteria to deconstruct input datasets into three levels of data entities: topographic, meso-level, and micro-level. Its output can directly serve multi-level fusion processes, resolving fusion distortion issues caused by resolution conflicts in cross-scale data. For example, based on the spatial scale characteristics of geographic entities, the data is divided into three levels: the topographic level encompasses large-scale digital elevation models and satellite remote sensing imagery, used to construct a macro-geographic framework; the meso-level includes river flow and water level data collected by hydrological sensor networks and vector boundaries of hydraulic engineering projects; and the micro-level consists of facility point cloud scans and building information models. This hierarchical division mechanism solves the information redundancy and accuracy contradiction problems encountered in traditional single-scale data fusion.

[0081] The terrain-level dataset is processed by fusion of basic geographic elements to generate terrain-level fused data.

[0082] Specifically, for elevation data and remote sensing data in a topographic-level dataset, pixel-level weighted overlay technology can be used for fusion. The raster elevation values ​​of the digital elevation model (DEM) are used as a base, and multispectral satellite imagery is overlaid to generate a surface texture-elevation composite model. For example, near-infrared and water system indices can be fused through band operations to enhance the expression of water body features. Alternatively, multi-source basic geographic data (such as DEM digital elevation and remote sensing land cover) can be extracted, and data conflicts can be eliminated through spatial overlay analysis and topological consistency verification. For instance, water body boundaries extracted from satellite remote sensing are fuzzily matched with vector water system maps, and a high-precision water system network is generated based on the principle of minimum geometric error. Furthermore, DEM data of different precisions are integrated using an elevation-weighted averaging algorithm to generate a seamless topographic surface. This embodiment forms a macro-geographic base through topographic-level fusion data, supporting large-scale analyses such as flood inundation simulations.

[0083] Dynamic monitoring data and geographic entity association processing are performed on the meso-level dataset to generate meso-level fused data.

[0084] Specifically, this technical solution dynamically associates monitoring data with geographic entities in meso-level datasets. It establishes spatial semantic associations between monitoring points (such as rain gauges and water quality sensors) and geographic entities (such as rivers and administrative regions). By matching spatial locations, monitoring points are linked to the nearest entities, and attribute mapping rules are constructed (e.g., associating station rainfall with its sub-basin). For example, in the Taihu Lake basin, monitoring data from 12 water quality stations around the lake are associated with the lake's surrounding river entities, forming an "entity-monitoring" binding relationship. This meso-level fusion of data enables logical linkage between dynamic indicators and geographic objects, providing structured input for regional situation analysis.

[0085] Micro-level datasets are integrated with facility 3D models to generate micro-level fused data.

[0086] Specifically, the extraction of micro-level datasets focuses on the spatial features of high-precision local facilities. This level accommodates discrete structural data with millimeter to centimeter precision, such as building information models and laser point cloud scans. Key structural point clusters of facilities (e.g., the structural point set of a pump station inlet) are identified using point cloud density clustering algorithms, and then categorized and grouped according to the facility's semantic type (e.g., "sluice gate" or "pipeline node"). This operation allows micro-level data, such as bridge deformation monitoring point clouds, to independently participate in subsequent model registration, avoiding scale conflicts with macro-level topographic data.

[0087] A time-dimensional sliding window weighting method is applied to topographic-level fused data, meso-level fused data, and micro-level fused data to generate a multi-level fused spatiotemporal dataset.

[0088] Specifically, this embodiment consists of three types of datasets: topographic, meso-level, and micro-level. This division creatively solves the fundamental contradiction in cross-scale data fusion at the technical level: the topographic level preserves large-scale continuous geographical features, providing a framework for flood simulation; the meso-level establishes a precise mapping relationship between "monitoring points and geographical entities," supporting dynamic process analysis; and the micro-level maintains the integrity of facility structures, ensuring the physical realism of scenarios such as dam break simulations. The resulting hierarchical dataset forms the input basis for subsequent multi-level spatial fusion processes. Compared to the simple spatial range clipping or attribute filtering in conventional geographic information systems, this method divides data levels based on spatial resolution physical indicators while integrating the semantic attributes of geographical entities to determine the data's functional affiliation. For example, in river water level analysis, centimeter-level precision dam structure monitoring data (micro-level) will be processed separately from ten-meter-level river boundary data (meso-level), whereas in traditional mixed processing, dam data would be smoothly masked by the river polygon boundaries. Through this type of gradation, the subsequent integration stage can superimpose the evolution of the meso-level water level field on the macro-topographic base and embed micro-structural deformations at key facility sites, forming a precise expression of multi-scale linkage.

[0089] In one embodiment, spatiotemporal grid aggregation and adaptive node filtering are performed on the multi-level fused spatiotemporal dataset to generate an optimized spatiotemporal node set, including:

[0090] Spatiotemporal grid cell partitioning is performed on the multi-level fused spatiotemporal dataset to generate a spatiotemporal grid cell set.

[0091] Specifically, a grid unit system is constructed on the three-dimensional spatial axes (longitude X, latitude Y, and elevation Z) and the time axis T: the spatial plane is divided into rectangular units using the equal-area Mercator projection, the elevation direction is layered according to hydrological characteristics (such as riverbed layer, floodplain layer, and high stratum), and the time axis is divided into time periods according to the disaster response cycle (for example, the flood evolution process is set into hourly time windows). The generated spatiotemporal grid unit set completely encloses the target area in all dimensions, providing a standardized calculation container for subsequent indicator statistics.

[0092] Multidimensional index statistical aggregation processing is performed on the spatiotemporal grid cell set to generate a grid statistical index set.

[0093] Specifically, within the framework of a spatiotemporal grid cell set, multi-dimensional data aggregation calculations are further performed. This method designs differentiated statistical approaches based on the characteristics of fused data at different levels: the elevation range and mean slope are statistically analyzed within topographic-level cells; the temporal mean and fluctuation variance of water level monitoring points are aggregated within meso-level cells; and the spatial distribution density of facility structure point clouds is extracted within micro-level cells. For example, the "maximum water level fluctuation within 24 hours" is calculated as a flood response sensitivity index for a specific river segment grid cell, forming a grid statistical index set containing three types of indicators: spatial characteristics, temporal dynamics, and structural stability. This result quantitatively expresses the spatiotemporal integrated state within the cell, overcoming the limitations of traditional single-indicator statistics.

[0094] High-density grid cell identification is performed based on the grid statistical index set to determine the redundant grid cell set.

[0095] Spatial clustering centers are extracted from redundant grid cell sets to generate representative node sets.

[0096] Specifically, redundant units can be determined by calculating the data density index based on the grid statistical index set, as shown in the following formula:

[0097]

[0098] Among them, I d N is the data density exponent. s V represents the number of sensors within a cell, reflecting the spatial distribution density of data acquisition points; A represents the planar projected area of ​​this grid cell (unit: square kilometers), used to standardize the spatial density, max(V t ) and min(V t ) represents the extreme values ​​of the current unit's index within the target time period (e.g., the maximum and minimum values ​​of the water level variation coefficient), characterizing the dynamic change amplitude; δ is the variation tolerance threshold set by domain experts, used to normalize the variation amplitude (e.g., a typical value of 0.2 in flood analysis), avoiding incomparability between indices of different dimensions, especially when equipment is densely packed within the unit ( High values ​​and drastic fluctuations in monitored values ​​(max(V)) t )-min(V t When the value is high, a binarization decision is made by setting a high-density threshold. d When the threshold is exceeded, it is identified as a high information redundancy area. For example, in urban flood control areas, the dense grid cells of underground pipe network sensors will trigger redundancy marking. Furthermore, for such redundant grid cell sets, an improved spectral clustering algorithm is used to extract feature nodes: first, a spatial proximity and index similarity matrix between grid cells is constructed, and then the cluster centers are obtained through feature vector decomposition to generate a representative node set that retains the core features of the area.

[0099] Low-density grid cell identification is performed based on the grid statistical index set to determine the sparse grid cell set.

[0100] Spatiotemporal interpolation is performed on the sparse grid cell set to generate a supplementary node set.

[0101] The representative node set and the supplementary node set are topologically integrated to generate an optimized spatiotemporal node set.

[0102] Specifically, for sparse grid cell sets with data density below a threshold in the grid statistical index set (such as missing areas of rain gauges in remote mountainous areas), a spatiotemporal collaborative interpolation mechanism is used to generate supplementary nodes. For example, in the spatial dimension, the Kriging algorithm is used to generate virtual monitoring points based on hydrological indicators of neighboring cells; in the temporal dimension, historical similar scenarios are combined to fill in time-series features. Furthermore, the representative node set and the supplementary node set are input into a graph theory topology model: a spatial adjacency matrix and elevation gradient-related edges are established between nodes, and then conflicting connections are eliminated using the minimum spanning tree algorithm, forming an optimized spatiotemporal node set with balanced spatial distribution and continuous topological structure.

[0103] In one embodiment, a dynamic association is constructed based on an optimized spatiotemporal node set to generate an association model, including:

[0104] Based on the acquired historical disaster events, the event trigger threshold is defined for the optimized spatiotemporal node set, and an event rule set is generated.

[0105] Specifically, this method quantitatively models the disaster response mechanism of nodes based on a historical disaster database. This database contains spatiotemporal evolution data of events such as floods, levee breaches, and pipeline bursts recorded over the past decade, covering multi-dimensional monitoring indicators such as water level, flow velocity, and facility stress. The method extracts dynamic parameter sequences strongly correlated with the disaster for each optimized node; for example, it extracts the hourly water level rise rate for river nodes and the motor vibration spectrum energy value for pump station nodes. Quantile regression analysis is used to determine the synergistic effect threshold of each parameter. For example, the triggering rule for urban flooding events is defined as follows: when the instantaneous flow rate at a drainage pipe node exceeds 85% of the historical peak and the variance of the water level rise rate at three adjacent manhole nodes is greater than 0.3, it is determined to be a high-risk flooding event, thus generating an event rule set containing multi-level constraints.

[0106] Event mapping is performed on the event rule set and the optimized spatiotemporal node set to generate an event response relationship set.

[0107] Specifically, when mapping event rule sets to real-time node datasets, dynamic Bayesian networks can be used to establish disaster transmission relationships. Threshold conditions in the rules are transformed into state variables of network nodes, and the intensity of disaster propagation between nodes is quantified using conditional probability tables. For example, in an upstream dam failure event, based on breach flow and downstream river gradient data, the inundation probability of each downstream node within a specific time period is calculated, generating an event response relationship set expressing the chain relationship of "dam failure - inundation range - infrastructure failure." This mapping process specifically introduces a hydraulic transmission time delay factor, causing the response probability of downstream nodes to exhibit an exponential decay characteristic over time, thus accurately reflecting the spatiotemporal propagation characteristics of flood waves.

[0108] Based on the pre-constructed spatial topology of hydraulic entities, a connectivity path analysis is performed on the optimized spatiotemporal node set to generate a node connectivity network.

[0109] For example, when constructing the connectivity path of hydraulic entities, this method breaks through the limitation of traditional hydrological networks that only consider natural river channels, integrates the logical blocking effect of artificial control facilities, and performs multi-level connectivity analysis based on a pre-constructed hydraulic spatial topology database (including the geometric connection relationship and hydraulic attribute parameters of river channels, culverts, gates, and pumping stations). It establishes physical connectivity paths based on the river centerline and pipeline orientation, and superimposes dynamic control parameters such as gate opening and pumping station start-stop status to generate a weighted transmission network. For example, an improved A* algorithm can be used for path search, converting the Manning coefficient into the basic component of path cost, and transforming the pumping station operating status into a Boolean blocking factor, so that the algorithm can automatically avoid currently closed diversion channels and output a node connectivity network reflecting the real-time control status.

[0110] Spatiotemporal correlation is calculated by connecting nodes to the network to generate an initial correlation strength matrix.

[0111] Specifically, based on the aforementioned connected network, the spatiotemporal lag correlation between nodes is further calculated. In the spatial dimension, this method uses a graph convolutional network to aggregate features of neighboring nodes, capturing the cascading impact of upstream water level changes on downstream areas. In the temporal dimension, a time-delay cross-correlation algorithm is used to analyze the phase difference of node indicator sequences. For example, when calculating the flow time-series correlation between two gate nodes, the maximum cross-correlation number and its corresponding time delay are detected through a sliding time window to generate an initial correlation strength matrix. This matrix not only records the spatial adjacency relationship between nodes but also quantifies the time lag effect of disaster propagation, such as the 6-hour delay correlation between the peak rainfall in mountainous areas and the downstream flood peak.

[0112] The initial association strength matrix is ​​dynamically weighted and optimized using a pre-trained machine learning model to generate an association model.

[0113] Specifically, the initial correlation strength is dynamically optimized through a pre-trained spatiotemporal graph neural network. This neural network uses the evolution sequence of node states in historical disaster events as training data and adopts a graph attention mechanism to adaptively adjust node weights. The network can automatically learn correlation patterns under different types of disaster scenarios. For example, in typhoon and rainstorm scenarios, the transmission weight of nodes with short-term heavy rainfall is enhanced, and in drought scenarios, the control weight of reservoir scheduling nodes is strengthened. The optimized correlation model can dynamically update the connection strength according to real-time weather forecasts. For example, after predicting the rainfall in the next three hours, the transmission coefficient of relevant river nodes is increased by 40%, thereby accurately simulating the disaster diffusion path under extreme weather conditions.

[0114] In summary, the multidimensional spatiotemporal data base construction method provided in this application performs spatial decoupling analysis on the original monitoring data, and stratifies it into topographic-level continuous curved surfaces, meso-level dynamic monitoring fields, and micro-level facility structure datasets based on the scale characteristics of geographic entities, thereby eliminating spatial semantic conflicts during cross-scale fusion. It adopts four-dimensional spatiotemporal grid aggregation technology, and performs multidimensional index statistical calculations on multi-level data based on unified spatial projection, elevation stratification, and time window division, quantifying the deformation index, dynamic variation coefficient, and structural density parameters of grid units, and generating a grid statistical index set that integrates physical characteristics and dynamic behavior. To address the issue of uneven data distribution, a dual-threshold density sensing mechanism is employed. Redundant grid cells are identified based on a composite index of equipment density and index variability. Feature nodes are extracted using spectral clustering algorithms to achieve dimensionality reduction in high-density areas. Simultaneously, for sparse regions, supplementary nodes are generated by combining Kriging spatial interpolation and historical scene reconstruction, forming an optimized spatiotemporal node set with balanced spatial coverage and complete feature preservation. Furthermore, by modeling dynamic associations, a multimodal knowledge fusion disaster response engine is constructed: node event trigger thresholds are defined based on a historical disaster rule base; the topological network of water conservancy entities is combined with spatiotemporal lag correlations; the cascading effects of neighboring nodes are captured using graph convolutional networks; and the phase characteristics of disaster propagation are quantified using a time-delay cross-correlation algorithm. Further, a pre-trained spatiotemporal graph neural network is introduced, and an attention mechanism is used to dynamically adjust node association weights, enabling the model to adapt to different meteorological conditions and control scenarios. The final output of this scheme, a multidimensional, multi-spatial data platform with an association model, overcomes the limitations of traditional static topology, achieving scenario perception and real-time simulation of disaster transmission paths. Through the above steps, this technical solution addresses the heterogeneity of data density and the differences in change frequency by using adaptive node selection, and captures the spatiotemporal coupling relationship of real-time evolution by using dynamic correlation modeling, forming a data base that covers global features and has real-time response capabilities, which can improve the decision-making accuracy and efficiency of disaster prevention and control and environmental monitoring such as water conservancy.

[0115] It should be understood that although the steps in the flowcharts of the embodiments described above 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 embodiments described above 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.

[0116] Based on the same inventive concept, this application also provides a multidimensional spatiotemporal data base construction system 10 for implementing the multidimensional spatiotemporal data base construction method described above. The solution provided by this system is similar to the solution described in the above method. Therefore, the specific limitations of one or more embodiments of the multidimensional spatiotemporal data base construction system 10 provided below can be found in the limitations of the multidimensional spatiotemporal data base construction method described above, and will not be repeated here.

[0117] In one exemplary embodiment, such as Figure 3 As shown, a multidimensional, multi-temporal data baseboard construction system 10 is provided, comprising:

[0118] The data standardization module 11 is used to standardize the acquired raw multi-source heterogeneous data and generate a standardized dataset.

[0119] Multi-spatiotemporal fusion module 12 is used to perform multi-level spatial fusion and dynamic temporal weighting processing based on standardized datasets to generate multi-level fused spatiotemporal datasets;

[0120] Node optimization module 13 is used to perform spatiotemporal grid aggregation and adaptive node filtering on multi-level fused spatiotemporal datasets to generate an optimized spatiotemporal node set;

[0121] The association modeling module 14 is used to construct dynamic associations based on the optimized spatiotemporal node set and generate an association model, which includes event rules, topological relationships and dynamic weights.

[0122] The data base generation module 15 is used to generate a multidimensional, multi-temporal data base based on the optimized spatiotemporal node set and the association model.

[0123] In one embodiment, the data standardization module 11 includes:

[0124] The spatiotemporal registration unit is used to perform spatial benchmark transformation and time series alignment on the original multi-source heterogeneous data to generate a unified benchmark dataset.

[0125] The attribute normalization unit is used to format attribute fields of the benchmark unified dataset to obtain a formatted dataset. The formatting process includes unit normalization, field type conversion, and null value imputation.

[0126] The data cleaning unit is used to perform anomaly detection and noise removal on formatted datasets to generate standardized datasets.

[0127] In one embodiment, the multi-temporal fusion module 12 includes:

[0128] Multi-level spatial partitioning units are used to perform spatial analysis on standardized datasets, dividing them into topographic-level datasets, meso-level datasets, and micro-level datasets.

[0129] The macro-element fusion unit is used to perform basic geographic element fusion processing on the terrain-level dataset to generate terrain-level fused data.

[0130] The dynamic entity association unit is used to dynamically associate monitoring data with geographic entities in the meso-level dataset to generate meso-level fused data.

[0131] The facility integration unit is used to integrate and process micro-level datasets into three-dimensional facility models, generating micro-level fused data.

[0132] The temporal weighting unit is used to perform time-dimensional sliding window weighting on terrain-level fused data, meso-level fused data, and micro-level fused data to generate a multi-level fused spatiotemporal dataset.

[0133] In one embodiment, the node optimization module 13 includes:

[0134] The grid partitioning unit is used to perform spatiotemporal grid unit partitioning on a multi-level fused spatiotemporal dataset to generate a spatiotemporal grid unit set.

[0135] The index aggregation unit is used to perform multi-dimensional index statistical aggregation processing on the spatiotemporal grid unit set to generate a grid statistical index set.

[0136] A redundancy identification unit is used to identify high-density grid cells based on a set of grid statistical indicators and determine the set of redundant grid cells.

[0137] Clustering optimization unit is used to extract spatial cluster centers from redundant grid cell sets and generate representative node sets.

[0138] Sparse identification unit is used to identify low-density grid cells based on grid statistical index set and determine sparse grid cell set;

[0139] Interpolation supplementation unit, used to perform spatiotemporal interpolation on sparse grid cell set to generate supplementary node set;

[0140] The topology integration unit is used to perform topology integration processing on the representative node set and the supplementary node set to generate an optimized spatiotemporal node set.

[0141] In one embodiment, the association modeling module 14 includes:

[0142] The event rule generation unit is used to define event triggering thresholds for the optimized spatiotemporal node set based on the acquired historical disaster events, and generate an event rule set.

[0143] The causal mapping unit is used to perform event mapping processing on the event rule set and the optimized spatiotemporal node set to generate an event response relationship set.

[0144] The topology analysis unit is used to perform connectivity path analysis on the optimized spatiotemporal node set based on the pre-constructed spatial topology relationship of hydraulic entities, and generate a node connectivity network.

[0145] The lag correlation unit is used to calculate the spatiotemporal lag correlation through the node-connected network and generate the initial correlation strength matrix.

[0146] The weight optimization unit is used to dynamically optimize the weights of the initial association strength matrix using a pre-trained machine learning model, thereby generating an association model.

[0147] In one embodiment, a computer device is provided, including a memory and a processor, the memory storing a computer program, the processor executing the computer program to implement the steps of the multidimensional multi-temporal data baseboard construction method as described above.

[0148] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps in the above method embodiments.

[0149] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The components described as separate parts may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this disclosure according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0150] The above-described embodiments are merely illustrative of several implementation methods of the embodiments of this application, and their descriptions are relatively specific and detailed. However, they should not be construed as limiting the scope of the patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the embodiments of this application, and these modifications and improvements all fall within the protection scope of the embodiments of this application.

Claims

1. A method for constructing a multidimensional, multi-temporal data base, characterized in that, The method includes: The acquired raw multi-source heterogeneous data is standardized to generate a standardized dataset; Multi-level spatial fusion and dynamic temporal weighting are performed on the standardized dataset to generate a multi-level fused spatiotemporal dataset. Spatiotemporal grid aggregation and adaptive node filtering are performed on the multi-level fused spatiotemporal dataset to generate an optimized spatiotemporal node set; Based on the optimized spatiotemporal node set, a dynamic association is constructed to generate an association model, which includes event rules, topological relationships, and dynamic weights. A multidimensional, multi-temporal data base is generated based on the optimized spatiotemporal node set and the association model.

2. The method according to claim 1, characterized in that, The standardization process for the acquired raw multi-source heterogeneous data to generate a standardized dataset includes: Spatial benchmark transformation and time series alignment are performed on the original multi-source heterogeneous data to generate a benchmark unified dataset; The attribute fields of the benchmark unified dataset are formatted to obtain a formatted dataset, wherein the formatting process includes unit standardization, field type conversion, and null value imputation; Anomaly detection and noise cleaning are performed on the formatted dataset to generate the standardized dataset.

3. The method according to claim 1, characterized in that, The process of performing multi-level spatial fusion and dynamic temporal weighting processing based on the standardized dataset to generate a multi-level fused spatiotemporal dataset includes: Spatial analysis was performed on the standardized dataset, which was divided into terrain-level dataset, meso-level dataset, and micro-level dataset. The terrain-level dataset is subjected to basic geographic element fusion processing to generate terrain-level fused data; Dynamic monitoring data and geographic entity association processing are performed on the aforementioned meso-level dataset to generate meso-level fused data; The micro-level dataset is subjected to facility 3D model integration processing to generate micro-level fused data; The terrain-level fused data, the meso-level fused data, and the micro-level fused data are weighted by a time-dimensional sliding window to generate the multi-level fused spatiotemporal dataset.

4. The method according to claim 1, characterized in that, The step of performing spatiotemporal grid aggregation and adaptive node filtering on the multi-level fused spatiotemporal dataset to generate an optimized spatiotemporal node set includes: The multi-level fused spatiotemporal dataset is divided into spatiotemporal grid cells to generate a spatiotemporal grid cell set; The spatiotemporal grid cell set is subjected to multidimensional index statistical aggregation processing to generate a grid statistical index set; Based on the set of grid statistical indicators, high-density grid cell identification processing is performed to determine the set of redundant grid cells; The redundant grid cell set is subjected to spatial clustering center extraction processing to generate a representative node set; Based on the set of grid statistical indicators, low-density grid cells are identified to determine the sparse grid cell set. Spatiotemporal interpolation is performed on the sparse grid cell set to generate a supplementary node set; The representative node set and the supplementary node set are subjected to topology integration processing to generate the optimized spatiotemporal node set.

5. The method according to claim 1, characterized in that, The step of constructing dynamic associations based on the optimized spatiotemporal node set and generating an association model includes: Based on the acquired historical disaster events, event trigger thresholds are defined for the optimized spatiotemporal node set, and an event rule set is generated. The event rule set and the optimized spatiotemporal node set are subjected to event mapping processing to generate an event response relationship set; Based on the pre-constructed spatial topology of hydraulic entities, the optimized spatiotemporal node set is subjected to connectivity path analysis to generate a node connectivity network. Spatiotemporal lag correlation is calculated through the node-connected network to generate an initial correlation strength matrix; The initial association strength matrix is ​​dynamically weighted and optimized using a pre-trained machine learning model to generate the association model.

6. A multidimensional, multi-temporal, and spatiotemporal data baseboard construction system, characterized in that, The system includes: The data standardization module is used to standardize the acquired raw multi-source heterogeneous data and generate a standardized dataset. The multi-temporal fusion module is used to perform multi-level spatial fusion and dynamic temporal weighting processing based on the standardized dataset to generate a multi-level fused spatiotemporal dataset. The node optimization module is used to perform spatiotemporal grid aggregation and adaptive node filtering on the multi-level fused spatiotemporal dataset to generate an optimized spatiotemporal node set. The association modeling module is used to construct dynamic associations based on the optimized spatiotemporal node set and generate an association model, which includes event rules, topological relationships and dynamic weights. The data base generation module is used to generate a multidimensional, multi-temporal data base based on the optimized spatiotemporal node set and the association model.

7. 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 method of any one of claims 1 to 5.

8. 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 method of any one of claims 1 to 5.

Images