A big data-based smart city planning decision support system

Abstract

The application discloses a kind of wisdom city planning decision support systems based on big data, it is related to data processing system technical field, the system includes: data acquisition module, space-time fusion module, load evaluation module, forecast early warning module and decision feedback module;The integrated architecture of data acquisition, space-time fusion, load evaluation, forecast early warning, decision feedback is built in the application, the static basic data and dynamic flow data of municipal infrastructure are completed standardization processing, to realize multi-source data weighted fusion with the space grid and unified time granularity construction space-time grid unit, respectively for point independent facilities, regional covering facilities carry out facility level and grid level load rate calculation, overload propagation path is combed in combination with facility network topology relationship, construct hierarchical clear, global coverage bearing capacity evaluation system, complete realization data processing, state determination, risk traceability The whole process landing, provide accurate operation state basis for wisdom city infrastructure planning.

Description

Technical Field

[0001] This invention relates to the field of data processing system technology, specifically to a smart city planning decision support system based on big data. Background Technology

[0002] With the deepening of smart city construction, the service scope and operational load of municipal infrastructure such as water, electricity, gas, transportation, and communication continue to grow. The refinement and intelligence of infrastructure carrying capacity management, operation and maintenance scheduling, and planning layout have become key directions for urban governance. Modern urban development has placed higher demands on the efficient management of public service facilities. The traditional extensive facility management and planning model can no longer meet the current needs of large-scale and complex infrastructure operation and management in cities.

[0003] However, existing smart city infrastructure planning and operation and maintenance technologies have some shortcomings. Static basic data and dynamic streaming data are isolated from each other, and a unified spatiotemporal fusion processing mechanism has not been formed, resulting in low data utilization. Load assessment only focuses on single-point facility calculations and lacks a grid-level global carrying capacity status determination system, which cannot reflect the overall operation of regional facilities. Predictive models do not take into account external influencing factors, resulting in insufficient accuracy of early warnings. Furthermore, they lack feedback optimization after decision execution, and the predictive models cannot be dynamically iterated, making it difficult to support full-process, scientific smart city planning decisions and meet the actual needs of urban infrastructure carrying capacity assessment, risk warning, and planning decisions. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of existing technologies and provide a smart city planning decision support system based on big data. This invention establishes an integrated architecture for data collection, spatiotemporal fusion, load assessment, prediction and early warning, and decision feedback. It standardizes the processing of static basic data and dynamic streaming data of municipal infrastructure, constructs spatiotemporal grid units with spatial grids and unified time granularity to achieve weighted fusion of multi-source data, and performs facility-level and grid-level load rate calculations for point-based independent facilities and regional coverage facilities. It also analyzes overload propagation paths by combining facility network topology relationships, and constructs a hierarchical and fully covered carrying capacity assessment system. This fully realizes the implementation of the entire process of data processing, status determination, and risk tracing, providing accurate operational status basis for smart city infrastructure planning.

[0005] To solve the above-mentioned technical problems, the present invention provides the following technical solution: a smart city planning decision support system based on big data, the system comprising: Data acquisition module: Collects static basic data and dynamic stream data through the acquisition interface, and adds timestamps and spatial location identifiers after preprocessing, thereby forming a standardized static parameter set and dynamic stream dataset, and then transmits them to the spatiotemporal fusion module; Spatiotemporal fusion module: Receives static parameter set and dynamic flow dataset, constructs spatial grid system, unifies time granularity for dynamic flow dataset, assigns static parameter set and dynamic flow dataset to corresponding spatial grid, performs weighted fusion of dynamic flow indicators of similar infrastructure within the same spatiotemporal grid, generates standard spatiotemporal grid coordinates, fused dynamic flow data matrix and grid static design capacity mapping table, and transmits them to load assessment module. Load assessment module: Receives standard spatiotemporal grid coordinates, fused dynamic flow data matrix and grid static design capacity mapping table, calculates facility-level load rate and grid-level load rate, classifies bearing capacity status and labels it, generates bearing capacity heat map vector data and real-time load rate dataset, and transmits the real-time load rate dataset to the prediction and early warning module. Prediction and early warning module: Receives real-time load rate dataset, constructs a multivariate time series prediction model, trains it with external features, predicts the load rate for future periods, triggers an early warning, determines the overload propagation path, generates a load rate prediction curve and a list of early warning events, and transmits the list of early warning events to the decision feedback module. Decision Feedback Module: Receives a list of early warning events, matches decision suggestions with the corresponding time scale, and feeds back the decision execution data and load rate change data to the multivariate time series prediction model for closed-loop optimization.

[0006] Furthermore, the data acquisition module includes a data direct connection interface and a message queue interface. The data direct connection interface collects static basic data such as municipal pipeline network GIS data, facility rated capacity and design parameters, and urban planning carrying capacity standards. The message queue interface collects dynamic streaming data such as water, electricity, and gas consumption, traffic flow at traffic sections, sewage treatment plant operation data, and communication base station service data. The collected static basic data and dynamic streaming data are then preprocessed, sequentially performing missing value marking, linear interpolation, and outlier removal. Timestamps and spatial location identifiers are uniformly added according to the data acquisition time and facility geographic information, resulting in a standardized static parameter set and dynamic streaming dataset.

[0007] Furthermore, in the spatiotemporal fusion module, a spatial grid system is constructed based on the city's geographical boundaries. This spatial grid system uses rectangular grids with a size of 500m × 500m. A reference time granularity of 15 minutes or 60 minutes is then set. Minute-level data from the dynamic flow dataset is aggregated to the reference time granularity using averages. Daily-level data from the dynamic flow dataset is decomposed into hourly-level data using historical patterns, thus unifying the time scale of the dynamic flow dataset. Next, the static parameter set and facility-related data from the dynamic flow dataset are matched to the corresponding spatial grids through spatial affiliation, thus unifying the regional... The corresponding data is matched to the corresponding spatial grid according to the area weight. Then, based on the historical data accuracy of each data source, the corresponding data fusion weight is determined. For dynamic flow data that has been matched with spatial grids and completed with time uniformity, spatiotemporal grid units are formed by combining a single spatial grid with a single reference time granularity. The fused flow value of a single spatiotemporal grid is calculated using a multi-source data weighted fusion formula. Then, standard spatiotemporal grid coordinates are generated based on the reference time granularity and spatial grid ID. All fused flow values ​​of single spatiotemporal grids are integrated to generate a fused dynamic flow data matrix. A grid static design capacity mapping table is generated based on the static parameter set.

[0008] Furthermore, in the spatiotemporal fusion module, the weighted fusion formula for multi-source data is: ,in, For single-temporal grid fused flow values, The fusion weight for the i-th type of data source is obtained by normalizing the historical data accuracy of each data source. , Let be the historical data accuracy of the i-th type of data source. The original flow value of the i-th type of data source within the target spatiotemporal grid cell is taken from the standardized dynamic flow dataset, and n is the total number of data sources of the same type, which is determined by the type of data source accessed.

[0009] Furthermore, in the load assessment module, based on the fused dynamic flow data matrix and the grid static design capacity mapping table, the facility-level load rate of point facilities in the static parameter set is calculated. The facility-level load rate is obtained by the ratio of the single-temporal grid fused flow value of the point facility to the design capacity. Point facilities include independent point infrastructure such as substations, water plants, pumping stations, and communication base stations. Then, based on the single-temporal grid fused flow value and the total design capacity of the grid facilities, the grid-level load rate of regional facilities is calculated using the grid-level load rate calculation formula. Regional facilities are regional coverage infrastructure such as municipal pipe networks, drainage pipe networks, and communication coverage areas. Then, based on the grid-level load rate, the carrying capacity status is divided and labeled, and finally, carrying capacity heat map vector data and real-time load rate dataset are generated. The real-time load rate dataset includes facility ID, grid ID, timestamp, facility-level load rate, grid-level load rate, and status label. The carrying capacity heat map vector data is used for visualization, and the update frequency of the status label is consistent with the reference time granularity.

[0010] Furthermore, in the load assessment module, the formula for calculating the grid-level load rate is: ,in, For grid-level load factor, when When, it is in a normal state, when When, it is in a state of attention. At that time, it is in a state of alert. At that time, it is in an overload state. This represents the sum of the spatiotemporal fusion flow values ​​of the k-th facility of the same type within the grid. The sum of the design capacities of the k-th facility of the same type within the grid. For single-facility, single-temporal-grid fusion of flow values, The design capacity for a single facility is obtained based on the grid static design capacity mapping table, where m is the total number of facilities of the same type within the grid.

[0011] Furthermore, in the prediction and early warning module, grid-level load rates within the real-time load rate dataset are continuously extracted according to timestamps to form historical time-series data. A multivariate time series prediction model is constructed based on this historical time-series data, consisting of an input layer, a time-series feature extraction layer, and an output layer. The input layer receives historical time-series data and quantized external features; the time-series feature extraction layer extracts time-series correlation features; and the output layer outputs the predicted load rate. The historical time-series data and quantized external features are used as training samples for training, fixing the basic structure of the multivariate time series prediction model. The quantized external features are obtained through external feature standardization. The external features include… Weather information, holiday information, and information on major events are collected. Historical time-series data of the target grid and the quantified external features of the current moment are input into a multivariate time series prediction model. The predicted load rate of the target grid for future periods is calculated using a load rate time-series prediction fitting formula. Based on the predicted load rate, corresponding level warnings are triggered. The overload propagation path is determined based on the upstream and downstream relationships of the power grid and the flow direction of the drainage network. The affected adjacent facilities and grids are identified, and a load rate prediction curve and a list of warning events are generated. The list of warning events includes the warning level, overloaded grid ID, overloaded facility ID, overload duration, warning trigger time, and overload propagation path information.

[0012] Furthermore, in the prediction and early warning module, the load rate time-series prediction fitting formula is: ,in, For the predicted load rate of the target grid over the next k time slices, when At that time, a yellow alert was triggered. At that time, an orange alert was triggered. At that time, a red alert was triggered. The current grid-level load rate of the target grid. The grid-level load rate for the target grid over m historical time slices is extracted from the real-time load rate dataset by timestamp. The fitting coefficients are determined from the training results of the multivariate time series prediction model. is the quantized value of the external feature at the current moment, and p is the input length of the historical time series data, which is determined by the training configuration of the multivariate time series prediction model.

[0013] Furthermore, in the decision feedback module, decision recommendations are matched based on the duration of overload and the frequency of warnings in the early warning events. If the overload duration is less than 2 hours and there is only one warning, a short-term off-peak scheduling recommendation is matched. If the overload duration is between 2 and 24 hours and there are 2-3 cumulative warnings, a medium-term temporary capacity expansion recommendation is matched. If the same facility or grid has ≥4 cumulative warnings within 7 days, a long-term capacity expansion recommendation is matched. The short-term off-peak scheduling recommendations include adjusting industrial power consumption periods, guiding traffic flow, and balancing communication loads. The medium-term temporary capacity expansion recommendations include activating backup units, dispatching mobile pump trucks, and temporary communication... The long-term expansion recommendations for resource leasing include new facility construction, pipeline expansion, and equipment upgrades. Decision execution data includes the progress of implementation, actual resource input, and any anomalies during implementation. Load rate change data includes facility-level and grid-level load rates for each baseline time slice after implementation, compared with load rate data before implementation. The decision execution data and load rate change data are then used as new training samples and fed back into the multivariate time series prediction model. Retraining is performed with a 7-day training cycle to optimize the fitting coefficients, the length of historical time series data input, correct prediction bias, and complete closed-loop optimization.

[0014] Compared with existing technologies, this smart city planning decision support system based on big data has the following beneficial effects: I. This invention establishes an integrated architecture encompassing data acquisition, spatiotemporal fusion, load assessment, prediction and early warning, and decision feedback. It standardizes the processing of static basic data and dynamic streaming data of municipal infrastructure. By constructing spatiotemporal grid units with spatial grids and unified time granularity, it achieves weighted fusion of multi-source data. It performs facility-level and grid-level load rate calculations for point-based independent facilities and regional coverage facilities, respectively. Combining the facility network topology, it identifies overload propagation paths and constructs a hierarchical and comprehensive carrying capacity assessment system. This fully realizes the implementation of the entire process of data processing, status determination, and risk tracing, providing accurate operational status data for smart city infrastructure planning.

[0015] Second, this invention constructs a multivariate time series prediction model by combining historical time series data with external features, enabling accurate projection of the future load status of infrastructure. Based on the prediction results, it triggers early warnings at different levels and matches decision recommendations with corresponding time scales. The decision execution data and load rate change data are fed back to the prediction model for retraining, forming a closed-loop operation mechanism of prediction, early warning, decision-making, and optimization. This continuously corrects the prediction bias of the prediction model, ensuring that early warning judgments and decision recommendations align with the actual operating patterns of infrastructure, and providing scientific and feasible decision support for the short-term scheduling, medium-term capacity expansion, and long-term capacity expansion of urban infrastructure.

[0016] Other advantages, objectives and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination or study, or may be learned from the practice of the invention. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without any creative effort.

[0018] Figure 1 A flowchart of a smart city planning decision support system based on big data; Figure 2 This is a framework diagram of a smart city planning decision support system based on big data. Figure 3 This is a framework diagram of the load assessment module in a smart city planning decision support system based on big data. Detailed Implementation

[0019] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structure, features and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0020] Example: In the integrated management and control scenario of municipal infrastructure in the central urban area of ​​a prefecture-level city, the area covers residential areas, commercial complexes, industrial parks, urban main roads and core corridors of municipal pipelines. On a daily basis, it is necessary to manage and control six major categories of infrastructure: water, electricity, gas, transportation, communication and drainage. Through the coordinated operation of five modules, namely data collection, spatiotemporal fusion, load assessment, prediction and early warning and decision feedback, the system can achieve full-area perception, accurate assessment, intelligent early warning and scientific decision-making of the infrastructure operation status in the area.

[0021] First, multi-source data is collected and standardized using the data acquisition module. Then, data related to the entire infrastructure is acquired through two types of acquisition interfaces: direct data connection interfaces and message queue interfaces. The direct data connection interfaces collect static basic data such as municipal pipeline GIS data, facility rated capacity and design parameters, and urban planning carrying capacity standards for the central urban area. This type of data represents the inherent attributes of the infrastructure and is the core basis for subsequent carrying capacity assessments. The message queue interfaces collect dynamic streaming data such as water, electricity, and gas consumption, traffic flow at traffic sections, sewage treatment plant operation data, and communication base station service data. This type of data can reflect the actual operating load of the infrastructure in real time. The preprocessing operations of missing value marking, linear interpolation, and outlier removal are sequentially performed on all collected static basic data and dynamic streaming data to effectively solve problems such as missing values ​​and anomalies in the original data and ensure data quality. According to the collection time and facility geographic information of various data types, timestamps and spatial location identifiers are uniformly added to all data, giving each data point clear time and spatial attributes. After processing, standardized static parameter sets and dynamic streaming datasets are formed, completely eliminating the format and dimensional differences of multi-source heterogeneous data, laying a stable data foundation for subsequent spatiotemporal fusion processing. The static parameter sets and dynamic streaming datasets are then transmitted to the spatiotemporal fusion module, such as... Figure 1 As shown.

[0022] Upon receiving the standardized static parameter set and dynamic stream dataset, the spatiotemporal fusion module immediately begins spatiotemporal integrated data processing. Based on the administrative geographical boundaries of the central urban area, a unified spatial grid system is constructed. This system uses rectangular grids with a uniform grid size of 500m × 500m, achieving a uniform division of the entire central urban area and ensuring that infrastructure operation data can be accurately matched to specific spatial units. Subsequently, a baseline time granularity of 15 minutes or 60 minutes is set. The minute-level high-frequency data in the dynamic stream dataset is aggregated to the baseline time granularity using an average method, avoiding computational pressure caused by high-frequency data redundancy. The daily-level low-frequency data in the dynamic stream dataset is decomposed into hourly data using a historical model, supplementing the time resolution of the low-frequency data and completing the time scale unification of the dynamic stream dataset, allowing all dynamic... The data possesses a consistent temporal statistical dimension. Then, the facility-related data in the static parameter set and dynamic flow dataset are matched to corresponding spatial grids through spatial affiliation. Regional data is matched to corresponding spatial grids according to area weight, achieving precise spatial binding of static and dynamic data. Next, based on the historical data accuracy of each data source, corresponding data fusion weights are determined to ensure that high-accuracy data sources occupy a reasonable proportion in the fusion results. For dynamic flow data that has been matched to spatial grids and achieved temporal uniformity, spatiotemporal grid units are formed by combining a single spatial grid with a single reference time granularity. This ensures that each computational unit possesses both spatial and temporal attributes. A multi-source data weighted fusion formula is used to calculate the fused flow value of a single spatiotemporal grid, completing the precise fusion of multi-source dynamic flow data. The multi-source data weighted fusion formula is as follows: ,in, For single-temporal grid fused flow values, The fusion weight for the i-th type of data source is obtained by normalizing the historical data accuracy of each data source. , Let be the historical data accuracy of the i-th type of data source. The original flow value of the i-th type of data source within the target spatiotemporal grid cell is taken from a standardized dynamic flow dataset, where n is the total number of data sources of the same type, determined by the type of data source accessed. Then, standard spatiotemporal grid coordinates are generated based on the reference time granularity and spatial grid ID. All single spatiotemporal grid fused flow values ​​are integrated to generate a fused dynamic flow data matrix. A grid static design capacity mapping table is generated based on the static parameter set. The standard spatiotemporal grid coordinates, the fused dynamic flow data matrix, and the grid static design capacity mapping table are then transmitted to the load assessment module. Figure 2 As shown, it provides complete spatiotemporal fusion data support for load assessment.

[0023] After receiving the standard spatiotemporal grid coordinates, the fused dynamic flow data matrix, and the grid static design capacity mapping table, the load assessment module performs load rate calculation and capacity status determination for all infrastructure in the region. Based on the fused dynamic flow data matrix and the grid static design capacity mapping table, it calculates the facility-level load rate for point-like facilities with concentrated static parameters. The facility-level load rate is obtained by the ratio of the fused flow value of the single spatiotemporal grid for the point-like facility to its design capacity. The point-like facilities include independent point-like infrastructure such as substations, water plants, pumping stations, and communication base stations, achieving accurate calculation of the operating load of independent point-like facilities. Then, based on the fused flow value of the single spatiotemporal grid and the total design capacity of the grid facilities, the grid-level load rate of regional facilities is calculated using the grid-level load rate calculation formula. The grid-level load rate calculation formula is as follows: ,in, For grid-level load factor, when When, it is in a normal state, when When, it is in a state of attention. At that time, it is in a state of alert. At this point, it is considered an overloaded state. The values ​​of 0.6, 0.85, and 0.95 are boundary values ​​set according to the regulations on the normal load rate range of various facilities in urban infrastructure planning standards, enabling each state to effectively distinguish the operational risks under different load levels. This represents the sum of the spatiotemporal fusion flow values ​​of the k-th facility of the same type within the grid. The sum of the design capacities of the k-th facility of the same type within the grid. For single-facility, single-temporal-grid fusion of flow values, The design capacity for a single facility is obtained based on a grid-based static design capacity mapping table. This involves determining the total number of similar facilities within a grid; regional facilities include municipal pipe networks, drainage pipe networks, and communication coverage areas, among other regional coverage infrastructure, and assessing the overall load of these facilities. Based on grid-level load rates, the carrying capacity status is then categorized and labeled, clearly presenting the operational risk levels of facilities and grids through a four-level status classification. Finally, carrying capacity heatmap vector data and a real-time load rate dataset are generated. The carrying capacity heatmap vector data can be directly connected to the urban operation and maintenance visualization platform for intuitive display, allowing operation and maintenance personnel to quickly locate high-load areas. The real-time load rate dataset includes facility ID, grid ID, timestamp, facility-level load rate, grid-level load rate, and status label, providing standardized input data for subsequent prediction and early warning. The update frequency of the status labels is consistent with the baseline time granularity, ensuring the real-time nature of status determination. The real-time load rate dataset is then transmitted to the prediction and early warning module, such as... Figure 3 As shown.

[0024] After receiving the real-time load rate dataset, the prediction and early warning module performs load prediction, tiered early warning, and risk tracing. It continuously extracts grid-level load rates from the real-time load rate dataset according to timestamps to form historical time-series data. Based on this historical time-series data, a multivariate time-series prediction model is constructed. This model consists of an input layer, a time-series feature extraction layer, and an output layer. The input layer receives historical time-series data and external feature quantification values; the time-series feature extraction layer extracts time-series correlation features; and the output layer outputs the predicted load rate, ensuring the multivariate time-series prediction model possesses the capability for feature extraction and prediction of time-series data. The historical time-series data and external feature quantification values ​​are then combined... The quantified values ​​of external features are used as training samples to train the multivariate time series prediction model, fixing its basic structure. The quantified values ​​of external features are obtained through external feature standardization, including weather information, holiday information, and information on major events, allowing the multivariate time series prediction model to incorporate external influencing factors and improve prediction accuracy. Subsequently, the historical time series data of the target grid and the quantified values ​​of external features at the current moment are input into the trained multivariate time series prediction model. The predicted load rate for future periods of the target grid is calculated using the load rate time series prediction fitting formula, enabling advance prediction of infrastructure load status. The load rate time series prediction fitting formula is as follows: ,in, For the predicted load rate of the target grid over the next k time slices, when At that time, a yellow alert was triggered. At that time, an orange alert was triggered. At that time, a red alert was triggered. The current grid-level load rate of the target grid. The grid-level load rate for the target grid over m historical time slices is extracted from the real-time load rate dataset by timestamp. The fitting coefficients are obtained by estimating the parameters of a multivariate time series prediction model using the least squares method, with historical time series data and external features as training samples. The goal is to minimize the mean squared error between the predicted load rate and the actual load rate. The model is then iteratively optimized. In subsequent closed-loop optimization, the multivariate time series prediction model will be retrained based on new samples, and the fitting coefficients will be adjusted accordingly. The quantized value of the external feature at the current moment. The input length of historical time series data is determined by the training configuration of the multivariate time series prediction model. Based on the predicted load rate, corresponding level warnings are triggered, achieving graded risk management through a three-level warning system. Simultaneously, the overload propagation path is determined based on the upstream and downstream relationships of the power grid and the flow direction of the drainage network, accurately locating affected adjacent facilities and grids, completing comprehensive risk tracing, and ultimately generating a load rate prediction curve and a list of warning events. The warning event list includes the warning level, overloaded grid ID, overloaded facility ID, overload duration, warning trigger time, and overload propagation path information, providing complete early warning basis for decision feedback. The warning event list is then transmitted to the decision feedback module.

[0025] After receiving the list of early warning events, the decision feedback module performs decision matching, execution feedback, and model closed-loop optimization. Based on the overload duration and frequency of the early warning events, it matches decision suggestions for the corresponding time scale. When the overload duration is less than 2 hours and it is a single early warning, a short-term peak-shifting scheduling suggestion is matched. This includes adjusting industrial power consumption periods, traffic flow guidance, and communication load balancing, suitable for temporary, small-scale high-load scenarios. When the overload duration is between 2 and 24 hours and 2-3 early warnings are triggered cumulatively, a medium-term temporary capacity expansion suggestion is matched. This includes activating backup units, dispatching mobile pump trucks, and leasing temporary communication resources, suitable for continuous, medium-scale overload scenarios. When the same facility or grid triggers ≥4 early warnings cumulatively within 7 days, a long-term expansion suggestion is matched. This includes new facility construction, pipeline expansion, and... The system is designed for upgrades and is suitable for infrastructure planning scenarios with long-term high loads. After the implementation of decision recommendations, it collects complete decision implementation data and load rate change data. The decision implementation data includes the progress of the measures, actual resources invested, and abnormal situations during the implementation process. The load rate change data includes the facility-level load rate and grid-level load rate of each baseline time slice after implementation, forming a direct comparison with the load rate data before implementation, clearly showing the effect of decision implementation. Then, the decision implementation data and load rate change data are used as new training samples and fed back into the multivariate time series prediction model. The model is retrained with a training cycle of 7 days to optimize the fitting coefficient and the length of historical time series data input, continuously correcting the prediction bias of the multivariate time series prediction model, so that the prediction results continuously conform to the actual operation law of infrastructure, completing closed-loop optimization, and realizing the long-term stable, accurate and efficient operation of the smart city planning decision support system.

[0026] In summary, by integrating spatiotemporal data processing to form a unified spatiotemporal grid unit, comprehensive carrying capacity assessments can be conducted on both point-based independent facilities and regional coverage facilities. Combined with external influencing factors, intelligent prediction and graded early warning of facility load status can be achieved, accurately locating overload risks and their spread. Based on the early warning information, appropriate scheduling and planning decision-making suggestions can be matched, and closed-loop optimization of multivariate time series prediction models can be achieved through data feedback. This provides complete and efficient technical support for the overall management, intelligent operation and maintenance, and scientific planning of urban infrastructure.

[0027] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A smart city planning decision support system based on big data, characterized in that, The system includes: Data acquisition module: Collects static basic data and dynamic stream data through the acquisition interface, and adds timestamps and spatial location identifiers after preprocessing, thereby forming a standardized static parameter set and dynamic stream dataset, and then transmits them to the spatiotemporal fusion module; Spatiotemporal fusion module: Receives static parameter set and dynamic flow dataset, constructs spatial grid system, unifies the time granularity of dynamic flow dataset, assigns static parameter set and dynamic flow dataset to corresponding spatial grid, and performs weighted fusion of dynamic flow indicators of similar infrastructure within the same spatiotemporal grid using multi-source data weighted fusion formula, generates standard spatiotemporal grid coordinates, fused dynamic flow data matrix and grid static design capacity mapping table, and transmits them to load assessment module; Load assessment module: Receives standard spatiotemporal grid coordinates, fused dynamic flow data matrix and grid static design capacity mapping table, calculates facility-level load rate and grid-level load rate, classifies bearing capacity status and labels it, generates bearing capacity heat map vector data and real-time load rate dataset, and transmits the real-time load rate dataset to the prediction and early warning module. Prediction and early warning module: Receives real-time load rate dataset, constructs a multivariate time series prediction model, trains it with external features, predicts the load rate for future periods, triggers early warning, determines the overload propagation path based on the facility network topology, generates a load rate prediction curve and a list of early warning events, and transmits the list of early warning events to the decision feedback module. Decision Feedback Module: Receives a list of early warning events, matches decision suggestions with the corresponding time scale, and feeds back the decision execution data and load rate change data to the multivariate time series prediction model for closed-loop optimization.

2. The smart city planning decision support system based on big data according to claim 1, characterized in that, The data acquisition module includes a data direct connection interface and a message queue interface. The data direct connection interface collects static basic data such as municipal pipeline GIS data, facility rated capacity and design parameters, and urban planning carrying capacity standards. The message queue interface collects dynamic streaming data such as water, electricity, and gas consumption, traffic flow at traffic sections, sewage treatment plant operation data, and communication base station service data. The collected static basic data and dynamic streaming data are then preprocessed, including missing value marking, linear interpolation, and outlier removal. Timestamps and spatial location identifiers are added uniformly according to the data acquisition time and facility geographic information, resulting in a standardized set of static parameters and a dynamic streaming dataset.

3. The smart city planning decision support system based on big data according to claim 1, characterized in that, In the spatiotemporal fusion module, a spatial grid system is constructed based on the city's geographical boundaries. The spatial grid system adopts a rectangular grid. Next, a reference time granularity of 15 minutes or 60 minutes is set. The minute-level data in the dynamic flow dataset is aggregated to the reference time granularity by averaging. The daily-level data in the dynamic flow dataset is decomposed into hour-level data according to historical patterns to achieve time scale unification of the dynamic flow dataset. Then, the static parameter set and the facility-corresponding data in the dynamic flow dataset are matched to the corresponding spatial grids through spatial affiliation. The regional corresponding data is matched to the corresponding spatial grids according to area weight. Then, the corresponding data fusion weights are determined based on the historical data accuracy of each data source. For the dynamic flow data that has been matched with spatial grids and achieved time unification, spatiotemporal grid units are formed by combining a single spatial grid with a single reference time granularity. The fused flow value of a single spatiotemporal grid is calculated using a multi-source data weighted fusion formula. Then, standard spatiotemporal grid coordinates are generated based on the reference time granularity and spatial grid ID. All fused flow values ​​of single spatiotemporal grids are integrated to generate a fused dynamic flow data matrix. A grid static design capacity mapping table is generated based on the static parameter set.

4. The smart city planning decision support system based on big data according to claim 3, characterized in that, In the spatiotemporal fusion module, the weighted fusion formula for multi-source data is: ,in, For single-temporal grid fused flow values, The fusion weight for the i-th type of data source is obtained by normalizing the historical data accuracy of each data source. , Let be the historical data accuracy of the i-th type of data source. The original flow values ​​of the i-th type of data source within the target spatiotemporal grid cell are taken from a standardized dynamic flow dataset. The total number of data sources of the same type is determined by the type of data source being accessed.

5. The smart city planning decision support system based on big data according to claim 1, characterized in that, In the load assessment module, based on the fused dynamic flow data matrix and the grid static design capacity mapping table, the facility-level load rate of point facilities in the static parameter set is calculated. The facility-level load rate is obtained by the ratio of the single spatiotemporal grid fused flow value of the point facility to the design capacity. Based on the flow rate value fused with the single spatiotemporal grid and the total design capacity of the grid facilities, the grid-level load rate of the regional facilities is calculated using the grid-level load rate calculation formula. Then, the carrying capacity status is divided and labeled based on the grid-level load rate, and finally, carrying capacity heat map vector data and real-time load rate dataset are generated. The real-time load rate dataset includes facility ID, grid ID, timestamp, facility-level load rate, grid-level load rate, and status label. The load-bearing capacity heatmap vector data is used for visualization, and the update frequency of the status labels is consistent with the base time granularity.

6. The smart city planning decision support system based on big data according to claim 5, characterized in that, In the load assessment module, the formula for calculating the grid-level load rate is: ,in, For grid-level load factor, when When, it is in a normal state, when When, it is in a state of attention. At that time, it is in a state of alert. At that time, it is in an overload state. This represents the sum of the spatiotemporal fusion flow values ​​of the k-th facility of the same type within the grid. The sum of the design capacities of the k-th facility of the same type within the grid. For single-facility, single-temporal-grid fusion of flow values, The design capacity for a single facility is obtained based on a grid-based static design capacity mapping table. This represents the total number of facilities of the same type within the grid.

7. The smart city planning decision support system based on big data according to claim 1, characterized in that, In the prediction and early warning module, grid-level load rates from the real-time load rate dataset are continuously extracted according to timestamps to form historical time-series data. A multivariate time-series prediction model is constructed based on this historical time-series data, consisting of an input layer, a time-series feature extraction layer, and an output layer. The input layer receives historical time-series data and quantized external feature values; the time-series feature extraction layer extracts time-series correlation features; and the output layer outputs the predicted load rate. The historical time-series data and quantized external feature values ​​are used as training samples to train the multivariate time-series prediction model, fixing its basic structure. The quantized external feature values ​​are obtained through external feature standardization, and these external features include weather information. Information on holidays and major events is collected. Historical time-series data of the target grid and the quantified values ​​of external features at the current moment are input into a multivariate time series prediction model. The predicted load rate of the target grid for future periods is calculated through the load rate time-series prediction fitting formula. Based on the predicted load rate, the corresponding level of warning is triggered. The overload propagation path is determined based on the upstream and downstream relationship of the power grid and the flow direction of the drainage network. The affected adjacent facilities and grids are identified, and a load rate prediction curve and a list of warning events are generated. The list of warning events includes the warning level, overloaded grid ID, overloaded facility ID, overload duration, warning trigger time, and overload propagation path information.

8. The smart city planning decision support system based on big data according to claim 7, characterized in that, In the prediction and early warning module, the load rate time-series prediction fitting formula is: ,in, For the predicted load rate of the target grid over the next k time slices, when At that time, a yellow alert was triggered. At that time, an orange alert was triggered. At that time, a red alert was triggered. The current grid-level load rate of the target grid. The grid-level load rate for the target grid over m historical time slices is extracted from the real-time load rate dataset by timestamp. The fitting coefficients are determined from the training results of the multivariate time series prediction model. The quantized value of the external feature at the current moment. The input length for historical time series data is determined by the training configuration of the multivariate time series prediction model.

9. A smart city planning decision support system based on big data according to claim 1, characterized in that, In the decision feedback module, decision recommendations are matched based on the duration of overload and the frequency of warnings in the early warning events. If the overload duration is less than 2 hours and there is only one warning, a short-term off-peak scheduling recommendation is matched. If the overload duration is between 2 and 24 hours and there are 2-3 cumulative warnings, a medium-term temporary capacity expansion recommendation is matched. If the same facility or grid has ≥4 cumulative warnings within 7 days, a long-term capacity expansion recommendation is matched. The short-term off-peak scheduling recommendations include adjusting industrial electricity usage periods, guiding traffic flow, and balancing communication loads. The medium-term temporary capacity expansion recommendations include activating backup units, dispatching mobile pump trucks, and providing temporary communication resources. Leasing and long-term expansion recommendations include new facility construction, pipeline expansion, and equipment upgrades; decision-making execution data includes the progress of measure implementation, actual resources invested, and abnormal situations during implementation; load rate change data includes facility-level load rate and grid-level load rate for each baseline time slice after implementation, compared with the load rate data before implementation; then, the decision-making execution data and load rate change data are used as new training samples, fed back into the multivariate time series prediction model, and retrained in a 7-day training cycle to optimize the fitting coefficients, the length of historical time series data input, correct prediction bias, and complete closed-loop optimization.

Images