Urban infrastructure digital twinborn management method supported by Beidou space-time base

Abstract

The invention relates to the technical field of intelligent cities, in particular to a Beidou space-time base supported urban infrastructure digital twinborn management method, which comprises the following steps of S1, space-time reference grid construction and code distribution: distributing Beidou grid codes for each urban infrastructure entity; s2, multi-modal data acquisition and space-time alignment processing: generating a sensing data stream with a space-time label; s3, feature extraction: generating an infrastructure state feature matrix; s4, space-time rule knowledge base establishment and anomaly detection: generating a space-time anomaly vector; s5, space-time memory enhanced prediction and early warning interval generation: generating a space-time early warning interval of a prediction deviation threshold value; s6, performing exception and early warning overlay analysis and causal reasoning optimization: generating a maintenance decision instruction with Beidou grid code positioning; according to the invention, the reliability and maintenance efficiency of the urban infrastructure are improved, and powerful technical support is provided for intelligent operation and maintenance of the urban infrastructure.

Description

Technical Field

[0001] The present invention relates to the field of smart city technology, and in particular to a digital twin management method for urban infrastructure supported by a Beidou space-time base. Background Art

[0002] With the acceleration of urbanization, the scale of urban infrastructure is expanding, involving a wide variety of facilities, including transportation, energy, communications and other facilities. In order to ensure the efficient, safe and reliable operation of these infrastructures, traditional urban infrastructure management methods can no longer meet the needs of modern urban management. The monitoring, early warning and maintenance of urban infrastructure urgently need to rely on advanced technical means, especially the integrated application of technologies such as the Internet of Things, big data and artificial intelligence. Digital twin technology, as a technology that maps physical entities through virtual models, can provide more accurate and real-time data support for the management of urban infrastructure, and is one of the key technologies for realizing smart city management.

[0003] Although the existing urban infrastructure management systems have made certain progress in some aspects, they generally have the following shortcomings. On the one hand, most traditional infrastructure management systems use static models and lack accurate perception and prediction of the real-time status of infrastructure, resulting in poor timeliness of fault prediction and maintenance decisions. On the other hand, the existing systems have limited capabilities in data fusion and spatiotemporal correlation analysis, and are unable to process multi-source and multi-dimensional sensor data, nor can they effectively perform spatiotemporal alignment and optimization of real-time data. In addition, traditional methods rely on experience and manual judgment in anomaly detection, fault diagnosis and maintenance decisions, resulting in frequent misjudgments and missed judgments. Existing technologies also lack adaptive learning mechanisms based on dynamic data, making it difficult to adapt to the rapidly changing urban infrastructure environment.

[0004] The purpose of this invention is to provide a digital twin management method for urban infrastructure based on the Beidou space-time base support, realize an adaptive intelligent management system, and effectively improve the operation efficiency and maintenance quality of urban infrastructure. Summary of the invention

[0005] The present invention provides a digital twin management method for urban infrastructure supported by a Beidou space-time base.

[0006] The urban infrastructure digital twin management method supported by the BeiDou space-time base includes the following steps: S1, spatiotemporal reference grid construction and code allocation: Based on the satellite-based ground-based augmented data of the BeiDou-3 system, a centimeter-level spatiotemporal reference grid covering the target area is generated, and a BeiDou grid code with spatiotemporal uniqueness is allocated to each urban infrastructure entity; S2, multimodal data collection and spatiotemporal alignment processing: real-time status data is collected through multimodal sensors deployed on urban infrastructure, and the real-time status data is spatiotemporally aligned with the Beidou grid code to generate a perception data stream with spatiotemporal labels; S3, feature extraction: construct a digital twin evolution model based on spatiotemporal graph neural network (ST-GNN), take the perception data stream input of spatiotemporal labels as input data, and generate the infrastructure status feature matrix with spatiotemporal correlation characteristics; S4, establishment of spatiotemporal rule knowledge base and anomaly detection: establish a spatiotemporal rule knowledge base for the evolution of urban infrastructure status, detect abnormal status of urban infrastructure and generate spatiotemporal anomaly vectors by comparing the infrastructure status feature matrix with the historical evolution pattern in the spatiotemporal rule knowledge base; S5, spatiotemporal memory enhanced prediction and warning interval generation: Construct a spatiotemporal memory enhanced prediction model, based on the current infrastructure status feature matrix and the spatiotemporal distribution characteristics of the Beidou grid code, predict the evolution trajectory of the infrastructure status within the future set time window, and generate the spatiotemporal warning interval of the prediction deviation threshold; S6, anomaly and warning superposition analysis and causal reasoning optimization: superimpose the spatiotemporal anomaly vector and the spatiotemporal warning interval for analysis, dynamically adjust the search space of the causal reasoning engine, output a fault diagnosis report with spatiotemporal coordinates, and generate maintenance decision instructions with Beidou grid code positioning based on the fault diagnosis report.

[0007] Optionally, the construction of the spatiotemporal reference grid and the allocation of codes in S1 include: S11, satellite-based augmentation data acquisition: obtain the original observation data of the target area through the BeiDou-3 system's satellite-based augmentation service (BDSBAS) and ground-based augmentation station network (CORS); S12, Spatiotemporal reference grid division and hierarchical design: Design a multi-level spatiotemporal grid division scheme based on the spatial scope and application requirements of the target area, including: Spatial level: Divide the target area into regular grid cells with adjustable side lengths, and the minimum cell size is 0.5m×0.5m; Time level: Time slices are defined based on Beidou timing signals, with a minimum time resolution of 1 second; S13, grid code generation and space-time binding: Beidou grid coding algorithm is used to generate a four-dimensional code including longitude, latitude, elevation and time for each grid unit. The four-dimensional code structure meets the following conditions: ; in, , , , are the longitude, latitude, elevation and time of the grid center point, , , , is the coordinate of the origin of the space-time reference, , , , is the grid spatial and temporal resolution, is the Beidou standard hash function, is the Beidou grid code; S14, Infrastructure entity coding mapping: The spatial coordinates of urban infrastructure entities are mapped With timing status Mapped to the corresponding grid code through a dynamic interpolation algorithm.

[0008] Optionally, the multimodal data collection and spatiotemporal alignment processing in S2 includes: S21, Multimodal sensor deployment and data collection: Deploy multimodal sensor networks on urban infrastructure, including structural health monitoring sensors, environmental sensors, and operating status sensors, to collect multimodal data in real time, including structural health data, environmental data, and operating status data, among which; The structural health data includes strain, vibration, and displacement; The environmental data include temperature, humidity, and wind speed; The operating status data includes load, flow, and pressure; S22, data preprocessing: preprocessing the collected multimodal data, including outlier removal, missing value filling, and noise filtering; S23, time-space benchmark unification and data alignment: the pre-processed multi-modal data is time-space aligned with the Beidou grid code, specifically including: Spatial alignment: Map the data in the sensor's local coordinate system to the Beidou grid coordinate system through the coordinate transformation matrix; Time alignment: Synchronize and calibrate sensor timestamps based on Beidou timing signals; S24, spatiotemporal label generation and data stream encapsulation: Add spatiotemporal labels to the aligned multimodal data to generate standardized perception data streams.

[0009] Optionally, the data preprocessing in S22 includes: S221, outlier removal: based on The criterion removes abnormal data from multimodal data; S222, missing value filling: using spatiotemporal interpolation to fill missing data in multimodal data; S223, Noise filtering: Remove high-frequency noise in multimodal data through wavelet transform.

[0010] Optionally, the feature extraction in S3 includes: S31, construction of space-time graph structure: using Beidou grid codes as nodes and space-time associations as edges to construct a dynamic space-time graph structure , where the node feature is expressed as , For the The perception data of a grid, is the embedding vector of Beidou grid code, and the edge weight is expressed as , is the spatial attenuation factor; S32, spatial graph convolution calculation: using gated graph convolution layer to extract spatial features; S33, time-gated recurrent unit: models the timing characteristics of each node; S34, Spatiotemporal Attention Mechanism: Dynamically Calculating Node Importance Weights ; S35, feature matrix generation: Sort the output of the spatiotemporal graph neural network by Beidou grid code to generate the infrastructure status feature matrix ; S36, model training: end-to-end training is performed using a hybrid loss function.

[0011] Optionally, the establishment of the spatiotemporal rule knowledge base and the anomaly detection in S4 include: S41, construction of spatiotemporal rule knowledge base: extract the spatiotemporal pattern of the evolution of urban infrastructure status based on historical status data, use multidimensional correlation analysis technology to abstract the spatiotemporal relationship of infrastructure status changes into rule triples (trigger conditions, spatiotemporal constraints, evolution results), and use spatiotemporal tensor decomposition technology to reduce the dimensionality of high-dimensional relationships and store them, forming a spatiotemporal rule knowledge base that includes normal state evolution paths and failure modes; S42, spatiotemporal anomaly detection and vector generation: The generated infrastructure status feature matrix is ​​compared with the historical evolution pattern in the spatiotemporal rule knowledge base in multiple dimensions. Through spatiotemporal similarity calculation and dynamic threshold judgment, the abnormal state that deviates from the normal evolution path is identified. For the detected abnormal state, a spatiotemporal anomaly vector including the anomaly type, spatiotemporal coordinates and degree of deviation is generated, and the potentially associated abnormal propagation path is marked.

[0012] Optionally, the construction of the spatiotemporal rule knowledge base in S41 includes: S411, Historical data preprocessing and spatiotemporal sequence generation: Perform spatiotemporal alignment and normalization on historical status data, divide spatiotemporal units according to Beidou grid codes, and generate a spatiotemporal sequence dataset of infrastructure status evolution , where a single space-time sequence is represented as: ; in, For the Space-time units in time The characteristic matrix of , is the total number of space-time units, is the number of time steps; S412, Multidimensional Association Rule Mining: Use the improved Apriori algorithm to extract spatiotemporal association rules in the form of triples ; S413, Spatiotemporal Tensor Decomposition and Dimensionality Reduction Storage: Constructing a Four-Dimensional Spatiotemporal Tensor , the dimensions represent the spatial grid, time window, state features, and association rules, respectively, and the dimensionality reduction is performed through high-order singular value decomposition (HOSVD); S414, Dynamic update of rule knowledge base: Design an incremental tensor decomposition algorithm, when new data is added , update the rule tensor.

[0013] Optionally, the spatiotemporal anomaly detection and vector generation in S42 includes: S421, spatiotemporal similarity calculation: Use spatiotemporal weighted cosine similarity algorithm to calculate the current feature matrix and the historical feature matrix in the knowledge base Similarity ; S422, dynamic threshold judgment: based on the statistical characteristics of the historical similarity distribution in the knowledge base, dynamically set the abnormality judgment threshold ; S423, Exception vector generation: for satisfying The abnormal state generates a spatiotemporal abnormal vector ; S424, Anomaly Propagation Path Labeling: Based on Spatiotemporal Graph Structure and the causal rules in the knowledge base, and uses the random walk algorithm to predict the abnormal propagation path .

[0014] Optionally, the generation of the spatiotemporal memory enhanced prediction and warning interval in S5 includes: S51, Spatiotemporal Memory Pool Construction: Constructing a spatiotemporal memory pool based on the historical infrastructure status feature matrix ; S52, Gated spatiotemporal attention mechanism: Calculate the current feature matrix Attention weights with historical feature matrix in memory pool ; S53, prediction trajectory generation: generating future time windows through gated recurrent units (GRU) The predicted trajectory ; S54, spatiotemporal warning interval calculation: Generate spatiotemporal warning interval based on the confidence of the predicted trajectory .

[0015] Optionally, the abnormality and warning superposition analysis and causal reasoning optimization in S6 includes: S61, anomaly and warning superposition analysis: the spatiotemporal anomaly vector and the spatiotemporal warning interval Conduct spatiotemporal superposition analysis and calculate the overlap coefficient of anomalies within the warning interval ; S62, Dynamic adjustment of the search space of the causal inference engine: based on the overlap coefficient Dynamically adjust the search space of the causal inference engine, including: Spatial constraint: limit the search range to the Beidou grid code where the abnormal event occurred Center, radius The space-time region, where (Take 10 times the smallest unit of the Beidou grid code); Rule filtering: only retain the subset of causal rules related to the current exception type Type; S63, fault diagnosis report generation: based on the output of the causal reasoning engine, a fault diagnosis report with time and space coordinates is generated, including the Beidou grid code of the root cause event (the spatial location of the fault or anomaly), the timestamp of the root cause event (the specific time when the fault occurred), and the diagnosis confidence (the reliability of the fault diagnosis); S64, maintenance decision instruction generation: Generate maintenance decision instructions with Beidou grid code positioning according to the fault diagnosis report, including priority sorting (arranging root cause events in descending order of confidence) and path planning (generating the optimal maintenance path based on the spatial distribution of Beidou grid codes).

[0016] Beneficial effects of the present invention: The present invention, through the construction and code allocation of the spatiotemporal reference grid, allocates a Beidou grid code with spatiotemporal uniqueness to each infrastructure entity, thereby achieving high-precision spatiotemporal positioning and data association. Combined with multimodal data acquisition and spatiotemporal alignment processing, it can acquire and accurately process the operation data of urban infrastructure in real time, significantly improving the integrity and consistency of data acquisition, and providing a reliable data foundation for subsequent state feature extraction, anomaly detection and prediction models.

[0017] The present invention, through the spatiotemporal memory enhanced prediction and warning interval generation mechanism, combined with the spatiotemporal graph neural network model and the spatiotemporal attention mechanism, effectively improves the accuracy of urban infrastructure status prediction and the timeliness of early warning. Through the spatiotemporal rule knowledge base constructed based on historical status data and the dynamically optimized causal reasoning engine, it can accurately detect abnormal states that deviate from the normal evolution path, and provide timely fault diagnosis reports and maintenance decision instructions. It not only improves the accuracy in the fault diagnosis process, but also enhances the adaptability and robustness of the prediction model in long-term operation, ensuring the efficient operation of urban infrastructure.

[0018] The present invention realizes the adaptive optimization of the intelligent management system through the closed-loop interaction between the digital twin and the physical entity, and enhances the adaptability of the system to complex scenarios. In the maintenance decision-making process, based on the superposition analysis and causal reasoning optimization of spatiotemporal anomalies and warning intervals, the system can generate the optimal maintenance path according to the diagnosis results, provide accurate fault location and maintenance instructions, and improve the response speed and decision-making efficiency of infrastructure faults. Through this intelligent and real-time management method, the reliability and maintenance efficiency of urban infrastructure are significantly improved, providing strong technical support for the intelligent operation and maintenance of urban infrastructure. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only for the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0020] Figure 1 A schematic diagram of a management method flow chart of an embodiment of the present invention; Figure 2 A schematic diagram of generating a spatiotemporal memory enhanced prediction and warning interval according to an embodiment of the present invention. DETAILED DESCRIPTION

[0021] The present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. At the same time, it is explained here that in order to make the embodiments more detailed, the following embodiments are the best and preferred embodiments, and those skilled in the art may also adopt other alternatives to implement some known technologies; and the accompanying drawings are only for more specific description of the embodiments, and are not intended to specifically limit the present invention.

[0022] It should be noted that the references to "one embodiment", "embodiment", "exemplary embodiments", "some embodiments" and the like in the specification indicate that the embodiments described may include specific features, structures or characteristics, but not every embodiment may include the specific features, structures or characteristics. In addition, when a specific feature, structure or characteristic is described in conjunction with an embodiment, it should be within the knowledge of a person skilled in the art to implement such feature, structure or characteristic in conjunction with other embodiments (whether or not explicitly described).

[0023] In general, a term can be understood, at least in part, from its use in context. For example, depending, at least in part, on the context, the term "one or more" as used herein can be used to describe any feature, structure, or characteristic in the singular sense, or can be used to describe a combination of features, structures, or characteristics in the plural sense. Additionally, the term "based on" can be understood as not necessarily intended to convey an exclusive set of factors, but can instead, depending, at least in part, on the context, allow for the presence of other factors that are not necessarily explicitly described.

[0024] like Figure 1-Figure 2 As shown in the figure, the urban infrastructure digital twin management method supported by the BeiDou space-time base includes the following steps: S1, spatiotemporal reference grid construction and code allocation: Based on the satellite-based ground-based augmented data of the BeiDou-3 system, a centimeter-level spatiotemporal reference grid covering the target area is generated, and a BeiDou grid code with spatiotemporal uniqueness is allocated to each urban infrastructure entity; S2, multimodal data collection and spatiotemporal alignment processing: real-time status data is collected through multimodal sensors deployed on urban infrastructure, and the real-time status data is spatiotemporally aligned with the Beidou grid code to generate a perception data stream with spatiotemporal labels; S3, feature extraction: construct a digital twin evolution model based on spatiotemporal graph neural network (ST-GNN), take the perception data stream input of spatiotemporal labels as input data, and generate the infrastructure status feature matrix with spatiotemporal correlation characteristics; S4, establishment of spatiotemporal rule knowledge base and anomaly detection: establish a spatiotemporal rule knowledge base for the evolution of urban infrastructure status, detect abnormal status of urban infrastructure and generate spatiotemporal anomaly vectors by comparing the infrastructure status feature matrix with the historical evolution pattern in the spatiotemporal rule knowledge base; S5, spatiotemporal memory enhanced prediction and warning interval generation: Construct a spatiotemporal memory enhanced prediction model, based on the current infrastructure status feature matrix and the spatiotemporal distribution characteristics of the Beidou grid code, predict the evolution trajectory of the infrastructure status within the future set time window, and generate the spatiotemporal warning interval of the prediction deviation threshold; S6, anomaly and warning superposition analysis and causal reasoning optimization: superimpose the spatiotemporal anomaly vector and the spatiotemporal warning interval, dynamically adjust the search space of the causal reasoning engine, output the fault diagnosis report with spatiotemporal coordinates, and generate maintenance decision instructions with Beidou grid code positioning based on the fault diagnosis report; Through the above content, accurate alignment and spatiotemporal coupling analysis of multi-source data of urban infrastructure are achieved. Combined with dynamic warning interval generation and causal reasoning optimization mechanism, the accuracy of abnormal diagnosis and the timeliness of maintenance decisions are significantly improved. At the same time, through the closed-loop interaction between digital twins and physical entities, an adaptive intelligent management system for urban infrastructure is formed.

[0025] The construction and coding of the spatiotemporal reference grid in S1 include: S11, satellite-based augmentation data acquisition: obtain the original observation data of the target area through the BeiDou-3 system's satellite-based augmentation service (BDSBAS) and ground-based augmentation station network (CORS); S12, Spatiotemporal reference grid division and hierarchical design: Design a multi-level spatiotemporal grid division scheme based on the spatial scope and application requirements of the target area, including: Spatial level: Divide the target area into regular grid cells with adjustable side lengths, and the minimum cell size is 0.5m×0.5m; Time level: Time slices are defined based on Beidou timing signals, with a minimum time resolution of 1 second; S13, grid code generation and space-time binding: Beidou grid coding algorithm is used to generate a four-dimensional code including longitude, latitude, elevation and time for each grid unit. The four-dimensional code structure meets the following conditions: ; in, , , , are the longitude, latitude, elevation and time of the grid center point, , , , is the coordinate of the origin of the space-time reference, , , , is the grid spatial and temporal resolution, It is the Beidou standard hash function, and the output is 128-bit code. is the Beidou grid code; S14, Infrastructure entity coding mapping: The spatial coordinates of urban infrastructure entities are mapped With timing status Mapped to the corresponding grid code through dynamic interpolation algorithm, expressed as: ; in, For entity to The Euclidean distance between neighboring grid centers, For the The encoding of the neighboring grids, Encoding infrastructure entities, For entity to The Euclidean distance between neighboring grid centers; Through the above content, the BeiDou-3 system's satellite-based enhanced data constructs a centimeter-level space-time reference grid, and combines multi-level grid division with four-dimensional coding algorithm to assign a BeiDou grid code with unique space-time characteristics to each urban infrastructure entity. This solves the problems of inconsistent space-time references, redundant coding, and difficulty in cross-system data fusion in traditional methods, significantly improving the accuracy and efficiency of urban infrastructure management. At the same time, a dynamic interpolation algorithm is used to achieve smooth mapping of entity codes, ensuring coding consistency and scalability in complex scenarios, providing reliable space-time reference support for the efficient operation of digital twin cities.

[0026] The multimodal data acquisition and spatiotemporal alignment processing in S2 includes: S21, Multimodal sensor deployment and data collection: Deploy multimodal sensor networks on urban infrastructure, including structural health monitoring sensors, environmental sensors, and operating status sensors, to collect multimodal data in real time, including structural health data, environmental data, and operating status data, among which; Structural health data includes strain, vibration, and displacement; Environmental data includes temperature, humidity, and wind speed; Operation status data includes load, flow, and pressure; S22, data preprocessing: preprocessing the collected multimodal data, including outlier removal, missing value filling, and noise filtering; S23, time-space benchmark unification and data alignment: the pre-processed multi-modal data is time-space aligned with the Beidou grid code, specifically including: Spatial alignment: Map the data in the sensor's local coordinate system to the Beidou grid coordinate system through the coordinate transformation matrix, expressed as: ; in, , , They are respectively coordinate, coordinate, coordinate, is the rotation matrix, , , They are respectively in the local coordinate system coordinate, coordinate, coordinate, , , They are direction, direction, The amount of translation in the direction; Time alignment: Synchronize and calibrate sensor timestamps based on Beidou timing signals to eliminate clock drift errors; S24, spatiotemporal label generation and data stream encapsulation: Add spatiotemporal labels to the aligned multimodal data to generate a standardized perception data stream. The label format is: ; in, is the Beidou grid code corresponding to the data, It is the Beidou synchronization timestamp. Identifies the data type (such as strain, temperature, etc.), Score the data quality, is the space-time label; Through the above content, a multimodal sensor network is deployed and high-precision spatiotemporal alignment of data is achieved based on the Beidou grid code, which solves the problems of inconsistent spatiotemporal benchmarks and difficult fusion of multi-source data in traditional methods, and significantly improves the integrity and consistency of data collection. At the same time, by encapsulating the perception data stream with standardized spatiotemporal labels, it supports efficient data retrieval and quality control, providing a reliable data foundation for the digital twin management of urban infrastructure, and enhancing the real-time, scalability and analysis accuracy of the system.

[0027] The data preprocessing in S22 includes: S221, outlier removal: based on The criterion removes abnormal data in multimodal data, which is expressed as: ; in, For the The value of the data point, is the mean of the data series, is the standard deviation of the data series; ; ; in, is the length of the data sequence; S222, missing value filling: Use spatiotemporal interpolation to fill in missing data in multimodal data, expressed as: ; ; ; in, are the missing values ​​to be filled, For the The values ​​of neighboring data points, For the The weights of neighboring data points, The missing point and The space-time distance between neighboring points, is the timestamp of the missing point, For the The timestamps of the neighboring points, is the Beidou grid code of the missing point, For the Beidou grid code of neighboring points, is the weight coefficient of spatial distance and temporal distance (take ), is the number of neighboring points involved in interpolation (take ); S223, noise filtering: removing high-frequency noise in multimodal data through wavelet transform, specifically including: Wavelet decomposition: ; in, is the original signal, is the wavelet basis function, is the scale function, is the wavelet coefficient (high frequency component), is the scale factor (low frequency component), is the number of decomposition layers; Threshold denoising: wavelet coefficients Perform soft threshold processing, expressed as: ; in, As the threshold, take ( is the noise standard deviation, is the signal length), is the wavelet coefficient after denoising; Wavelet reconstruction: ; in, is the signal after denoising; Through the above content, the quality and reliability of multimodal data are improved. Outlier elimination effectively removes obviously erroneous data, missing value filling ensures the spatiotemporal continuity of data, and noise filtering accurately separates and removes high-frequency noise through wavelet transform, enhancing the robustness of the system and the accuracy of the analysis results.

[0028] Feature extraction in S3 includes: S31, construction of space-time graph structure: using Beidou grid codes as nodes and space-time associations as edges to construct a dynamic space-time graph structure , where the node feature is expressed as , For the The perception data of a grid, is the embedding vector of Beidou grid code, and the edge weight is expressed as , is the spatial attenuation factor; S32, spatial graph convolution calculation: The gated graph convolution layer is used to extract spatial features, expressed as: ; in, For the Tier The hidden state of a node, For the Tier The hidden state of a node, , is the trainable weight matrix, is the Hadamard product, is the Sigmoid function, For the Tier The hidden state of a node, For Node The degree of For Node The degree of is the activation function, For Node The set of neighbors of ; S33, time-gated recurrent unit: models the timing characteristics of each node, expressed as: ; ; ; ; in, , , is the time-gated weight matrix, is the time step, To update the gate, To reset the gate, is a candidate hidden state, is the updated hidden state, is the hidden state at the current time step, is the hidden state of the previous time step; S34, Spatiotemporal Attention Mechanism: Dynamically Calculating Node Importance Weights , expressed as: ; in, is the attention vector, is the attention weight matrix, For Node With Node The attention weight, For Node The hidden state of For Node The hidden state of is the activation function, ; S35, feature matrix generation: Sort the output of the spatiotemporal graph neural network by Beidou grid code to generate the infrastructure status feature matrix , expressed as: ; in, is the final time step, is the total number of grids, For Node The hidden state at the final time step is ; S36, model training: end-to-end training is performed using a hybrid loss function, expressed as: ; in, is the true state label, is the physical constraint prior distribution, , is the balance coefficient, is the mixed loss function, is the mean absolute error, is the KL divergence, Output distribution for the model; Through the above content, the feature extraction model based on the spatiotemporal graph neural network, combined with spatial graph convolution, time-gated recurrent unit and spatiotemporal attention mechanism, has achieved efficient fusion and feature extraction of multi-source data of urban infrastructure. Spatial graph convolution captures the spatial dependency between nodes, time-gated recurrent unit models the dynamic changes of timing, and the spatiotemporal attention mechanism dynamically adjusts the importance of nodes, and finally generates an infrastructure status feature matrix with spatiotemporal correlation characteristics, which significantly improves the accuracy and robustness of feature expression.

[0029] The establishment of spatiotemporal rule knowledge base and anomaly detection in S4 include: S41, construction of spatiotemporal rule knowledge base: extract the spatiotemporal pattern of the evolution of urban infrastructure status based on historical status data, use multidimensional correlation analysis technology to abstract the spatiotemporal relationship of infrastructure status changes into interpretable rule triples (trigger conditions, spatiotemporal constraints, evolution results), and use spatiotemporal tensor decomposition technology to reduce the dimensionality of high-dimensional relationships and store them, forming a spatiotemporal rule knowledge base that includes normal state evolution paths and failure modes; S42, spatiotemporal anomaly detection and vector generation: The generated infrastructure status feature matrix is ​​compared with the historical evolution pattern in the spatiotemporal rule knowledge base in multiple dimensions. The abnormal state that deviates from the normal evolution path is identified through spatiotemporal similarity calculation and dynamic threshold judgment. For the detected abnormal state, a spatiotemporal anomaly vector including the anomaly type, spatiotemporal coordinates and degree of deviation is generated, and the potentially associated abnormal propagation path is marked. Through the above content, the spatiotemporal rule knowledge base based on historical operation and maintenance data, combined with multidimensional correlation analysis and spatiotemporal tensor decomposition technology, has achieved accurate modeling and rule extraction of the state evolution of urban infrastructure. Through multi-dimensional comparison and dynamic threshold judgment, it can efficiently identify abnormal states that deviate from the normal evolution path, and generate spatiotemporal anomaly vectors containing anomaly type, spatiotemporal coordinates and degree of deviation, which significantly improves the accuracy and interpretability of anomaly detection. At the same time, it supports dynamic updating and adaptive optimization of the knowledge base, enhancing the intelligence level and long-term applicability of the system.

[0030] The construction of the spatiotemporal rule knowledge base in S41 includes: S411, Historical data preprocessing and spatiotemporal sequence generation: Perform spatiotemporal alignment and normalization on historical status data, divide spatiotemporal units according to Beidou grid codes, and generate a spatiotemporal sequence dataset of infrastructure status evolution , where a single space-time sequence is represented as: ; in, For the Space-time units in time The characteristic matrix of , is the total number of space-time units, is the number of time steps; S412, Multidimensional Association Rule Mining: Use the improved Apriori algorithm to extract spatiotemporal association rules in the form of triples , including: Support calculation: ; Confidence calculation: ; in, As the trigger condition, For space-time constraints, As a result of evolution; S413, Spatiotemporal Tensor Decomposition and Dimensionality Reduction Storage: Constructing a Four-Dimensional Spatiotemporal Tensor , the dimensions represent the spatial grid, time window, state features, and association rules respectively, and the dimensionality reduction is performed through high-order singular value decomposition (HOSVD), which is expressed as: ; in, is a rule tensor, which represents the core rule structure after dimensionality reduction. , , , is a factor matrix, corresponding to the spatial, temporal, feature and rule dimensions, for - modular product; S414, Dynamic update of rule knowledge base: Design an incremental tensor decomposition algorithm, when new data is added When , the update rule tensor is expressed as: ; in, is the updated regular tensor, For the Transpose of the dimension factor matrix; Through the above content, a spatiotemporal rule knowledge base based on historical status data is constructed. Combined with multi-dimensional association rule mining and spatiotemporal tensor decomposition technology, accurate modeling and rule extraction of urban infrastructure state evolution are achieved. Through the dynamic update mechanism, new data can be learned online and rule weights can be optimized, which significantly improves the adaptability and timeliness of the knowledge base. At the same time, it supports interpretable rule triplets (trigger conditions, spatiotemporal constraints, evolution results), providing high-quality knowledge support for anomaly detection and fault diagnosis, and enhancing the intelligence level and long-term applicability of the system.

[0031] Spatiotemporal anomaly detection and vector generation in S42 include: S421, spatiotemporal similarity calculation: Use spatiotemporal weighted cosine similarity algorithm to calculate the current feature matrix and the historical feature matrix in the knowledge base Similarity , expressed as: ; in, is the spatiotemporal weight, is the attenuation coefficient, is the current feature matrix The feature vector of the grid, The first History Model The feature vector of the grid, is the total number of grids; S422, dynamic threshold judgment: based on the statistical characteristics of the historical similarity distribution in the knowledge base, dynamically set the abnormality judgment threshold , expressed as: ; in, is the mean historical similarity, is the standard deviation of historical similarity, is the sensitivity adjustment factor (take ); S423, Exception vector generation: for satisfying The abnormal state generates a spatiotemporal abnormal vector , expressed as: ; in, Encode the exception type, is the Beidou grid code of the abnormal location, is the abnormal timestamp, is the deviation (defined as ); S424, Anomaly Propagation Path Labeling: Based on Spatiotemporal Graph Structure and the causal rules in the knowledge base, and uses the random walk algorithm to predict the abnormal propagation path , expressed as: ; in, is the causal rule in the knowledge base, Taking the outlier point as the starting point, the propagation path is selected according to the edge weight and causal rule probability; Causal rules in knowledge base The settings are based on historical status data, including: Historical state data preprocessing: Perform spatiotemporal alignment based on historical state data to ensure the spatiotemporal consistency of the data, and remove noise in the data through normalization processing; Multidimensional association rule mining: mining the causal relationship between different urban infrastructure states from historical data and constructing causal rules , expressed as: ; in, As the trigger condition, As a result of evolution; ; ; in, Indicates the conditions and results The number of simultaneous occurrences, is the size of the dataset, It is a condition The number of occurrences of Decomposition and dimensionality reduction of space-time tensors: transforming causal rules Stored in an efficient spatiotemporal rule knowledge base for easy subsequent retrieval and application, and reduce the dimensionality of high-dimensional data through spatiotemporal tensor decomposition to extract the most important causal rules; Incremental rule updating: When new data is added, the causal rules in the knowledge base are updated using an incremental tensor decomposition method. ; Through the above content, it is possible to accurately identify abnormal states that deviate from the normal evolution path, and generate a spatiotemporal anomaly vector containing the anomaly type, spatiotemporal coordinates and degree of deviation, which significantly improves the accuracy and interpretability of anomaly detection. At the same time, combined with a set of causal rules to mark potentially associated anomaly propagation paths, it achieves full process coverage from anomaly detection to risk warning, providing efficient and reliable technical support for the intelligent operation and maintenance of urban infrastructure.

[0032] The spatiotemporal memory-enhanced prediction and warning interval generation in S5 includes: S51, Spatiotemporal Memory Pool Construction: Constructing a spatiotemporal memory pool based on the historical infrastructure status feature matrix , expressed as: ; in, It is a space-time memory pool The historical state feature matrix in ; S52, Gated spatiotemporal attention mechanism: Calculate the current feature matrix Attention weights with historical feature matrix in memory pool , expressed as: ; in, For the space-time memory pool A historical feature matrix, is the attention weight matrix, is the attention vector, is the spatial attenuation coefficient, To calculate the current Beidou grid code With Beidou grid code of historical feature matrix The spatial distance between is the total number of historical state matrices stored in the spatiotemporal memory pool, For the space-time memory pool A historical feature matrix, To calculate the current Beidou grid code With Beidou grid code of historical feature matrix The spatial distance between S53, prediction trajectory generation: generating future time windows through gated recurrent units (GRU) The predicted trajectory , expressed as: ; in, is the predicted infrastructure status feature matrix, representing the future time window The state of infrastructure in is the length of the prediction time window, that is, how many time steps are predicted in the future. is the time step; S54, spatiotemporal warning interval calculation: Generate spatiotemporal warning interval based on the confidence of the predicted trajectory , expressed as: ; in, is the confidence level coefficient, is the standard deviation of the prediction error, To calculate the historical state Beidou grid code With reference grid code The spatial distance between Beidou grid code for calculating current infrastructure status With reference grid code The spatial distance between Through the above content, combined with the historical infrastructure status feature matrix and current status data, the gated spatiotemporal attention mechanism is used to weight the historical patterns to generate accurate prediction trajectories and reliable warning intervals. It can make full use of the historical data in the spatiotemporal memory pool, improve the model's adaptability to complex spatiotemporal changes, and support dynamic adjustment and real-time prediction. Through precise spatiotemporal correlation calculation, it can not only predict the future state of infrastructure, but also generate spatiotemporal warning intervals based on the confidence of the prediction, identify potential anomalies or failures in advance, and ensure the intelligence and efficiency of urban infrastructure management. In addition, the dynamic update mechanism of the model enables the system to self-optimize as new data is continuously added, further improving the prediction accuracy and long-term applicability.

[0033] The anomaly and warning overlay analysis and causal reasoning optimization in S6 include: S61, anomaly and warning superposition analysis: the spatiotemporal anomaly vector and the spatiotemporal warning interval Conduct spatiotemporal superposition analysis and calculate the overlap coefficient of anomalies within the warning interval , expressed as: ; in, is the abnormal deviation, indicating the magnitude or degree of abnormality. The Beidou grid code where the abnormal event occurred indicates the spatial location of the abnormal event. is the Beidou grid code of the warning interval, indicating the spatial location of the warning interval; S62, Dynamic adjustment of the search space of the causal inference engine: based on the overlap coefficient Dynamically adjust the search space of the causal inference engine, including: Spatial constraint: limit the search range to the Beidou grid code where the abnormal event occurred Center, radius The space-time region, where (Take 10 times the smallest unit of the Beidou grid code); Rule filtering: only retain the subset of causal rules related to the current exception type Type; S63, fault diagnosis report generation: based on the output of the causal reasoning engine, a fault diagnosis report with time and space coordinates is generated, including the Beidou grid code of the root cause event (the spatial location of the fault or anomaly), the timestamp of the root cause event (the specific time when the fault occurred), and the diagnosis confidence (the reliability of the fault diagnosis); Diagnostic confidence It is expressed as: ; in, is the weight of the causal rule; S64, maintenance decision instruction generation: Generate maintenance decision instructions with Beidou grid code positioning according to the fault diagnosis report, including priority sorting (arranging root cause events in descending order of confidence) and path planning (generating the optimal maintenance path based on the spatial distribution of Beidou grid code); Through the above content, the accuracy and real-time performance of fault diagnosis are effectively improved. By calculating the overlap coefficient between the anomaly and the warning interval, it is possible to accurately evaluate whether the abnormal event is within the warning range, and weight the impact of the anomaly according to the spatial distance attenuation characteristics, thereby reducing the risk of false alarms and missed alarms. At the same time, causal reasoning optimization ensures that the causal reasoning engine can focus on the most relevant areas and rules through rule screening based on anomaly types and dynamic adjustment of search space, thereby improving the efficiency of fault location and cause analysis, enhancing the system's adaptability to complex scenarios, and improving the accuracy and response speed of infrastructure fault warnings and maintenance decisions, thereby achieving more intelligent and efficient urban infrastructure management.

[0034] The present invention covers any substitution, modification, equivalent method and scheme made on the essence and scope of the present invention. In order to make the public have a thorough understanding of the present invention, specific details are described in detail in the following preferred embodiments of the present invention, but those skilled in the art can fully understand the present invention without the description of these details. In addition, in order to avoid unnecessary confusion about the essence of the present invention, well-known methods, processes, procedures, components and circuits are not described in detail.

[0035] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A digital twin management method for urban infrastructure supported by BeiDou space-time base, characterized in that: The following steps are involved: S1, spatiotemporal reference grid construction and code allocation: Based on the satellite-based ground-based augmented data of the BeiDou-3 system, a centimeter-level spatiotemporal reference grid covering the target area is generated, and a BeiDou grid code with spatiotemporal uniqueness is allocated to each urban infrastructure entity; S2, multimodal data collection and spatiotemporal alignment processing: real-time status data is collected through multimodal sensors deployed on urban infrastructure, and the real-time status data is spatiotemporally aligned with the Beidou grid code to generate a perception data stream with spatiotemporal labels; S3, feature extraction: construct a digital twin evolution model based on spatiotemporal graph neural network, take the spatiotemporal labeled perception data stream input as input data, and generate the infrastructure status feature matrix with spatiotemporal correlation characteristics; S4, establishment of spatiotemporal rule knowledge base and anomaly detection: establish a spatiotemporal rule knowledge base for the evolution of urban infrastructure status, detect abnormal status of urban infrastructure and generate spatiotemporal anomaly vectors by comparing the infrastructure status feature matrix with the historical evolution pattern in the spatiotemporal rule knowledge base; S5, spatiotemporal memory enhanced prediction and warning interval generation: Construct a spatiotemporal memory enhanced prediction model, based on the current infrastructure status feature matrix and the spatiotemporal distribution characteristics of the Beidou grid code, predict the evolution trajectory of the infrastructure status within the future set time window, and generate the spatiotemporal warning interval of the prediction deviation threshold; S6, anomaly and warning superposition analysis and causal reasoning optimization: superimpose the spatiotemporal anomaly vector and the spatiotemporal warning interval for analysis, dynamically adjust the search space of the causal reasoning engine, output a fault diagnosis report with spatiotemporal coordinates, and generate maintenance decision instructions with Beidou grid code positioning based on the fault diagnosis report.

2. The urban infrastructure digital twin management method supported by the Beidou space-time base according to claim 1 is characterized in that: The construction and coding allocation of the spatiotemporal reference grid in S1 include: S11, satellite-based and ground-based augmentation data acquisition: obtain the original observation data of the target area through the satellite-based augmentation service of the BeiDou-3 system and the ground-based augmentation station network; S12, Spatiotemporal reference grid division and hierarchical design: Design a multi-level spatiotemporal grid division scheme based on the spatial scope and application requirements of the target area, including: Spatial level: Divide the target area into regular grid cells with adjustable side lengths, and the minimum cell size is 0.5m×0.5m; Time level: Time slices are defined based on Beidou timing signals, with a minimum time resolution of 1 second; S13, grid code generation and space-time binding: Beidou grid coding algorithm is used to generate a four-dimensional code including longitude, latitude, elevation and time for each grid unit. The four-dimensional code structure meets the following conditions: ； in, , , , are the longitude, latitude, elevation and time of the grid center point, , , , is the coordinate of the origin of the space-time reference, , , , is the grid spatial and temporal resolution, is the Beidou standard hash function, is the Beidou grid code; S14, Infrastructure entity coding mapping: The spatial coordinates of urban infrastructure entities are mapped With timing status Mapped to the corresponding grid code through a dynamic interpolation algorithm.

3. The urban infrastructure digital twin management method supported by the Beidou space-time base according to claim 1 is characterized in that: The multimodal data acquisition and spatiotemporal alignment processing in S2 includes: S21, Multimodal sensor deployment and data collection: Deploy multimodal sensor networks on urban infrastructure, including structural health monitoring sensors, environmental sensors, and operating status sensors, to collect multimodal data in real time, including structural health data, environmental data, and operating status data, among which; The structural health data includes strain, vibration, and displacement; The environmental data include temperature, humidity, and wind speed; The operating status data includes load, flow, and pressure; S22, data preprocessing: preprocessing the collected multimodal data, including outlier removal, missing value filling, and noise filtering; S23, time-space benchmark unification and data alignment: the pre-processed multi-modal data is time-space aligned with the Beidou grid code, specifically including: Spatial alignment: Map the data in the sensor's local coordinate system to the Beidou grid coordinate system through the coordinate transformation matrix; Time alignment: Synchronize and calibrate sensor timestamps based on Beidou timing signals; S24, spatiotemporal label generation and data stream encapsulation: Add spatiotemporal labels to the aligned multimodal data to generate standardized perception data streams.

4. The urban infrastructure digital twin management method supported by the Beidou space-time base according to claim 3 is characterized in that: The data preprocessing in S22 includes: S221, outlier removal: based on The criterion removes abnormal data from multimodal data; S222, missing value filling: using spatiotemporal interpolation to fill missing data in multimodal data; S223, Noise filtering: Remove high-frequency noise in multimodal data through wavelet transform.

5. The urban infrastructure digital twin management method supported by the Beidou space-time base according to claim 1 is characterized in that: The feature extraction in S3 includes: S31, construction of space-time graph structure: using Beidou grid codes as nodes and space-time associations as edges to construct a dynamic space-time graph structure , where the node feature is expressed as , For the The perception data of a grid, is the embedding vector of Beidou grid code, and the edge weight is expressed as , is the spatial attenuation factor; S32, spatial graph convolution calculation: using gated graph convolution layer to extract spatial features; S33, time-gated recurrent unit: models the timing characteristics of each node; S34, Spatiotemporal Attention Mechanism: Dynamically Calculating Node Importance Weights ; S35, feature matrix generation: Sort the output of the spatiotemporal graph neural network by Beidou grid code to generate the infrastructure status feature matrix ; S36, model training: end-to-end training is performed using a hybrid loss function.

6. The urban infrastructure digital twin management method supported by the Beidou space-time base according to claim 5 is characterized in that: The establishment of the spatiotemporal rule knowledge base and anomaly detection in S4 include: S41, construction of spatiotemporal rule knowledge base: extract the spatiotemporal pattern of the evolution of urban infrastructure status based on historical status data, use multidimensional correlation analysis technology to abstract the spatiotemporal relationship of infrastructure status changes into rule triples, and use spatiotemporal tensor decomposition technology to reduce the dimensionality of high-dimensional relationships and store them, forming a spatiotemporal rule knowledge base that includes normal state evolution paths and failure modes; S42, spatiotemporal anomaly detection and vector generation: The generated infrastructure status feature matrix is ​​compared with the historical evolution pattern in the spatiotemporal rule knowledge base in multiple dimensions. Through spatiotemporal similarity calculation and dynamic threshold judgment, the abnormal state that deviates from the normal evolution path is identified. For the detected abnormal state, a spatiotemporal anomaly vector including the anomaly type, spatiotemporal coordinates and degree of deviation is generated, and the potentially associated abnormal propagation path is marked.

7. The urban infrastructure digital twin management method supported by the Beidou space-time base according to claim 6 is characterized in that: The construction of the spatiotemporal rule knowledge base in S41 includes: S411, Historical data preprocessing and spatiotemporal sequence generation: Perform spatiotemporal alignment and normalization on historical status data, divide spatiotemporal units according to Beidou grid codes, and generate a spatiotemporal sequence dataset of infrastructure status evolution , where a single space-time sequence is represented as: ； in, For the Space-time units in time The characteristic matrix of , is the total number of space-time units, is the number of time steps; S412, Multidimensional Association Rule Mining: Use the improved Apriori algorithm to extract spatiotemporal association rules in the form of triples ; S413, Spatiotemporal Tensor Decomposition and Dimensionality Reduction Storage: Constructing a Four-Dimensional Spatiotemporal Tensor , the dimensions represent the spatial grid, time window, state features, and association rules, respectively, and the dimensionality reduction is performed through high-order singular value decomposition; S414, Dynamic update of rule knowledge base: Design an incremental tensor decomposition algorithm, when new data is added , update the rule tensor.

8. The urban infrastructure digital twin management method supported by the Beidou space-time base according to claim 7 is characterized in that: The spatiotemporal anomaly detection and vector generation in S42 include: S421, spatiotemporal similarity calculation: Use spatiotemporal weighted cosine similarity algorithm to calculate the current feature matrix and the historical feature matrix in the knowledge base Similarity ; S422, dynamic threshold judgment: based on the statistical characteristics of the historical similarity distribution in the knowledge base, dynamically set the abnormality judgment threshold ; S423, exception vector generation: for satisfying The abnormal state generates a spatiotemporal abnormal vector ; S424, Anomaly Propagation Path Labeling: Based on Spatiotemporal Graph Structure and the causal rules in the knowledge base, and uses the random walk algorithm to predict the abnormal propagation path .

9. The urban infrastructure digital twin management method supported by the Beidou space-time base according to claim 8 is characterized in that: The generation of the spatiotemporal memory enhanced prediction and warning interval in S5 includes: S51, Spatiotemporal Memory Pool Construction: Constructing a spatiotemporal memory pool based on the historical infrastructure status feature matrix ; S52, Gated spatiotemporal attention mechanism: Calculate the current feature matrix Attention weights with historical feature matrix in memory pool ; S53, prediction trajectory generation: generating future time windows through gated recurrent units The predicted trajectory ; S54, spatiotemporal warning interval calculation: Generate spatiotemporal warning interval based on the confidence of the predicted trajectory .

10. The urban infrastructure digital twin management method supported by the Beidou space-time base according to claim 9 is characterized in that: The anomaly and warning superposition analysis and causal reasoning optimization in S6 include: S61, anomaly and warning superposition analysis: the spatiotemporal anomaly vector and the spatiotemporal warning interval are combined Perform spatiotemporal superposition analysis and calculate the overlap coefficient of anomalies within the warning interval ; S62, Dynamic adjustment of the search space of the causal inference engine: based on the overlap coefficient Dynamically adjust the search space of the causal inference engine, including: Spatial constraint: limit the search range to the Beidou grid code where the abnormal event occurred Center, radius The space-time region, where ; Rule filtering: only retain the subset of causal rules related to the current exception type Type; S63, fault diagnosis report generation: based on the output of the causal reasoning engine, a fault diagnosis report with time and space coordinates is generated, including the Beidou grid code of the root cause event, the timestamp of the root cause event, and the diagnosis confidence; S64, maintenance decision instruction generation: generating maintenance decision instructions with Beidou grid code positioning according to the fault diagnosis report, including priority sorting and path planning.

Images