A simulation and deduction method, apparatus and equipment based on risk propagation diagram

Abstract

This invention provides a simulation and deduction method, apparatus, and equipment based on a risk propagation graph. By collecting multi-source heterogeneous data from a logistics and transportation system, and after cleaning, alignment, and standardization, multi-dimensional fused risk features are extracted to characterize the current risk status of each transportation node and segment. A risk propagation graph is constructed based on the transportation network topology, using the fused risk features as the initial risk status of nodes and edges. Discrete event-driven simulation and deduction are performed at a preset time step. The propagation increment between adjacent nodes is calculated according to risk propagation rules and updated by aggregation to obtain the evolution results within the simulation period. Finally, a comprehensive evaluation index is calculated, and an emergency dispatch plan is generated. This invention can complete a forward-looking quantitative deduction of the risk transmission path and its superimposed impact before the risk actually reaches downstream nodes, ensuring that the generation of the emergency dispatch plan precedes the actual risk transmission process.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of simulation deduction, in particular to a simulation deduction method and device based on a risk propagation graph and equipment. BACKGROUND

[0002] With the rapid development of the logistics industry, risk prediction and emergency scheduling in the transportation process have become the key to ensuring the safety of goods and reducing operating costs. The logistics transportation system involves various transportation modes such as highways, railways, and air routes. During operation, it is affected by the coupling of traffic conditions, weather conditions, equipment performance, and unexpected events. Abnormalities in any link can spread and cause a chain reaction in the network.

[0003] The existing logistics transportation risk prediction and scheduling system usually adopts a processing architecture of "single-point detection - independent alarm - manual scheduling": the system independently analyzes a single data source (such as GPS trajectory or weather information), triggers an alarm when a certain indicator exceeds the preset threshold, and then the operator formulates an emergency scheduling plan based on the alarm information.

[0004] Under this processing architecture, since each risk factor is detected and evaluated in isolation, the risk correlation between adjacent transportation nodes cannot be identified and utilized by the system, resulting in the inability to identify and quantitatively evaluate the chain reaction caused by the spread of the risk along the transportation path to the downstream nodes and the mutual superposition of other risk sources before the risk actually reaches the downstream nodes. Therefore, the formulation of the emergency scheduling plan lags behind the actual conduction process of the risk, missing the best emergency disposal window, causing avoidable transportation delays and economic losses.

[0005] Therefore, the present application is proposed. SUMMARY

[0006] The present application discloses a simulation deduction method, device and equipment based on a risk propagation graph, aiming to solve the problem that the prior art cannot quantitatively deduce the conduction and superposition of risks in the logistics transportation network in advance, resulting in lagging emergency scheduling.

[0007] The first embodiment of the present application provides a simulation deduction method based on a risk propagation graph, comprising: Collecting multi-source heterogeneous data of a logistics transportation system, performing cleaning, alignment and standardization processing on the multi-source heterogeneous data to obtain standardized transportation data, and the multi-source heterogeneous data comprising transportation network topology data and risk factor data; Performing multi-dimensional feature extraction and fusion on the standardized transportation data to obtain fusion risk features representing the current risk state of each transportation node and transportation section; construct a risk propagation graph based on transportation network topology data in the standardized transportation data, take the fused risk features as initial risk states of corresponding nodes and edges in the risk propagation graph, and configure risk propagation rule parameters; perform discrete event driven simulation deduction on the risk propagation graph according to preset time steps, calculate a propagation increment of risk between adjacent nodes according to the risk propagation rule parameters, and perform convergence update on multi-source risk increments received by each node to obtain an evolution result of the risk propagation graph in a deduction period; calculate a comprehensive evaluation index based on the evolution result, and generate a logistics transportation emergency scheduling scheme according to the comprehensive evaluation index.

[0008] The second embodiment of the present application provides a simulation deduction device based on a risk propagation graph, comprising: a data management unit configured to collect multi-source heterogeneous data of a logistics transportation system, perform cleaning, alignment and standardization processing on the multi-source heterogeneous data to obtain standardized transportation data, and wherein the multi-source heterogeneous data comprises transportation network topology data and risk factor data; a feature modeling unit configured to perform multi-dimensional feature extraction and fusion on the standardized transportation data to obtain fused risk features representing current risk states of each transportation node and transportation section; a risk graph construction unit configured to construct a risk propagation graph based on transportation network topology data in the standardized transportation data, take the fused risk features as initial risk states of corresponding nodes and edges in the risk propagation graph, and configure risk propagation rule parameters; a simulation deduction unit configured to perform discrete event driven simulation deduction on the risk propagation graph according to preset time steps, calculate a propagation increment of risk between adjacent nodes according to the risk propagation rule parameters, and perform convergence update on multi-source risk increments received by each node to obtain an evolution result of the risk propagation graph in a deduction period; a decision unit configured to calculate a comprehensive evaluation index based on the evolution result, and generate a logistics transportation emergency scheduling scheme according to the comprehensive evaluation index.

[0009] The third embodiment of the present application provides a simulation deduction device based on a risk propagation graph, comprising a memory and a processor, wherein the memory stores a computer program, and the computer program can be executed by the processor to implement the simulation deduction method based on the risk propagation graph according to any one of the above embodiments.

[0010] The fourth embodiment of the present application provides a computer readable storage medium storing a computer program, which can be executed by a processor of a device where the computer readable storage medium is located to implement the open office overall work efficiency decay prediction method according to any one of the preceding embodiments.

[0011] Based on the simulation deduction method, device and equipment based on the risk propagation graph provided by the present application, the standardized transportation data is obtained by collecting the multi-source heterogeneous data of the logistics transportation system and performing cleaning, alignment and standardization processing, and then the fusion risk features representing the current risk states of each transportation node and transportation section are obtained by performing multi-dimensional feature extraction and fusion on the data; on this basis, the risk propagation graph is constructed based on the transportation network topology data, and the fusion risk features are taken as the initial risk states of the corresponding nodes and edges in the graph, and then the simulation deduction driven by discrete events is performed according to the preset time step, the propagation increment of the risk between adjacent nodes is calculated according to the risk propagation rule parameters, and the multi-source risk increments received by each node are updated, and the evolution result of the risk propagation graph in the deduction period is obtained; finally, the comprehensive evaluation index is calculated based on the evolution result, and the logistics transportation emergency scheduling scheme is generated accordingly, so that the prospective quantitative deduction of the risk conduction path and the superimposed influence is completed before the risk actually reaches the downstream node, and the generation of the emergency scheduling scheme precedes the actual conduction process of the risk. BRIEF DESCRIPTION OF DRAWINGS

[0012] Figure 1 is a flow diagram of a simulation deduction method based on a risk propagation graph provided by the first embodiment of the present application; Figure 2 is a module diagram of a simulation deduction device based on a risk propagation graph provided by the second embodiment of the present application. DETAILED DESCRIPTION

[0013] The technical solutions in the embodiments of the present application will be described in detail below with reference to the drawings of the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0014] In order to better understand the technical solutions of the present application, the embodiments of the present application will be described in detail below with reference to the drawings.

[0015] The present application discloses a simulation deduction method, device and equipment based on a risk propagation graph, which aims to solve the problem that the prior art cannot perform prospective quantitative deduction on the conduction diffusion and superimposed influence of risks in the logistics transportation network, resulting in lag of emergency scheduling.

[0016] The first embodiment of the present invention provides a simulation and extrapolation method based on a risk propagation graph, which can be executed by a simulation device (hereinafter referred to as a system), specifically by one or more processors within the simulation device, to at least implement the following steps. S101, Collect multi-source heterogeneous data from the logistics and transportation system, clean, align and standardize the multi-source heterogeneous data to obtain standardized transportation data, wherein the multi-source heterogeneous data includes transportation network topology data and risk factor data; In this embodiment, the system simultaneously accesses multi-source heterogeneous data from the logistics and transportation scenario through a preset data acquisition interface. This multi-source heterogeneous data specifically includes historical transportation data categorized by route and product type, real-time dynamic data (such as vehicle speed, location, and temperature) reported by in-vehicle IoT devices and GPS positioning devices, real-time traffic data provided by road traffic and flight route management systems, and external environmental data provided by meteorological services and policy dissemination platforms. For the collected continuous time-series data (such as GPS speed and temperature), when an observation value at a certain time point t is missing, the system selects the nearest known time point before that time point. and the most recent known time point thereafter and according to the formula Linear interpolation is performed to obtain the filled value at time t, where , They are respectively , The system calculates the observed values ​​at each time point. For discrete attribute data (e.g., event labels), it uses a forward or backward neighbor-based fill method to fill the data based on the nearest known label in time. After data cleaning, the system further performs time and spatial alignment processing on each data source. For time alignment, high-frequency data such as GPS is aggregated to a unified low-frequency time axis using a downsampling strategy. For low-frequency data such as meteorological data, the aforementioned linear interpolation method is used for upsampling to match the high-frequency time axis. For spatial alignment, a map matching method based on an implicit Markov model is used for the location and direction of travel data of transportation vehicles. According to the formula Calculate the transition probability. According to the formula Calculation, where GPS observation point at time t for In candidate road sections Projection points on GPS error standard deviation, For the shortest path distance within the road segment network, The straight-line distance between adjacent GPS points Using the scale parameter, the optimal matching route sequence for transportation vehicles is obtained by maximizing the joint posterior probability of the observation probability and the transition probability; for data such as meteorology, air quality, and temperature collected according to geographic grids, a bilinear interpolation method is used according to the formula... To achieve grid matching, where , The normalized coordinates of the target GPS point within its grid. , , , These are the observation values ​​of the four vertices of the grid. Finally, the data after the above interpolation, padding, time alignment and spatial alignment processes are normalized through a preset standardized API interface, so that data from different data sources have unified field naming rules, data type definitions and event description structures, resulting in standardized transportation data and stored in the data warehouse.

[0017] S102, Multi-dimensional feature extraction and fusion are performed on the standardized transportation data to obtain fused risk features that characterize the current risk status of each transportation node and transportation segment; In this embodiment, feature vectors from K data sources are extracted from standardized transportation data according to the dimensions of transportation nodes and transportation segments. Where K is a positive integer greater than or equal to 2. Let represent the feature vector extracted from the i-th data source (i = 1, 2, ..., K). Specifically, the first data source can be device status features such as speed, temperature, and vibration reported by in-vehicle IoT devices; the second data source can be traffic status features such as congestion index and accident rate provided by the road condition system; the third data source can be environmental status features such as rainfall intensity, wind speed, and visibility provided by the meteorological service platform; and so on for the remaining data sources. After extracting the feature vectors from each data source, the system selects a fusion method based on the reliability and importance differences of each data source in the current business scenario. When the reliability of each data source is roughly equivalent or their weight information is not yet available, the feature vector is concatenated according to the formula. The K feature vectors are concatenated end-to-end to form a fused feature vector x_fused, thus preserving all original feature dimensions of each data source. For example, the feature vector of the equipment status of a transportation node at a certain moment is... (Speed, temperature, vibration), traffic state feature vector is Congestion index, accident rate), meteorological state feature vector is (Rainfall intensity, wind speed, visibility), then the fused feature vector of that node is obtained after splicing. When there are significant differences in the reliability or importance of various data sources to risk assessment, a weighted fusion method should be adopted according to the public... The K feature vectors are weighted and summed to form a fused feature vector. ,in The weight corresponding to the i-th data source and satisfying The weight The values ​​can be pre-set based on business experience (e.g., giving higher weight to meteorological data during typhoon season), or learned through backpropagation or least squares based on the prediction accuracy of each data source on the historical validation set, so that data sources with high reliability and large risk contribution occupy a higher proportion in the fused features; finally, the system will combine the fused feature vectors output by at least one of the above fusion methods. As a fusion risk feature representing the current risk status of the transportation node or transportation segment, it is persistently stored in the feature data warehouse indexed by node ID or segment ID, so that it can be called and attached to the corresponding initial risk status one by one by node and edge when constructing the risk propagation graph.

[0018] S103, Construct a risk propagation graph based on the transportation network topology data in the standardized transportation data, use the fused risk features as the initial risk state of the corresponding nodes and edges in the risk propagation graph, and configure risk propagation rule parameters; In this embodiment, a graph structure is constructed based on the transportation network topology data in standardized transportation data. Where V is the set of nodes, including at least one of cities, transit stations, intersections, and waypoints, and E is the set of edges, where each edge represents a transportation segment (such as a highway segment, railway section, or air route segment) connecting two nodes. For example, a regional logistics network can be modeled as a directed weighted graph with nodes such as "Shanghai Transit Station," "Hangzhou Transit Station," "Shanghai-Hangzhou Expressway G60-1 Section," and "Shanghai-Hangzhou Expressway G60-2 Section," and transportation segments between nodes as edges. After the graph structure is constructed, the system retrieves the corresponding fused risk features from the feature data warehouse according to the node ID and segment ID, and uses the fused risk feature corresponding to node v as the initial risk state of node v at the start of the simulation. (Values ​​are normalized to the [0, 1] interval; for example, a transfer station is initially assigned R0(v) = 0.65 due to high surrounding rainfall intensity and increased equipment failure rate.) The fusion risk characteristics corresponding to the edge are attached to the edge as attribute parameters (such as road traffic volume, accident probability, and average travel time) to complete the initialization of the risk propagation graph. Subsequently, the system configures risk propagation rule parameters, including: risk attenuation coefficient. This is used to characterize the rate at which risk decays with the distance it travels between nodes. A higher value indicates faster attenuation. Its value can be obtained by fitting historical cascading disaster data (such as measured attenuation curves of risk values ​​of adjacent road sections under the influence of historical typhoons), with a typical range of 0.01 to 1.0; Risk trigger threshold. This is used to determine whether a node is qualified to propagate risk to neighboring nodes, only if the node's risk status is... Exceed The propagation event is only triggered at this time. The value can be set according to the business's risk sensitivity requirements (e.g., for high-value cold chain transportation). For conventional transportation ); Propagation intensity coefficient The node vulnerability coefficient is used to control the impact intensity of risk sources on adjacent nodes. It can be trained based on the regression relationship between the "source node risk value and the actual incremental value of adjacent nodes" in historical cascading disaster data. This parameter is used to characterize the sensitivity of node v to incoming risks. Its value is related to factors such as the node's emergency response capabilities, the risk resistance level of its infrastructure, and the adequacy of its backup resources. For example, a transfer station equipped with a comprehensive emergency plan and a backup fleet can be configured with this parameter. For intersections located in remote areas without backup resources, backup can be set up. Finally, the system will use the above parameters. , , and the corresponding nodes Together with scenario files (such as typhoon scenarios, rainstorm scenarios, and large-scale equipment failure scenarios), they are standardized and packaged to form a configurable risk propagation rule base, and the risk propagation graph and corresponding rule base are output after initialization and parameter configuration.

[0019] S104, Perform discrete event-driven simulation on the risk propagation graph at a preset time step, calculate the risk propagation increment between adjacent nodes according to the risk propagation rule parameters, and aggregate and update the multi-source risk increments received by each node to obtain the evolution result of the risk propagation graph within the simulation period. In this embodiment, the risk propagation graph that has completed initialization and parameter configuration is received. After obtaining the corresponding risk propagation rule base, the system will proceed according to the preset time step. (For example, every 5 minutes) Initiate a discrete event-driven simulation. At each time step t, the engine traverses all nodes in the risk propagation graph and determines their current risk status one by one. Does it exceed the risk trigger threshold θ, for those that meet the requirements? For node v, the system triggers a risk propagation event, causing node v to propagate risk to each of its neighboring nodes u according to the formula: Calculate and send the propagation increment, for example, the risk state of a relay station v at time t. (Exceeding θ = 0.6), the vulnerability coefficient of its downstream adjacent road segment u Network distance between v kilometers, propagation delay Step, and take , , When, the propagation increment received by segment u at time t. When neighboring node u receives data from k upstream source nodes simultaneously at time t. When the propagation increment is reached, the system follows the formula The risk status of node u at time t+1 is aggregated and updated to reflect the superposition effect of multi-path risks and ensure that the risk value is always normalized to the interval [0, 1]. If the updated risk status is... If the threshold θ is exceeded again, node u becomes a new risk source in the next time step and continues to propagate to its downstream neighboring nodes, thus forming a ripple effect of risk spreading step by step along the transportation network. For multi-path propagation scenarios with multiple parallel or circular transportation routes in the transportation network (e.g., the same goods can reach their destination via two parallel paths, "Shanghai-Hangzhou Expressway—Hangzhou-Jinhua-Quzhou Expressway" or "Shanghai-Kunming Expressway—Hangzhou-Jinhua-Quzhou Expressway"), the simulation unit further initiates a random walk propagation subprocess, according to the formula... Calculate the probability of risk shifting from node v to each adjacent node u, where the edge weights are... Based on the historical traffic volume and business relevance of the road segment (such as the transportation frequency of the same cargo owner), the risk status of all nodes is determined iteratively until the change in the overall risk value is less than a preset convergence threshold (e.g., 10). -3 This allows for a stable global risk distribution. Simultaneously, as the risk propagation graph progresses step-by-step, the simulation unit runs in parallel dedicated models mounted on each node and edge. These dedicated models include at least one of the following: a traffic flow model (for estimating road segment speed and congestion evolution), a vehicle fuel consumption model (for estimating transportation energy consumption and range), and a cargo loss model (for estimating the probability of loss for temperature-sensitive cold chain goods). Each dedicated model represents the risk state of node v at time t. It is used as one of the input parameters in the calculation (e.g., in traffic flow models). (When the estimated traffic speed is high, the estimated delay time is lowered), and the calculated output of the estimated delay time, the estimated probability of cargo loss, and the estimated energy consumption increment are fed back to the risk propagation graph as new risk factors. The risk status of the corresponding nodes and / or edges is updated a second time, thus forming a closed-loop coupled simulation between the risk graph evolution and the calculation of the special model. The above simulation continues until the preset simulation end time T is reached. The simulation unit summarizes the state evolution trajectory of each node and edge of the risk propagation graph, the detailed output of each special model, and the final global risk situation during the entire simulation cycle, forming a structured evolution result dataset and outputting it.

[0020] S105, Calculate the comprehensive evaluation index based on the evolution results, and generate an emergency dispatch plan for logistics transportation based on the comprehensive evaluation index.

[0021] In this embodiment, after receiving the structured evolution result dataset, a comprehensive evaluation index is first calculated based on the evolution results. The comprehensive evaluation index is used to quantitatively assess the risk pressure borne by the entire logistics and transportation network within the current simulation period and the level of economic losses that may result therefrom. Specifically, it includes the cumulative risk exposure. At least one of the following three factors: risk propagation range Ψ and expected economic loss Θ, wherein the cumulative risk exposure is calculated according to the formula Calculations are performed by summing the risk status of all nodes at each time step within the simulation period T to reflect the overall risk pressure borne by the logistics network throughout the entire simulation period (e.g., the simulation result Φ = 1250 under a typhoon scenario). The risk propagation range is determined by the formula... Calculations are performed to determine the proportion of nodes in a high-risk state (i.e., the risk level exceeds the trigger threshold θ) at the end of the simulation, reflecting the breadth of risk propagation (e.g., ...). (This indicates that 35% of the nodes in the entire network are in a high-risk state at the end of the simulation), and the expected economic loss is calculated according to the formula. The calculation is performed by weighting the value of goods per unit time for each edge at each time step by the risk state. The post-summation reflects the expected economic losses that network risks may cause throughout the entire simulation period (e.g., the losses in a particular simulation). (in ten thousand yuan); after the comprehensive evaluation indicators are calculated, the scheduling decision-making unit uses at least one of the above indicators as the optimization objective (such as minimizing the expected economic loss Θ as the optimization objective), and calculates it according to the formula. The optimal scheduling scheme is solved by recursively working backward from the end time T to the start time 1. For example, when state s represents "congestion on the Shanghai-Hangzhou Expressway G60-1 section and continuous rainfall", and the decision set A(s) contains three candidate decisions n1, n2, and n3: "detour via Hangzhou-Jinhua-Quzhou Expressway", "delay departure for 2 hours", and "change transportation mode to railway", the system calculates the current stage cost corresponding to each decision one by one. (e.g., additional mileage fees for detours, late payment penalties for delayed departures, transshipment costs for changing transportation methods) and post-transfer status. Optimal cumulative cost The decision that minimizes the sum of the two factors is taken as the optimal decision for this stage (e.g., the decision with the lowest total cost of the n1 detour decision is selected), and the complete emergency dispatch plan is obtained by backtracking in time sequence. After the emergency dispatch plan is generated, it is output to the unified dispatch center for feasibility assessment. For the feasible plan, dispatch instructions are directly generated and pushed to the third-party transportation management system, fleet management system, or warehouse management system via API for execution (e.g., automatically issuing detour route instructions to the driver terminal). For the infeasible plan (e.g., the detour route has been marked as high risk by the road condition system during the execution period), the reasons for the assessment are returned to the decision model for learning. After the plan is executed, the unified dispatch center collects the actual execution feedback data of this plan (e.g., actual delay time, actual loss amount, customer satisfaction score), and calculates it according to the formula. The decision model is iteratively updated using Q-learning, where the immediate feedback values... Based on the deviation between the actual and expected results (e.g., positive feedback is given if the actual loss is lower than expected, and negative feedback is given if it is higher than expected), the learning rate h typically ranges from 0.1 to 0.3, and the discount factor m typically ranges from 0.8 to 0.95. This allows the decision-making model to continuously optimize its action value function under various states in each round of the "deduction-scheme generation-execution-feedback" cycle. Ultimately, this forms a closed-loop intelligent decision-making optimization system that connects data collection, risk simulation, solution generation, execution feedback, and model iteration.

[0022] The second embodiment of the present invention provides a simulation and deduction device based on a risk propagation graph, comprising: Data governance unit 201 is used to collect multi-source heterogeneous data from the logistics and transportation system, clean, align and standardize the multi-source heterogeneous data to obtain standardized transportation data, wherein the multi-source heterogeneous data includes transportation network topology data and risk factor data. The feature modeling unit 202 is used to extract and fuse multi-dimensional features from the standardized transportation data to obtain fused risk features that characterize the current risk status of each transportation node and transportation segment. The risk graph construction unit 203 is used to construct a risk propagation graph based on the transportation network topology data in the standardized transportation data, take the fused risk features as the initial risk state of the corresponding nodes and edges in the risk propagation graph, and configure risk propagation rule parameters. The simulation and deduction unit 204 is used to perform discrete event-driven simulation and deduction on the risk propagation graph at a preset time step, calculate the propagation increment of risk between adjacent nodes according to the risk propagation rule parameters, and aggregate and update the multi-source risk increments received by each node to obtain the evolution result of the risk propagation graph within the deduction period. The decision-making unit 205 is used to calculate the comprehensive evaluation index based on the evolution results, and generate an emergency dispatch plan for logistics transportation based on the comprehensive evaluation index.

[0023] The third embodiment of the present invention provides a simulation and deduction device based on a risk propagation graph, including a memory and a processor. The memory stores a computer program, which can be executed by the processor to implement a simulation and deduction method based on a risk propagation graph as described in any of the above embodiments.

[0024] The fourth embodiment of the present invention provides a computer-readable storage medium storing a computer program, which can be executed by a processor of the device where the computer-readable storage medium is located, to implement a simulation and deduction method based on a risk propagation graph as described in any of the above embodiments.

[0025] Based on the simulation and deduction method, apparatus, and equipment based on risk propagation graph provided by this invention, standardized transportation data is obtained by collecting multi-source heterogeneous data from the logistics and transportation system and processing it through cleaning, alignment, and standardization. Multi-dimensional feature extraction and fusion of this data yields fused risk features characterizing the current risk status of each transportation node and transportation segment. On this basis, a risk propagation graph is constructed based on the transportation network topology data, and the fused risk features are used as the initial risk status of corresponding nodes and edges in the graph. Discrete event-driven simulation and deduction are then performed at a preset time step. The propagation increment of risk between adjacent nodes is calculated according to risk propagation rule parameters, and the multi-source risk increments received by each node are aggregated and updated to obtain the evolution result of the risk propagation graph within the simulation period. Finally, a comprehensive evaluation index is calculated based on this evolution result, and an emergency dispatch plan for logistics and transportation is generated accordingly. This allows for a forward-looking quantitative deduction of the risk transmission path and its superimposed impact before the actual arrival of the risk at downstream nodes, ensuring that the generation of the emergency dispatch plan precedes the actual risk transmission process.

[0026] Exemplary examples show that the computer program described in the third and fourth embodiments of the present invention can be divided into one or more modules, which are stored in the memory and executed by the processor to complete the present invention. The one or more modules can be a series of computer program instruction segments capable of performing specific functions, which describe the execution process of the computer program in implementing a simulation and deduction device based on a risk propagation graph. For example, the apparatus described in the second embodiment of the present invention.

[0027] The processor referred to can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor. The processor is the control center of the aforementioned simulation and deduction method based on risk propagation diagrams, connecting various parts of the entire simulation and deduction method using various interfaces and lines.

[0028] The memory can be used to store the computer program and / or modules. The processor, by running or executing the computer program and / or modules stored in the memory, and by calling the data stored in the memory, implements various functions of a simulation and deduction method based on a risk propagation diagram. The memory may mainly include a program storage area and a data storage area. The program storage area may store the operating system, at least one application program required for a function (such as sound playback function, text conversion function, etc.), etc.; the data storage area may store data created based on the use of the mobile phone (such as audio data, text message data, etc.). In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, memory, plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0029] If the implemented module is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the above embodiments of the present invention can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the content included in the computer-readable medium can be appropriately added or removed according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electrical carrier signals and telecommunication signals.

[0030] It should be noted that the device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Furthermore, in the accompanying drawings of the device embodiments provided by this invention, the connection relationships between modules indicate that they have communication connections, which can be specifically implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement this without any creative effort.

[0031] The above description is merely a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A simulation and deduction method based on risk propagation diagrams, characterized in that, include: Collect multi-source heterogeneous data from the logistics and transportation system, clean, align, and standardize the multi-source heterogeneous data to obtain standardized transportation data. The multi-source heterogeneous data includes transportation network topology data and risk factor data. Multi-dimensional feature extraction and fusion are performed on the standardized transportation data to obtain fused risk features that characterize the current risk status of each transportation node and transportation segment; A risk propagation graph is constructed based on the transportation network topology data in the standardized transportation data. The fused risk features are used as the initial risk states of the corresponding nodes and edges in the risk propagation graph, and risk propagation rule parameters are configured. Discrete event-driven simulations are performed on the risk propagation graph at preset time steps. The propagation increment of risk between adjacent nodes is calculated according to the risk propagation rule parameters. The multi-source risk increments received by each node are aggregated and updated to obtain the evolution result of the risk propagation graph within the simulation period. Based on the evolution results, a comprehensive evaluation index is calculated, and an emergency dispatch plan for logistics transportation is generated according to the comprehensive evaluation index.

2. The simulation and deduction method based on risk propagation diagrams according to claim 1, characterized in that, The process of cleaning, aligning, and standardizing the multi-source heterogeneous data to obtain standardized transport data specifically involves: For continuous time series data in the aforementioned multi-source heterogeneous data, linear interpolation is used for data interpolation processing. The interpolation calculation formula is as follows: Where t represents the time point to be interpolated. Let be a known time point that is before t and adjacent to t. Let be a known time point that is after t and adjacent to t. and These represent the points in time. and The known observations, This represents the interpolation result at time point t; For discrete attribute data in the multi-source heterogeneous data, the nearest neighbor value filling method is used for filling; To address the issue of inconsistent sampling frequencies among different data sources in the multi-source heterogeneous data, time alignment and spatial alignment processing are performed. The data, after undergoing interpolation, padding, time alignment, and spatial alignment, is normalized through a pre-defined standardized API interface to ensure that the data from each data source has a unified field naming, data type, and event description structure, thus obtaining the standardized transportation data.

3. The simulation and deduction method based on risk propagation diagrams according to claim 2, characterized in that, The spatial alignment process includes map matching and grid matching, wherein: For the location and direction of travel data of transportation vehicles in the multi-source heterogeneous data, map matching based on the hidden Markov model is used, and the observation probability is... Calculate using the following formula: in, Let be the GPS observation point at time t. Candidate road sections for On the road section The projection point on the surface, This represents the standard deviation of GPS error. Matching road segments from the previous moment Move to the current road segment transition probability Calculate using the following formula: in, This represents the shortest path distance within the road segment network. The straight-line distance between GPS points. For scale parameters, The GPS observation point at time t-1; For the meteorological, air quality, and temperature data collected according to geographic grids in the aforementioned multi-source heterogeneous data, a bilinear interpolation method is used for grid matching: in, , These are the normalized coordinates of the GPS point within its respective grid. , , , These are the observations at the four vertices of the grid. This is the output value after grid matching.

4. The simulation and deduction method based on risk propagation diagrams according to claim 1, characterized in that, The process of extracting and fusing multi-dimensional features from the standardized transportation data to obtain fused risk features is as follows: Extract feature vectors from K data sources from the standardized transportation data, categorized by transportation node and transportation segment. ; The feature vectors of the K data sources are fused using at least one of the following fusion methods: Feature vector concatenation method: The feature vectors from the K data sources are directly concatenated into a long vector. ; Weighted fusion method: in, The feature vector extracted from the i-th data source. The fused feature vector is formed by concatenating the K feature vectors end to end. The weight of the i-th data source and satisfying The fusion risk characteristics are obtained and persistently stored.

5. The simulation and deduction method based on risk propagation diagrams according to claim 1, characterized in that, A risk propagation diagram is constructed based on the transportation network topology data in the standardized transportation data, specifically as follows: A graph structure is constructed based on the transportation network topology data. , where V is a set of nodes, the nodes include at least one of cities, transit stations, intersections and waypoints, and E is a set of edges, the edges represent transportation segments between nodes; The feature corresponding to node v in the fused risk features is used as the initial risk state of node v. The features corresponding to the edges are used as attribute parameters of the edges; The risk propagation rule parameters include: a risk decay coefficient, which characterizes the rate attenuation of risk with propagation distance. Risk trigger threshold used to determine whether a node triggers risk propagation The propagation intensity coefficient characterizes the impact of the risk source on adjacent nodes. And the node vulnerability coefficient, which characterizes the node's sensitivity to incoming risks. .

6. The simulation and deduction method based on risk propagation diagrams according to claim 5, characterized in that, The risk propagation increment between adjacent nodes is calculated based on the risk propagation rule parameters, specifically as follows: At each time step Traverse all nodes in the risk propagation graph; when node v is in risk state at time t... Exceeding the aforementioned risk trigger threshold When this occurs, node v is triggered to propagate the risk to all its neighboring nodes u, with the propagation increment... Calculate using the following formula: in, For the propagation intensity coefficient, Let be the node vulnerability coefficient of the adjacent node u. This is the risk attenuation coefficient. This represents the spatial or network distance between node v and node u. The time decay factor has a value between 0 and 1. The time delay required for the risk to propagate from node v to node u; Node u receives data from the source node After multiple propagation increments, update its risk status using the following formula: in, Let's consider the risk state of node u at time t. For the i-th source node The propagation increment generated to node u at time t, when Exceeding the aforementioned risk trigger threshold hour, Given the risk state at time t+1, node u becomes a new risk source and propagates outward in the next time step.

7. The simulation and deduction method based on risk propagation diagrams according to claim 1, characterized in that, The comprehensive evaluation indicators include at least one of the following three: Cumulative risk exposure : Where T is the total duration of the simulation cycle, and V is the set of nodes. The risk state of node v at time t; Scope of risk transmission : Where |V| is the total number of nodes in the node set V. Let V be the risk state of node v at the end of the simulation at time T. As the risk trigger threshold, () is an indicator function that takes the value 1 when the condition inside the parentheses is true, and takes the value 0 otherwise; Expected economic loss : Where E is the set of edges, Let e ​​be the unit time value of the goods on the transport segment corresponding to edge e. Let e ​​be the risk state of edge e at time t.

8. The simulation and deduction method based on risk propagation diagrams according to claim 1, characterized in that, The comprehensive evaluation indicators generate an emergency dispatch plan for logistics transportation, specifically as follows: Using at least one of the comprehensive evaluation indicators as the optimization objective, a multi-stage decision-making model is constructed, and the optimal scheduling scheme is solved recursively using the following formula: Where s represents the state of the transportation system at stage t. Let s be the set of allowed scheduling decisions in state s, and n be the set of decisions. One of the scheduling decisions, The stage cost incurred in making decision n in stage t at state s. To determine the state in the next stage after making decision n. The optimal cumulative cost is calculated from state s in stage t to the end of the simulation period. To be in the next stage t+1 from state The optimal cumulative cost incurred from departure to the end of the simulation period. Let n be the state transition function, representing the state reached by the transportation system after making decision n in state s at stage t and transitioning to the next stage t+1; The execution feedback data of the emergency dispatch plan is fed back, and the decision model is iteratively updated using Q-learning according to the following formula: in, Let t be the state of the transportation system. The scheduling decision made at time t. For state Make a decision Action value function, For learning rate, To execute scheduling decisions Immediate feedback obtained from the environment afterwards For future reward discount factors, To execute scheduling decisions The state of the system at time t+1, For state All possible scheduling decisions to be made. For state The maximum action value among all possible decisions.

9. A simulation and deduction device based on a risk propagation diagram, characterized in that, include: The data governance unit is used to collect multi-source heterogeneous data from the logistics and transportation system, clean, align, and standardize the multi-source heterogeneous data to obtain standardized transportation data. The multi-source heterogeneous data includes transportation network topology data and risk factor data. The feature modeling unit is used to extract and fuse multi-dimensional features from the standardized transportation data to obtain fused risk features that characterize the current risk status of each transportation node and transportation segment. The risk graph construction unit is used to construct a risk propagation graph based on the transportation network topology data in the standardized transportation data, take the fused risk features as the initial risk state of the corresponding nodes and edges in the risk propagation graph, and configure risk propagation rule parameters. The simulation and deduction unit is used to perform discrete event-driven simulation and deduction on the risk propagation graph at a preset time step, calculate the propagation increment of risk between adjacent nodes according to the risk propagation rule parameters, and aggregate and update the multi-source risk increments received by each node to obtain the evolution result of the risk propagation graph within the deduction period. The decision-making unit is used to calculate the comprehensive evaluation index based on the evolution results, and generate an emergency dispatch plan for logistics transportation based on the comprehensive evaluation index.

10. A simulation and deduction device based on a risk propagation diagram, characterized in that, The system includes a memory and a processor. The memory stores a computer program that can be executed by the processor to implement a simulation and deduction method based on a risk propagation graph as described in any one of claims 1 to 8.

Images