Networking charging data synchronization optimization method based on multi-level distributed computing architecture

Abstract

The application discloses a networking charging data synchronization optimization method based on a multi-level distributed computing architecture, S1, collecting real-time traffic data of each charging site and performing preliminary cleaning and preprocessing; S2, using a space-time convolutional neural network to perform space-time feature extraction on the preprocessed traffic data, generating a space-time feature representation; S3, based on a Neural Prophet model, performing short-term and long-term traffic prediction, generating a traffic prediction value; S4, based on the traffic prediction value, constructing a dynamic data synchronization strategy; S5, transmitting the optimized data synchronization strategy to the regional center layer, adjusting the specific execution details of data synchronization; S6, the cloud layer generates a global optimization scheme according to the optimization information fed back by the regional center layer; S7, continuously evaluating the execution effect of the synchronization strategy. The application can provide an efficient and scientific optimization scheme in the networking charging data synchronization optimization, and brings significant technical value and economic benefits for practical application.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data synchronization optimization, in particular to a networking charging data synchronization optimization method based on a multi-level distributed computing architecture. BACKGROUND

[0002] In modern networking charging systems, with the continuous growth of traffic flow and the increasing complexity of charging systems, real-time synchronization and accurate prediction of charging data have become an important issue to improve system efficiency and optimize operation. Traditional networking charging data synchronization methods mostly rely on simple time-based synchronization strategies, usually using time series analysis-based prediction models to predict the traffic of toll sites. However, these traditional methods often ignore the mutual influence of traffic between toll sites, i.e., the spatio-temporal dependence, resulting in the inability to effectively capture the traffic fluctuation patterns between toll sites in actual applications, thereby affecting the accuracy and efficiency of data synchronization.

[0003] Existing traffic prediction techniques are mainly based on time series models such as ARIMA and ExponentialSmoothing. These models can perform time series prediction on the traffic of a single toll site, but cannot handle spatial correlations between sites. In traditional methods, traffic prediction for each toll site is performed independently, which leads to significant errors in traffic synchronization between sites. For example, the traffic fluctuations of some toll sites may be influenced by the traffic fluctuations of neighboring sites, but traditional methods cannot capture this spatio-temporal correlation, resulting in insufficient consideration of the interdependence between sites during data synchronization, thereby affecting the accuracy of synchronization.

[0004] In addition, current data synchronization strategies also have limitations. Most traditional synchronization methods rely on fixed synchronization periods, usually based on fixed time windows for data synchronization. This approach ignores the dynamic nature of traffic, such as during peak hours or special holidays when traffic changes dramatically. Fixed synchronization periods cannot effectively handle these changes, leading to synchronization delays or over-synchronization, and cannot flexibly adapt to various traffic change scenarios.

[0005] To address these issues, in recent years, more and more research has begun to explore spatio-temporal data modeling techniques based on deep learning, such as spatio-temporal convolutional neural networks and neural network models. These methods can consider both time and space dimensions, thereby more accurately capturing the spatio-temporal dependence of traffic data. For example, ST-CNN can better understand the interaction between different toll sites and the traffic fluctuation patterns by simultaneously processing traffic data in both time and space dimensions, thereby improving the accuracy of prediction. Meanwhile, the Neural Prophet model can combine historical data and real-time data to perform short-term and long-term traffic prediction, and dynamically adjust model parameters to better adapt to traffic changes.

[0006] Although these emerging methods have made some progress in handling spatio-temporal dependencies and prediction accuracy, there are still some shortcomings in the existing technology. First, although ST-CNN can capture spatio-temporal dependencies, in large-scale networked toll systems, the traffic data volume of each toll site is huge, and how to efficiently process and store these data is still a challenge. The computational complexity of existing methods is high, especially in the data transmission process between the cloud and the edge computing, which may cause delay and affect the efficiency of real-time data synchronization. Second, although the Neural Prophet model has made breakthroughs in traffic prediction, in practical applications, the traffic changes between toll sites are influenced by many factors

[0007] Therefore, the existing technology still has shortcomings in the real-time, accuracy and adaptability of data synchronization in networked toll systems. In order to better cope with the spatio-temporal dependencies, dynamic changes and system scale challenges of traffic, a new method is urgently needed that can fully utilize advanced deep learning techniques to improve the overall performance and data synchronization efficiency of networked toll systems through spatio-temporal feature extraction, traffic prediction and dynamic synchronization strategy optimization. SUMMARY

[0008] One object of the present application is to provide a networked toll data synchronization optimization method based on a multi-level distributed computing architecture. The present application can provide an efficient and scientific optimization scheme for networked toll data synchronization optimization, bringing significant technical value and economic benefits to practical applications.

[0009] The networked toll data synchronization optimization method based on a multi-level distributed computing architecture according to an embodiment of the present application comprises the following steps:

[0010] S1, collecting real-time traffic data of each toll site and performing preliminary cleaning and preprocessing of the collected traffic data at the edge computing layer;

[0011] S2, using a spatio-temporal convolutional neural network to extract spatio-temporal features from the preprocessed traffic data, using a multi-dimensional convolution kernel to perform convolution operations on both time and space dimensions, and generating spatio-temporal feature representations;

[0012] S3, based on the Neural Prophet model, combining historical traffic data and spatio-temporal feature representations, performing short-term and long-term traffic prediction, and simultaneously performing error correction and seasonal adjustment to generate traffic prediction values;

[0013] S4, based on the traffic prediction values, constructing a dynamic data synchronization strategy, optimizing the time window and frequency of data synchronization according to the traffic trend of each toll site, and dynamically adjusting the data synchronization strategy of different sites;

[0014] S5. The optimized data synchronization strategy is transmitted to the regional center layer, which then adjusts the specific execution details of data synchronization based on the traffic situation and synchronization strategy of each toll station.

[0015] S6. The cloud layer generates a global optimization plan based on the optimization information fed back by the regional center layer and performs global data synchronization scheduling.

[0016] S7. Through real-time data monitoring and feedback mechanisms, continuously evaluate the execution effect of the synchronization strategy, dynamically adjust the data synchronization strategy, and optimize it.

[0017] Optionally, S1 includes the following steps:

[0018] S11. In the network toll collection system, real-time traffic data of each toll station is collected. The traffic data includes the number of vehicles passing through each toll station, the toll amount, the traffic time period, the geographical location information of the toll station, and the timestamp.

[0019] S12. Perform preliminary cleaning on the collected raw traffic data to remove outliers and noisy data, and supplement the missing data using interpolation methods;

[0020] S13. Based on the geographical location information of the toll stations, spatial coordinates are used to mark the traffic data of each toll station, and the accurate location of each toll station is obtained to generate a traffic dataset with spatial location information.

[0021] S14. Timestamp the preprocessed traffic data to form a multidimensional time series dataset.

[0022] Optionally, S2 includes the following steps:

[0023] S21. Construct a spatiotemporal data matrix, assuming the preprocessed traffic dataset is... Where T represents the number of time steps and N represents the number of charging stations. For each charging station i and time step t, the traffic data is represented as M(t,i), where M(t,i) represents the traffic data of charging station i at time t.

[0024] S22. Define a multidimensional convolution kernel, and let the size of the time-dimensional convolution window be k. t The spatial dimension convolution window size is k n The convolution kernel is Perform a spatiotemporal convolution operation to obtain the convolution result.

[0025]

[0026] Wherein F(t) ′ i′ ) represents a time step t ′ and a site i ′ , a convolution result of is a time convolution weight, is a space convolution weight, b is a bias term, t ′ = 1, 2, …, T-k t + 1, i ′ = 1, 2, …, N-k n + 1;

[0027] S23, a ReLU activation function is used to perform nonlinear activation processing on the convolution result F to obtain an activation output F ′ ;

[0028] S24, a maximum pooling operation is performed on the activation output F ′ , a pooling window size of p t × p n , and a normalization processing is performed to obtain a normalized spatiotemporal feature representation G norm .

[0029] Optionally, the S3 includes the following steps:

[0030] S31, the spatiotemporal feature representation G is combined with historical traffic data real-time traffic data as an input sequence X of the Neural Prophet model: t

[0031] X t = {M historical (t-k), M real-time (t-k+1), …, M real-time (t), G norm};

[0032] S32, the Neural Prophet model is used to predict future time traffic, and the prediction output is M true (t+1) is the actual traffic at time t+1,

[0033] The Neural Prophet model updates the parameters by optimizing the following loss function:

[0034]

[0035] where M true (t+1) is the actual traffic at time t+1, is the predicted traffic at time t+1, λ is a regularization coefficient, θ is a learning parameter of the model, and ‖.‖2 is an L2 regularization term.

[0036] S33, calculating the flow prediction error

[0037]

[0038] wherein M true (t+1) is the actual flow at time t+1, is the predicted flow at time t+1;

[0039] The error e(t+1) is input to an error correction module, and error correction is performed, and the corrected flow prediction value is:

[0040]

[0041] wherein R(e(t+1),a) is a correction function based on the error e(t+1) and the correction coefficient a;

[0042] S34, the corrected flow prediction value is seasonally adjusted:

[0043]

[0044] wherein g is a seasonal adjustment factor, T season is the length of the seasonal period, is used to capture seasonal fluctuations, is the flow prediction value after seasonal adjustment.

[0045] Optionally, the S4 comprises the following steps:

[0046] S41, according to the generated flow prediction value a dynamic data synchronization strategy is constructed, and the size W t and the synchronization frequency F t in the synchronization strategy are determined by the following optimization function:

[0047]

[0048] wherein is the synchronized flow data, F t is the synchronization frequency, and a is an adjustment factor that balances the synchronization error and the frequency, and the optimization objective is to minimize the difference between the synchronization error and the synchronization frequency

[0049] S42, according to the output of the synchronization strategy, the synchronization time and the synchronization frequency of each toll station are calculated:

[0050]

[0051] wherein, Δt sync is the synchronization delay time, τ i (t+1) is the synchronization timing of site i, F i (t+1) is the synchronization frequency of site i;

[0052] S43, dynamically adjusting the synchronization strategy, updating the parameters of the synchronization strategy based on real-time monitoring feedback, the real-time monitoring error is ∈(t+1):

[0053]

[0054] wherein, is the real-time traffic data at time t+1;

[0055] If the monitoring error ∈(t+1) exceeds the preset threshold ∈ threshold , the synchronization strategy is optimized by adjusting the synchronization timing and frequency parameters:

[0056]

[0057] wherein, γ is the adjustment factor, is the adjusted synchronization traffic data.

[0058] Optionally, the S5 comprises the following steps:

[0059] S51, the optimized data synchronization strategy is transmitted to the regional center layer, the regional center layer analyzes the execution parameters of the synchronization strategy and further adjusts according to the traffic conditions of each regional toll site, and the synchronization strategy of each region is obtained as

[0060] S52, according to the synchronization strategy , the synchronization timing τ i (t+1) and the synchronization frequency F i (t+1) of each toll site i in the region are calculated;

[0061] S53, according to the global optimization scheme of the regional center layer, the updated synchronization strategy is applied to all toll sites in the region, and the synchronization operation parameters of each site are transmitted to the edge computing layer, and the edge computing layer realizes the synchronization operation according to the synchronization timing τ i (t+1) and the synchronization frequency F i (t+1);

[0062] S54, during the synchronization execution process, if it is found that the traffic data of some regions or toll sites exists abnormal fluctuation, the regional center layer will adjust the synchronization strategy through real-time monitoring;

[0063] S55, the synchronization strategy is dynamically updated during execution to cope with changes in factors such as traffic fluctuations, network delays, etc., and the regional center layer adjusts the synchronization timing and frequency parameters in real time according to feedback information.

[0064] Optionally, S6 comprises the following steps:

[0065] S61, the synchronization strategy is passed to the cloud layer for global data synchronization scheduling, the cloud layer aggregates the synchronization strategy parameters of each region, and coordinates through a global optimization algorithm to generate a global synchronization scheme

[0066] S62, the global synchronization scheme is applied to the global data synchronization scheduling system of the cloud layer, and the scheduling system issues synchronization instructions to different regions and sites according to the global synchronization strategy;

[0067] S63, through the real-time monitoring and feedback mechanism of the cloud layer, the effect of the global synchronization strategy is continuously evaluated;

[0068] S64, according to the global synchronization scheme and real-time feedback, dynamically adjust the parameters of the global synchronization strategy.

[0069] The beneficial effects of the present application are:

[0070] (1) The present application introduces a networking charging data synchronization optimization method based on a multi-level distributed computing architecture, which fully solves the problems of insufficient spatiotemporal dependence processing, low traffic prediction accuracy, fixed data synchronization strategy, etc. in the prior art, significantly improves the overall performance and data synchronization efficiency of the networking charging system, and by using the spatiotemporal convolutional neural network and the Neural Prophet model, the present application can accurately capture the spatiotemporal dependence relationship between different charging sites, so that the traffic prediction not only depends on the time series, but also considers the mutual influence of the traffic between the sites, thereby improving the accuracy of traffic prediction.

[0071] (2) The present application combines a multi-level distributed computing architecture, and the system can optimize data synchronization between different levels (such as edge computing layer, regional center layer, cloud layer). The edge computing layer dynamically adjusts the synchronization timing and synchronization frequency according to the local real-time traffic data and the global synchronization strategy, thereby realizing accurate real-time data synchronization. The regional center layer and the cloud layer optimize the synchronization strategy according to the traffic characteristics and synchronization requirements of different regions, so as to flexibly cope with the problem of excessive synchronization or delayed synchronization under the condition of large traffic fluctuations or special circumstances.

[0072] (3) The dynamic data synchronization strategy proposed by the present application can adjust the synchronization time and frequency in real time when facing traffic fluctuations, greatly improving the adaptability and real-time performance of synchronization. The system can dynamically adjust the synchronization strategy according to the actual situation. In the period of high peak or special holidays with large traffic fluctuations, through accurate synchronization time and frequency optimization, the waste of bandwidth resources and synchronization delay are avoided, and the synchronization efficiency and accuracy of data are ensured. BRIEF DESCRIPTION OF DRAWINGS

[0073] The accompanying drawings are included to provide a further understanding of the present application, and constitute a part of the specification, illustrate the present application and explain the principles of the present application, and do not constitute a limitation of the present application. In the drawings:

[0074] Fig. 1 Flow chart of the networking charging data synchronization optimization method based on the multi-level distributed computing architecture proposed by the present application;

[0075] Fig. 2 Flow chart of the Neural Prophet model generating traffic prediction value in the networking charging data synchronization optimization method based on the multi-level distributed computing architecture proposed by the present application. DETAILED DESCRIPTION

[0076] The present application will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, and only illustrate the basic structure of the present application in a schematic manner, and therefore only show the components related to the present application.

[0077] Reference Figs. 1-2 The networking charging data synchronization optimization method based on the multi-level distributed computing architecture, comprising the following steps:

[0078] S1, collecting real-time traffic data of each toll station, and performing preliminary cleaning and preprocessing on the collected traffic data in the edge computing layer;

[0079] S2, using a spatio-temporal convolutional neural network to extract spatio-temporal features from the preprocessed traffic data, using a multi-dimensional convolution kernel to simultaneously perform convolution operation on time and space dimensions, and generating spatio-temporal feature representation;

[0080] S3, based on the Neural Prophet model, combining historical traffic data and spatio-temporal feature representation, performing short-term and long-term traffic prediction, and simultaneously performing error correction and seasonal adjustment to generate traffic prediction value;

[0081] S4, based on the traffic prediction value, constructing a dynamic data synchronization strategy, optimizing the time window and frequency of data synchronization according to the traffic trend of each toll station, and dynamically adjusting the data synchronization strategy of different stations;

[0082] S5, the optimized data synchronization strategy is transmitted to the regional center layer, and the regional center layer adjusts specific execution details of data synchronization according to traffic conditions and the synchronization strategy of each toll site;

[0083] S6, the cloud layer generates a global optimization scheme according to the optimization information fed back by the regional center layer, and executes global data synchronization scheduling;

[0084] S7, through the real-time data monitoring and feedback mechanism, the execution effect of the synchronization strategy is continuously evaluated, and the data synchronization strategy is dynamically adjusted and optimized.

[0085] The present application can effectively remove noise data and fill in missing values by collecting real-time traffic data of each toll site and performing preliminary cleaning and preprocessing, thereby ensuring the quality of the traffic data input to the subsequent processing. Data preprocessing not only helps to improve the reliability of the data, but also reduces the prediction error caused by data quality problems. In addition, the addition of time stamp marking and geographic location information enables the traffic data to be combined with the dimensions of time and space, providing more comprehensive background information for subsequent spatio-temporal feature extraction and analysis.

[0086] In the embodiment, S1 includes the following steps:

[0087] S11, in the networked toll system, collecting real-time traffic data of each toll site, the traffic data including the number of vehicles passing through each toll site, the toll amount, the traffic time period, the geographic location information of the toll station and the time stamp;

[0088] S12, the collected original traffic data is preliminarily cleaned to remove abnormal values and noise data, and the missing data is supplemented by interpolation method;

[0089] S13, according to the geographic location information of the toll site, the traffic data of each toll site is spatially coordinate calibrated, and the accurate position of each toll site is obtained, and a traffic data set with spatial position information is generated;

[0090] S14, the preprocessed traffic data is time stamped to form a multi-dimensional time series data set.

[0091] The present application can effectively capture the spatio-temporal dependence relationship between different toll sites by using the spatio-temporal convolutional neural network to extract the spatio-temporal features of the traffic data. This process solves the defect that the traditional method cannot handle the spatio-temporal correlation, so that the traffic prediction not only considers the change of time dimension, but also considers the spatial relationship between the toll sites, thereby improving the prediction accuracy and accuracy. The extraction of spatio-temporal features provides more rich information support for subsequent traffic prediction and data synchronization strategy optimization.

[0092] In this embodiment, S2 includes the following steps:

[0093] S21, construct a spatio-temporal data matrix, let the pre-processed traffic data set be wherein T represents the number of time steps, N represents the number of toll stations, for each toll station i and time step t, the traffic data is represented as M(t, i), and M(t, i) represents the traffic data of toll station i at time t;

[0094] S22, define a multi-dimensional convolution kernel, let the time dimension convolution window size be k t , the spatial dimension convolution window size be k n , and the convolution kernel be perform a spatio-temporal convolution operation to obtain a convolution result

[0095]

[0096] wherein F(t ′ ,i ′ ) represents the convolution result of time step t ′ and station i ′ , w represents a time convolution weight, represents a spatial convolution weight, and b represents a bias term, t ′ = 1, 2, …, T-k t +1, i ′ = 1, 2, …, N-k n +1;

[0097] S23, perform a nonlinear activation processing on the convolution result F using a ReLU activation function to obtain an activation output F ′ .

[0098] S24, perform a max-pooling operation on the activation output F ′ , a pooling window size of p t × p n , and a normalization processing to obtain a normalized spatio-temporal feature representation G norm .

[0099] The application further enhances the prediction ability of the system through traffic prediction based on the Neural Prophet model, combines historical data and real-time data, and the Neural Prophet can perform short-term and long-term traffic prediction and can handle complex factors such as seasonal changes and holiday effects, which significantly improves the flexibility and adaptability of traffic prediction and can more accurately reflect the traffic change trend in different time periods. In addition, the self-adaptive error correction function in the prediction process can reduce the prediction error, making the future traffic prediction result more reliable, thereby providing accurate basis for the optimization of the synchronization strategy.

[0100] In this embodiment, S3 includes the following steps:

[0101] S31, combining spatiotemporal feature representation Historical traffic data Real-time traffic data As the input sequence X of the Neural Prophet model t :

[0102] X t ={M historical (t-k),M real-time (t-k+1),…,M real-time (t),G norm};

[0103] S32, using the Neural Prophet model to predict the traffic at future time, the prediction output is The predicted traffic at time t+1;

[0104] The Neural Prophet model updates the parameters by optimizing the following loss function:

[0105]

[0106] Where M true (t+1) is the actual traffic at time t+1, is the predicted traffic at time t+1, λ is the regularization coefficient, θ is the learning parameter of the model, and ‖.‖2 is the L2 regularization term;

[0107] S33, calculate the traffic prediction error

[0108]

[0109] Where M true (t+1) is the actual traffic at time t+1, is the predicted traffic at time t+1;

[0110] The error e(t+1) is input to an error correction module to perform error correction, and a corrected flow prediction value is obtained.

[0111]

[0112] wherein R(e(t+1), a) is a correction function based on the error e(t+1) and the correction coefficient a;

[0113] S34, the corrected flow prediction value is seasonally adjusted.

[0114]

[0115] wherein g is a seasonal adjustment factor, and T season is the length of the seasonal cycle, for capturing seasonal fluctuations, is the flow prediction value after seasonal adjustment.

[0116] The dynamic data synchronization strategy constructed based on the flow prediction result in the application can dynamically adjust the synchronization strategy according to the flow changes of different toll stations by optimizing the synchronization time and frequency. This method overcomes the shortcomings of the traditional fixed cycle synchronization strategy, can flexibly adapt to flow fluctuations and real-time changes, avoids the resource waste or delay problem caused by fixed synchronization frequency, and through the combination of space-time characteristics, the strategy can more accurately reflect the mutual dependence relationship between stations, ensuring efficient synchronization during high flow periods and special time periods.

[0117] In this embodiment, S4 includes the following steps:

[0118] S41, according to the generated flow prediction value a dynamic data synchronization strategy is constructed, and the time window size W t and the synchronization frequency F t are determined by the following optimization function:

[0119]

[0120] wherein, is the synchronized flow data, F t is the synchronization frequency, and a is an adjustment factor for balancing the synchronization error and the frequency, and the optimization target is to minimize the difference between the synchronization error and the synchronization frequency

[0121] S42, according to the output of the synchronization strategy, the synchronization time and synchronization frequency of each toll station are calculated:

[0122]

[0123] Wherein, Δt sync is the synchronization delay time, τ i (t+1) is the synchronization time of the site i, F i (t+1) is the synchronization frequency of the site i.

[0124] S43, dynamically adjust the synchronization strategy, update the parameters of the synchronization strategy based on real-time monitoring feedback, and the real-time monitoring error is ∈(t+1):

[0125]

[0126] Wherein, is the real-time flow data at time t+1;

[0127] If the monitoring error ∈(t+1) exceeds the preset threshold ∈ threshold , the synchronization strategy is optimized by adjusting the synchronization time and frequency parameters:

[0128]

[0129] Wherein, γ is the adjustment factor, is the adjusted synchronization flow data.

[0130] The further optimization of the synchronization strategy by the regional center layer makes the synchronization operation of each region more coordinated, and the regional center layer can effectively adjust the synchronization strategy according to the flow conditions and mutual dependence of each toll site, avoid the synchronization error between different toll sites in the region, and improve the overall efficiency of the system. The optimization not only improves the overall efficiency of the system, but also ensures the synchronization accuracy and real-time performance between regions in the multi-region coordination process, thereby improving the performance of the overall system.

[0131] In this embodiment, S5 includes the following steps:

[0132] S51, the optimized data synchronization strategy is transmitted to the regional center layer, the regional center layer analyzes the execution parameters of the synchronization strategy and further adjusts according to the flow conditions of each regional toll site, and obtains the synchronization strategy of each region as

[0133] S52, according to the synchronization strategy the synchronization time τ i (t+1) and the synchronization frequency F i (t+1) of each toll site i in the region are calculated.

[0134] S53, according to the global optimization scheme of the regional center layer, the updated synchronization strategy The synchronization operation parameters of each station are transmitted to the edge computing layer, and the edge computing layer adjusts the synchronization time τ i (t+1) and the synchronization frequency F i (t+1) to realize the synchronization operation.

[0135] S54, during the synchronization execution process, if it is found that the traffic data of some regions or toll stations has abnormal fluctuations, the regional center layer will adjust the synchronization strategy through real-time monitoring.

[0136] S55, the synchronization strategy is dynamically updated during the execution process to cope with changes in traffic fluctuations, network delays and other factors, and the regional center layer adjusts the synchronization time and frequency parameters in real time according to feedback information.

[0137] The global synchronization scheduling of the cloud layer of the application is globally optimized according to the feedback information of the regional center layer, so that the system can realize efficient synchronization scheduling in a large-scale networked toll system. The scheduling system of the cloud layer not only accurately adjusts the synchronization time and frequency according to the global optimization scheme, but also flexibly adjusts the synchronization strategy when the system load is high or abnormal conditions occur. Through global optimization, the data synchronization between different regions can be better coordinated, the synchronization efficiency and response speed of the overall system can be improved, and the data synchronization can be efficiently and stably operated in a complex traffic and toll environment.

[0138] In this embodiment, S6 includes the following steps:

[0139] S61, the synchronization strategy is transmitted to the cloud layer for global data synchronization scheduling, and the cloud layer collects the synchronization strategy parameters of each region and coordinates them through a global optimization algorithm to generate a global synchronization scheme

[0140] S62, the global synchronization scheme is applied to the global data synchronization scheduling system of the cloud layer, and the scheduling system issues synchronization instructions to different regions and stations according to the global synchronization strategy.

[0141] S63, through the real-time monitoring and feedback mechanism of the cloud layer, the effect of the global synchronization strategy is continuously evaluated.

[0142] S64, according to the global synchronization scheme and real-time feedback, the parameters of the global synchronization strategy are dynamically adjusted.

[0143] Embodiment:

[0144] The embodiment is applied to a large city's networked toll collection system, which is composed of multiple toll sites, each responsible for a different toll area, and the traffic volume of each area fluctuates greatly at different time periods. Due to changes in traffic conditions, holiday effects, and adverse weather, the traffic fluctuations between toll sites are often interdependent, which poses a great challenge to traditional traffic prediction and data synchronization.

[0145] Before the implementation, the system used a fixed time window synchronization strategy, with each toll site performing data synchronization every 30 minutes. However, this strategy did not take into account the traffic fluctuations of different sites, resulting in a failure to respond in time during traffic surges, with a large number of delayed or repeated synchronization problems. This not only consumed excessive bandwidth resources but also affected the overall charging efficiency. Specifically, during some high-traffic periods, the data transmission delay was as high as 5 minutes, while during low-traffic periods, the synchronization frequency was too high, leading to waste of bandwidth resources.

[0146] To solve this problem, the implementer collected real-time traffic data from each toll site at the edge computing layer of the toll collection system, including the number of vehicles passing through each toll site per minute, the toll amount, and timestamp information. Then, a spatio-temporal convolutional neural network was used to process these data and extract spatio-temporal features to capture the traffic dependency between different toll sites. These spatio-temporal feature information was passed to the Neural Prophet model as input data, which made traffic predictions based on historical traffic data and real-time data.

[0147] Through the prediction of the Neural Prophet model, the system can accurately predict the traffic trend of each toll site within the next 30 minutes. This prediction takes into account not only the historical traffic of each toll site but also external factors such as holiday effects and weather changes, ensuring the accuracy of the prediction.

[0148] Based on the traffic prediction results, the system dynamically adjusts the data synchronization strategy for each toll site. For high-traffic sites, the synchronization frequency is increased, while for low-traffic sites, the synchronization frequency is appropriately reduced. This dynamic traffic prediction-based synchronization strategy significantly reduces bandwidth waste, improves data synchronization efficiency, and avoids the problems of excessive synchronization or synchronization delay.

[0149] During the implementation process, the system not only performs global scheduling and optimization in the cloud but also performs local optimization at the regional center layer. Each region adjusts its synchronization strategy based on its traffic characteristics, and the cloud layer further optimizes the global synchronization scheduling based on the needs of different regions to ensure that data synchronization in all regions is completed at the most appropriate time.

[0150] The following is a comparison table of data synchronization before and after implementation of a certain toll station:

[0151] Table 1 Comparison of synchronization effect of a certain toll station before and after implementation

[0152]

[0153] As can be seen from the table, the system synchronization delay after implementation is significantly reduced, especially during peak hours, the data synchronization delay is reduced from 5.2 minutes before implementation to 2.1 minutes, about 59% reduction, this improvement greatly improves the real-time response ability of the system, so that the toll system can more efficiently process a large amount of traffic data.

[0154] Through the implementation of the method of the present application, the data synchronization efficiency of the toll station has been significantly improved, the system can dynamically adjust the synchronization strategy according to the real-time traffic prediction, avoiding the delay and resource waste caused by traffic fluctuations in the traditional synchronization method, in addition, the precise traffic prediction and optimized synchronization strategy greatly improve the response speed and user experience of the system, the above data fully proves the beneficial effects and technical advantages of the present application in practical application.

[0155] The above is only the preferred specific implementation of the present application, but the protection scope of the present application is not limited to this, any skilled person in the art can make equivalent replacement or change according to the technical solution and the inventive concept of the present application within the technical range disclosed by the present application, which should be covered within the protection scope of the present application.

Claims

1. A networking charging data synchronization optimization method based on a multi-level distributed computing architecture, characterized in that, Comprise the following steps: S1, collecting real-time traffic data of each toll site, and performing preliminary cleaning and preprocessing on the collected traffic data in the edge computing layer; S2, using a spatio-temporal convolutional neural network to extract spatio-temporal features from the preprocessed traffic data, using a multi-dimensional convolution kernel to simultaneously perform convolution operations on the time and space dimensions, and generating spatio-temporal feature representations; S3, based on the Neural Prophet model, combining historical traffic data and spatio-temporal feature representations, performing short-term and long-term traffic prediction, and simultaneously performing error correction and seasonal adjustment to generate traffic prediction values; S4, based on the traffic prediction values, constructing a dynamic data synchronization strategy, optimizing the time window and frequency of data synchronization according to the traffic trend of each toll site, and dynamically adjusting the data synchronization strategy of different sites; S5, passing the optimized data synchronization strategy to the regional center layer, and adjusting the specific execution details of data synchronization according to the traffic situation and synchronization strategy of each toll site; S6, the cloud layer generates a global optimization scheme according to the optimization information fed back by the regional center layer, and executes global data synchronization scheduling; S7, through real-time data monitoring and feedback mechanism, continuously evaluate the execution effect of synchronization strategy, dynamically adjust the data synchronization strategy and optimize.

2. The method for networking charging data synchronization optimization based on multi-level distributed computing architecture according to claim 1, characterized in that, The S1 comprises the following steps: S11, collecting real-time traffic data of each toll site in the networked toll system, the traffic data including the number of vehicles passing through each toll site, the toll amount, the traffic time period, the geographic location information of the toll station and the timestamp; S12, performing preliminary cleaning on the collected original traffic data, removing abnormal values and noise data, and supplementing missing data through interpolation method; S13, according to the geographic location information of the toll site, the traffic data of each toll site is spatially coordinate calibrated, and the accurate position of each toll site is obtained, and a traffic data set with spatial position information is generated; S14, time stamp marking is performed on the preprocessed traffic data to form a multi-dimensional time series data set.

3. The method for networking charging data synchronization optimization based on multi-level distributed computing architecture according to claim 1, characterized in that, The S2 comprises the following steps: S21. Construct a spatiotemporal data matrix, assuming the preprocessed traffic dataset is... ,in, Indicates the number of time steps. This indicates the number of toll stations. For each toll station... and time step Traffic data is represented as , Indicates at time Toll stations Traffic data; S22, define a multi-dimensional convolution kernel, set the time dimension convolution window size as , the spatial dimension convolution window size as , the convolution kernel as , perform the space-time convolution operation to obtain the convolution result : ； wherein, denotes a time step and a site convolution result, is a time convolution weight, is a spatial convolution weight, is a bias term, , ; S23, adopt ReLU activation function to the convolution result perform nonlinear activation processing to obtain an activation output ; S24, to the activation output A max pooling operation is performed, with a pooling window size of and a normalization process is performed to obtain a normalized spatiotemporal feature representation .

4. The method for networking charging data synchronization optimization based on multi-level distributed computing architecture according to claim 1, characterized in that, The S3 comprises the following steps: S31, spatio-temporal feature representation , historical traffic data , real-time traffic data as input sequences to the Neural Prophet model : ； S32, using the Neural Prophet model to predict the flow at future time, the prediction output is , the flow at time predicted; The Neural Prophet model updates parameters by optimizing the following loss function: ； wherein, is the time actual flow, is the time predicted flow, is a regularization coefficient, is a learning parameter of the model, is an L2 regularization term; S33, calculate flow prediction error : ； wherein, is the time actual flow, is the time predicted flow; The error is input to an error correction module, and the flow rate prediction value after error correction is calculated as: Q = Q0 + K * e ； wherein is a correction function based on an error and a correction coefficient ; S34, to the corrected flow forecast value Seasonal adjustment is made: ； wherein, is a seasonal adjustment factor, is a length of a seasonal cycle, for capturing seasonal fluctuations, is a flow prediction value after seasonal adjustment.

5. The method for networking charging data synchronization optimization based on multi-level distributed computing architecture according to claim 1, characterized in that, The S4 comprises the following steps: S41、According to the generated traffic prediction value , a dynamic data synchronization strategy is constructed, and the time window size in the synchronization strategy and the synchronization frequency are determined by the following optimization function: ； wherein, is the synchronized flow data, is the synchronized frequency, is a balancing adjustment factor between the synchronization error and the frequency, the optimization objective is optimized by minimizing the difference between the synchronization error and the synchronized frequency ; S42, according to the output of the synchronization strategy, calculate the synchronization time and synchronization frequency of each toll site: ； ； in, For synchronization delay time, For the site The timing of synchronization For the site The synchronization frequency; S43, dynamically adjusting the synchronization strategy, updating the parameters of the synchronization strategy based on real-time monitoring feedback, and the real-time monitoring error is : ； wherein, is the time real-time traffic data; If the monitoring error exceeds a preset threshold then the synchronization strategy is optimized by adjusting the synchronization timing and frequency parameters: ； wherein, is a modulating factor, is the adjusted synchronized traffic data.

6. The method for networking charging data synchronization optimization based on multi-level distributed computing architecture according to claim 1, characterized in that, The S5 comprises the following steps: S51, the optimized data synchronization strategy is transmitted to the regional center layer, the regional center layer analyzes the execution parameters of the synchronization strategy and further adjusts according to the traffic conditions of the toll stations in each region, and the synchronization strategy of each region is obtained as ; S52, according to the synchronization strategy for each toll site synchronization opportunities within a zone and synchronization frequencies performing the calculation; S53、According to the global optimization scheme of the regional center layer, the updated synchronization strategy is applied to all charging sites in the region, and the synchronization operation parameters of each site are transmitted to the edge computing layer, and the edge computing layer transmits the synchronization operation parameters of each site to the regional center layer according to the synchronization time and synchronization frequency to realize synchronization operation; S54, in the synchronization execution process, if it is found that the traffic data of some regions or toll sites exists abnormal fluctuation, the regional center layer will adjust the synchronization strategy through real-time monitoring; S55, the synchronization strategy will be dynamically updated during execution to cope with the changes of traffic fluctuation and network delay factors, and the regional center layer will adjust the synchronization time and frequency parameters in real time according to the feedback information.

7. The method for networking charging data synchronization optimization based on multi-level distributed computing architecture according to claim 1, characterized in that, The S6 comprises the following steps: S61、Synchronization strategy The global data synchronization scheduling is performed to the cloud layer, the cloud layer collects the synchronization strategy parameters of each region, and coordinates through a global optimization algorithm to generate a global synchronization scheme ​ S62、the global synchronization scheme The global data synchronization scheduling system applied to the cloud layer, the scheduling system issues synchronization instructions to different regions and sites according to the global synchronization strategy; S63, through the real-time monitoring and feedback mechanism of the cloud layer, continuously evaluate the effect of global synchronization strategy; S64, dynamically adjust the parameters of global synchronization strategy according to global synchronization scheme and real-time feedback.

Images