A charging pile site selection method and system

Abstract

The present application relates to a kind of charging pile site selection method and system.The method obtains the topography of target rural area, road trajectory, historical residence and energy supplement behavior, meteorological environment, energy supply and safety risk data, constructs rural space-time graph and candidate node set;Respectively, energy supplement demand intensity model, comprehensive accessible cost model and deployment feasibility model are established;The above results are input into space-time prediction model, and the demand income prediction value and risk disturbance prediction value of candidate node are obtained;Further, the robust site selection optimization model with budget constraint, service constraint and risk opportunity constraint is constructed, and the charging facility site selection scheme is solved;And through load disturbance simulation and extreme environment simulation, the site selection scheme is feedback corrected.This method can improve the demand matching ability, engineering implementability and robustness under complex environment of rural charging facility site selection.

Description

Technical Field

[0001] This invention relates to the field of charging facility planning and intelligent optimization technology, and in particular to a charging pile site selection method and system. Specifically, it relates to an intelligent site selection method and system for charging facilities that is designed for complex rural environments and combines spatiotemporal demand prediction, accessibility assessment, deployment feasibility analysis, risk opportunity constraint modeling, and robust optimization solution. Background Technology

[0002] With the increasing application of electric equipment, mobile work platforms, unmanned transportation equipment, rural inspection equipment, and portable energy storage terminals in rural scenarios, rural charging capacity is gradually becoming a crucial factor affecting continuous operation capability, energy replenishment efficiency, and task execution stability. Compared to urban or fixed industrial park scenarios, rural areas typically feature undulating terrain, complex road conditions, difficult energy access, strong environmental disturbances, rapid weather changes, and high uncertainty in safety risks. Therefore, the site selection for rural charging facilities cannot simply follow the conventional urban charging station deployment approach.

[0003] Existing charging facility site selection methods mostly focus on site planning under urban road networks, fixed traffic flow, and stable power grid access conditions. Their basic approach typically involves estimating charging demand based on historical demand data, then combining this with distance, cost, or coverage indicators to determine charging station layout using clustering algorithms, integer programming models, or heuristic algorithms. However, these methods still have the following shortcomings in rural scenarios:

[0004] First, existing methods are usually based on static historical statistical data, which makes it difficult to accurately reflect the dynamic fluctuations in energy replenishment demand in rural scenarios caused by changes in tasks, personnel stay, temporary route changes and weather disturbances, resulting in insufficient adaptability of site selection results to time-varying demand.

[0005] Secondly, existing methods do not adequately consider factors such as terrain slope, road resistance, differences in accessibility, and complex environmental factors, and often fail to effectively characterize the real access costs between rural demand nodes and candidate deployment points, thus affecting the engineering feasibility of the site selection results.

[0006] Third, existing methods do not adequately consider risk factors such as energy supply conditions, construction and maintenance conditions, extreme weather, geological disasters, communication loss and energy interruption. They lack unified modeling of deployment feasibility and continuous service capability under uncertain environments, which may lead to the theoretically optimal solution failing to be implemented or having poor operational stability in actual deployment.

[0007] Fourth, existing technologies mostly adopt a single offline planning method, which lacks an online correction mechanism based on simulation feedback or real operation feedback after the site selection results are formed. It is impossible to dynamically adjust the model parameters and site selection strategies according to load deviation, coverage deviation and risk changes.

[0008] Therefore, there is an urgent need to provide a charging pile site selection method and system to solve the problems of insufficient adaptability to rural scenarios, inaccurate spatiotemporal demand characterization, insufficient consideration of risks in complex environments, and low robustness of site selection results in existing technologies. Summary of the Invention

[0009] The purpose of this invention is to provide a charging pile site selection method and system to solve the problem that the existing technology does not adequately consider complex terrain environment, dynamic energy replenishment demand, energy access conditions and multi-source risk factors in the site selection of rural charging facilities, thereby improving the demand matching ability, feasibility, environmental adaptability and operational robustness of the site selection results.

[0010] To achieve the aforementioned objective, this invention provides a method for selecting the location of a charging pile, comprising:

[0011] S1. Obtain relevant data for the target rural area, perform grid-based discretization on the target rural area, and construct a candidate node set. Set of node connection relationships To form a rural time and space map ;

[0012] S2, Based on the candidate node set And the corresponding historical trajectory, dwell information, refueling behavior information and task scenario information, to calculate each candidate node At any moment Energy replenishment demand intensity ;

[0013] S3. Based on the aforementioned rural spatiotemporal map The required nodes are calculated based on road connectivity, terrain raster attributes, and meteorological environment attributes. to candidate node The overall attainable cost ;

[0014] S4. Based on the energy attributes, construction and maintenance convenience, cost conditions, and risk conditions of the candidate nodes, calculate the parameters for each candidate node. Deployment feasibility ;

[0015] S5, Intensity of Energy Replenishment Demand Comprehensive achievable cost and deployment feasibility As a joint input, it is fed into the spatiotemporal prediction model to obtain candidate nodes. Demand and revenue forecasts With risk disturbance forecast ;

[0016] S6. Construct a robust addressing optimization model based on the following formula:

[0017]

[0018] in, To comprehensively optimize the objective function, Candidate nodes Location decision variables, ; For service; fairness indicators, To optimize the trade-off coefficients; For construction costs;

[0019] And satisfy:

[0020]

[0021] in, This is the upper limit of the total budget;

[0022]

[0023] in, For demand nodes The maximum acceptable service cost;

[0024]

[0025] in, The maximum permissible risk threshold, This refers to the risk tolerance coefficient.

[0026] S7, to Perform load disturbance simulation and extreme environment simulation. When the simulation results do not meet the preset threshold, update the spatiotemporal prediction model and the robust location optimization model based on the simulation feedback results, and re-execute step S6 until the final charging facility location result is output.

[0027] To achieve the aforementioned objective, the present invention also provides a charging pile site selection system, comprising:

[0028] The data construction module is used to acquire terrain raster data, road trajectory data, historical stay and refueling behavior data, meteorological and environmental data, energy supply data, and safety risk data of the target rural area, and to perform grid-based discretization of the target rural area to construct a candidate node set. Node connection set and rural time and space map ;

[0029] The requirement modeling module is used to model based on the candidate node set. And the corresponding historical trajectory, dwell information, refueling behavior information and task scenario information, to calculate each candidate node Energy replenishment demand intensity ;

[0030] The reachability cost modeling module is used to calculate the comprehensive reachability cost from demand nodes to candidate nodes based on road connectivity, terrain grid attributes, and meteorological environment attributes. ;

[0031] Deploy a feasibility modeling module to calculate the feasibility of each candidate node based on its energy attributes, ease of construction and maintenance, cost conditions, and risk conditions. Deployment feasibility ;

[0032] The prediction module takes the energy replenishment demand intensity, overall reachability cost, and deployment feasibility as joint inputs and feeds them into the spatiotemporal prediction model to obtain the demand and revenue prediction values ​​for candidate nodes. With risk disturbance forecast ;

[0033] A robust optimization module is used to optimize based on the predicted demand and revenue values. Risk disturbance prediction value Comprehensive achievable cost and construction costs A robust site selection optimization model with budget constraints, service constraints, and risk opportunity constraints is constructed and solved to obtain a site selection scheme for charging facilities.

[0034] The feedback correction module is used to perform load disturbance simulation and extreme environment simulation on the charging facility site selection scheme, and update the spatiotemporal prediction model and the robust site selection optimization model when the simulation results do not meet the preset threshold.

[0035] The output module is used to output the final location results of the charging facilities.

[0036] Compared with the prior art, the present invention has at least the following beneficial effects:

[0037] 1. Improves the accuracy of rural energy replenishment demand characterization: This invention constructs an energy replenishment demand intensity model, which jointly models factors such as historical trajectory, dwell time, node popularity and task type. This enables the site selection process to not only rely on static historical statistical results, but also to more accurately reflect the real energy replenishment demand distribution in rural scenarios over time, thereby improving the matching degree between site selection results and actual needs.

[0038] 2. Improves the adaptability of site selection results to complex rural traffic conditions: This invention constructs a comprehensive reachability cost model, which incorporates path distance, slope, road condition impedance, and environmental additional impedance into a unified evaluation framework. This enables the site selection results to reflect the actual traffic difficulty in rural areas and avoids the problem that the theoretically optimal site deployment location may be difficult to reach in practice due to only considering geometric distance.

[0039] 3. Improves the engineering feasibility of charging facility deployment schemes: This invention constructs a deployment feasibility model to comprehensively evaluate energy availability, construction and maintenance convenience, construction costs, and risk penalties, thereby making the selection and decision-making process of candidate nodes closer to actual deployment conditions and improving the feasibility and implementability of the final site selection scheme in rural scenarios.

[0040] 4. Improves the robustness of site selection schemes under uncertain environments: This invention constructs a robust site selection optimization model by introducing demand and revenue forecasts, risk disturbance forecasts, budget constraints, service constraints, and risk opportunity constraints. This enables the final scheme to maintain good service coverage and operational stability under disturbance scenarios such as extreme weather, path changes, and energy outages, thereby improving the adaptability of the site selection scheme to complex rural environments.

[0041] 5. Improves service balance between different regions: This invention introduces a service fairness index into the optimization objective, which not only considers the overall coverage benefits, but also the dispersion of service costs between different demand nodes, thereby avoiding the problem of over-concentration in some areas and insufficient service in other areas, and improving the overall coordination and fairness of charging facility layout.

[0042] 6. Enables dynamic optimization and continuous correction of site selection schemes: This invention sets up load disturbance simulation and extreme environment simulation, and uses the simulation feedback results to update the parameters of the spatiotemporal prediction model and the robust optimization model, thereby forming a closed-loop adaptive site selection mechanism. This allows the site selection scheme to be dynamically optimized as demand changes, environmental changes and load changes occur, thereby improving long-term operating performance.

[0043] 7. Enhances the intelligence level of algorithm decision-making: This invention combines rural spatiotemporal map modeling, spatiotemporal prediction, robust optimization, and feedback correction to form an intelligent algorithm framework for rural charging facility site selection. This enables the system to not only output static site selection results, but also achieve higher-quality intelligent decision-making under complex constraints, multi-source disturbances, and dynamic demand conditions. Attached Figure Description

[0044] Figure 1 This is a flowchart of the charging pile site selection method provided by the present invention. Detailed Implementation

[0045] Figure 1 This is a flowchart of the charging pile site selection method provided by the present invention, such as... Figure 1 As shown, the charging pile site selection method provided by the present invention is executed by a processor and is applicable to rural operation areas, mountain patrol areas, forest protection areas, remote camp areas, disaster emergency protection areas, or other areas with complex terrain, complex road conditions, limited energy access, and significant environmental disturbances.

[0046] S1. Construction of Rural Spatiotemporal Map and Generation of Candidate Nodes

[0047] Specifically, this includes acquiring topographic raster data, road trajectory data, historical dwelling and refueling behavior data, meteorological and environmental data, energy supply data, and safety risk data for the target rural area. Among these, the topographic raster data includes at least elevation, slope, aspect, surface roughness, and landform type; the road trajectory data includes at least road grade, historical travel trajectory, travel frequency, and transfer route; the historical dwelling and refueling behavior data includes at least dwelling location, dwelling duration, historical refueling frequency, and refueling time distribution; the meteorological and environmental data includes at least rainfall, wind speed, temperature, snow accumulation, and information on mud or water wading; the energy supply data includes at least grid connection information, mobile energy storage deployment capacity, distributed photovoltaic conditions, or hybrid energy supply conditions; and the safety risk data includes at least information on landslides, collapses, floods, fire hazards, communication outages, and energy interruptions.

[0048] The target rural area is discretized into a grid, dividing the target area into multiple spatial units. A set of candidate nodes is determined based on deployable locations, path intersections, hotspot locations, energy access locations, or combinations thereof.

[0049]

[0050] in, For the set of candidate nodes, For the first Candidate nodes, This represents the total number of candidate nodes.

[0051] Furthermore, a set of inter-node connection relationships is constructed based on road connectivity, drivability, energy coupling, and risk association. This creates a rural spatiotemporal map:

[0052]

[0053] in, A time-space map of the countryside. For the set of candidate nodes, It is a set of node connection relationships.

[0054] In one embodiment, spatial units can be initially screened based on historical trajectory density, dwell time, and terrain deployability. Then, a second screening can be conducted based on power supply accessibility, road accessibility, and risk ceiling, thereby eliminating unconstructable areas, severely high-risk areas, and obviously low-value areas, resulting in a set of candidate nodes for subsequent steps. Rural Time and Space Map .

[0055] The output data for this step includes: a set of candidate nodes. Node connection set Rural Time and Space Map The output data includes the terrain attributes, path attributes, energy attributes, and risk attributes of each candidate node. This output data is then used in steps S2, S3, and S4, respectively.

[0056] In this step, the present invention transforms the originally scattered terrain, road, energy and risk information in the rural scene into a computable, associative and disseminable data structure by uniformly representing the target area through grid and graph structure. This provides a unified data foundation for subsequent demand modeling, accessibility modeling, feasibility modeling and joint optimization, thereby improving the data consistency, engineering adaptability and computational feasibility of the entire site selection process.

[0057] S2, Energy Demand Intensity Modeling

[0058] Specifically, this includes: the set of candidate nodes output from step S1. Based on its corresponding historical trajectory, dwell information, refueling behavior information, and mission scenario information, each candidate node is calculated. At any moment The intensity of energy replenishment demand. This intensity of energy replenishment demand can be expressed as:

[0059]

[0060] in, Candidate nodes At any moment The intensity of energy replenishment demand, Candidate nodes At any moment The node heat index is used to characterize the intensity of trajectory convergence or activity frequency near the candidate node. Candidate nodes At any moment Average length of stay; Candidate nodes At any moment Frequency of historical refueling events; Candidate nodes At any moment The task type or scenario intensity factor is used to characterize the degree of impact of different tasks such as inspection, transportation, rescue, stationing or temporary operations on energy replenishment demand; Model weighting coefficients for demand.

[0061] In one embodiment, the node heat The average dwell time can be obtained by combining the number of trajectories passing through the buffer zone around the candidate node, the number of dwell times, and the average activity density within a unit time window; The frequency of the historical recharge events can be obtained from the time statistics of historical dwell segments around the candidate node; It can be obtained by segmenting and statistically analyzing historical recharge records on a timeline; the task type or scene intensity factor. It can be mapped based on the task plan, operation level, emergency level, or regional support level.

[0062] Furthermore, demand intensity can be calculated at different time granularities to form the time-series demand characteristics of candidate nodes:

[0063]

[0064] in, Candidate nodes The sequence of energy replenishment demand intensity within a preset time window This represents the number of time steps.

[0065] The output data for this step includes: the energy replenishment demand intensity of each candidate node. and replenishment demand intensity sequence Its output will be used as one of the inputs for step S5.

[0066] This step directly calls upon the candidate node spatial locations, historical trajectory density, dwell records, historical refueling events, and task region labels from step S1 to generate time-dependent demand intensity modeling results. The output of this step is... and Together they constitute the demand-side input for step S5.

[0067] In this step, the present invention integrates trajectory heat, dwell behavior, refueling history and task intensity to improve the traditional static heat distribution into a refueling demand characterization result with time resolution. This enables subsequent prediction and optimization models to make site selection decisions based on dynamic demand, thereby improving the matching degree between the site selection result and the actual rural refueling demand.

[0068] S3, Comprehensive Accessibility Cost Modeling

[0069] Specifically, this includes: the set of candidate nodes output from step S1. Road connectivity, terrain grid attributes, meteorological environmental attributes, and the set of demand nodes are considered for any given demand node. With candidate nodes The service paths between them are searched, and the comprehensive reachability cost is calculated. The comprehensive reachability cost can be expressed as:

[0070]

[0071] in, For demand nodes to candidate node The overall attainable cost, For demand nodes to candidate node Path distance, For demand nodes to candidate node The cost of the slope, For demand nodes to candidate node Road condition impedance, For demand nodes to candidate node Meteorological or environmental additional impedance, Weighting coefficients are used to model accessibility.

[0072] Among them, path distance It can be determined by the shortest path length of the road network, the off-road passable length, or the overall passable length; gradient cost. The road condition impedance can be obtained from the average slope, maximum slope, or cumulative slope value of the grids traversed along the path. The impedance can be determined based on road grade, road surface conditions, water wading depth, mud level, turning complexity, and traffic bottleneck severity; meteorological or environmental additional impedance. It can be determined based on factors such as rainfall, snow accumulation, icing, rockfall warnings, or decreased visibility.

[0073] In one embodiment, different accessibility parameter libraries can be set for different types of vehicles or different power supply methods. For example, for light electric inspection equipment, mobile energy storage vehicles, and fixed construction support equipment, different path passability thresholds, slope tolerance thresholds, and wading capabilities can be set respectively, thereby obtaining differentiated comprehensive accessibility cost matrices.

[0074]

[0075] in, To synthesize the reachability cost matrix, The total number of nodes required. This represents the total number of candidate nodes.

[0076] The output data for this step includes: the combined reachability cost between each demand node and each candidate node. and the comprehensive reachability cost matrix Its output will be used for joint prediction in step S5 and for service constraint and fairness calculation in step S6.

[0077] This step directly uses the road, terrain, weather, and risk map data from step S1 to form the path cost structure; the comprehensive reachability cost matrix in this step... As the spatial service-side input for step S5; this step It is also used for constraints, site allocation, and fairness evaluation.

[0078] In this step, the present invention expands the simple geometric distance into a comprehensive service cost that takes into account slope, road conditions and environmental factors, which can more realistically reflect the actual difficulty of "accessibility" and "serviceability" in rural areas, thereby avoiding the deployment deviation caused by site selection based solely on straight-line distance or ordinary road distance, and improving the actual usability and service accessibility of the site selection results.

[0079] S4. Deployment Feasibility Modeling

[0080] S4 specifically includes: based on the energy attributes, construction and maintenance convenience, cost conditions, and risk conditions of the candidate nodes output in step S1, for each candidate node... The deployment feasibility is calculated. The deployment feasibility can be expressed as:

[0081]

[0082] in, Candidate nodes Deployment feasibility, Candidate nodes Energy availability Candidate nodes Convenience of construction and operation and maintenance Candidate nodes Construction costs, Candidate nodes Risk penalty items, Weighting coefficients are used to model feasibility.

[0083] Among them, energy availability Can be used to characterize candidate nodes The feasibility of power supply when using grid connection, mobile energy storage, photovoltaic energy storage, or hybrid energy supply; ease of construction and operation and maintenance. It can be used to characterize the ease of equipment transportation, ground flatness, construction difficulty, maintenance accessibility, and ease of construction and operation; construction cost. This should at least include equipment purchase costs, civil engineering costs, transportation and installation costs, and operation and maintenance costs; risk penalty items. It can be represented as:

[0084]

[0085] in, Risk penalty item for candidate nodes, The geological hazard risk value of the candidate node. The extreme weather risk value for candidate nodes. This represents the risk value of communication failure for candidate nodes. The energy outage risk value for candidate nodes. , , , This is the risk weighting coefficient.

[0086] In one embodiment, the normalized demand intensity, normalized deployment feasibility, and normalized risk penalty term can be calculated for each candidate node first, and then a pre-screening score can be performed:

[0087]

[0088] in, Candidate nodes Pre-screening scores Candidate nodes Normalized demand intensity Candidate nodes Normalization deployment feasibility, Candidate nodes Normalized risk penalty items, These are the pre-screening weighting coefficients. The top performers are selected based on their pre-screening scores, from highest to lowest. The candidate nodes are the focus of optimization in subsequent steps S5 and S6, in order to reduce the solution scale and improve the solution efficiency.

[0089] The output data for this step includes: Deployment Feasibility. Risk penalty items Construction costs And the pre-screened subset of candidate nodes; its output will be used as one of the inputs to step S5, and as the basic data for optimizing costs and risks in step S6.

[0090] This step directly uses the energy supply data, construction and maintenance convenience data, and safety risk data from step S1 to form the deployment feasibility modeling results; the output of this step... , and As a feasibility and risk input for step S5, and in step S5 They jointly enter the objective function and constraints.

[0091] In this step, the present invention incorporates "whether it can be built, whether it is easy to build, whether the construction is cost-effective, and whether it is stable after deployment" into the deployment feasibility calculation framework. This allows for the elimination of candidate nodes that are obviously unfeasible or too costly before optimization, thereby improving the effectiveness of subsequent prediction and optimization and making the final site selection result closer to the actual engineering deployment conditions in rural scenarios.

[0092] S5, Spatiotemporal Prediction Model Construction, Training, and Joint Prediction

[0093] Specifically, this includes: the energy replenishment demand intensity output in step S2. and demand intensity sequence The comprehensive reachability cost matrix output in step S3 Deployment feasibility output in step S4 Construction costs and risk penalty items Perform concatenation encoding to form candidate nodes. Input feature vector:

[0094]

[0095] in, Candidate nodes At any moment The input feature vector, Candidate nodes The average comprehensive achievable cost for each demand node, the It can be obtained from the comprehensive reachability cost matrix Statistics show that...

[0096] Based on rural spatiotemporal maps and input feature vector A spatiotemporal prediction model is constructed to predict the demand and revenue forecasts for each candidate node within a preset period. With risk disturbance forecast In one embodiment, the spatiotemporal prediction model employs a fusion structure of graph neural networks and temporal networks, wherein the graph neural network is used to learn the spatial correlation features between candidate nodes, and the temporal network is used to learn the temporal evolution features of energy replenishment demand; the input of the graph neural network includes a rural spatiotemporal map. The system includes node features, and outputs spatial aggregation features. The temporal network can employ Long Short-Term Memory (LSTM) networks, gated recurrent unit (GRU) networks, temporal convolutional networks, or attention-based temporal coding networks to output temporal prediction results. Finally, fully connected prediction heads are used to output the results. and .

[0097] The training samples for the spatiotemporal prediction model can be composed of "node characteristics - real revenue - real risk" pairs within historical time periods. The real revenue label can be determined by historical coverage demand, historical service volume, historical replenishment frequency, historical revenue discount values, or a combination thereof; the real risk label can be determined by historical failure rate, extreme weather interruption rate, communication interruption rate, energy interruption rate, or a combination thereof. A joint loss function can be used during training.

[0098]

[0099] in, For the joint loss function, Forecast losses for demand revenue. To predict losses due to risk disturbances, This is the loss weighting coefficient.

[0100] In one embodiment, the demand revenue forecast loss Mean squared error loss, mean absolute error loss, or Huber loss can be used; the risk disturbance prediction loss Mean squared error loss, cross-entropy loss, or quantile regression loss can be used. During training, historical datasets can be divided into training, validation, and test sets in chronological order. A combination of offline pre-training and online incremental updates can be used, first training the model parameters offline using historical data, and then fine-tuning online using recent simulation or running samples in the feedback phase of step S7. To improve training stability, learning rate decay, gradient clipping, early stopping mechanisms, and regularization strategies can be employed.

[0101] After the model is trained, joint prediction is performed on each candidate node within a preset period, and the output is:

[0102] ,

[0103] ,

[0104] in, This is the set of predicted demand and revenue values ​​for candidate nodes. This is the set of predicted risk disturbance values ​​for candidate nodes.

[0105] The output data for this step includes: demand and revenue forecasts. Risk disturbance prediction value And the corresponding candidate node prediction set; its output will serve as the core input for step S6.

[0106] This step involves jointly encoding the intensity of energy replenishment demand, the overall reachability cost, and the deployment feasibility to achieve the fusion of demand-side, service-side, and deployment-side information in the prediction model. and The process directly enters the comprehensive objective function and opportunity constraints, and the obtained demand and revenue forecasts and risk disturbance forecasts are directly input into the comprehensive objective function and risk opportunity constraints in step S6.

[0107] This invention uses rural spatiotemporal maps, time series information, and multi-source deployment information to jointly predict candidate nodes. It can estimate the service benefits and risk levels of different candidate nodes in future periods before solving the site selection problem. This makes the optimization process no longer limited to historical static statistical results, but can make forward-looking site selection based on future expected demand and future disturbance risks, thereby improving the accuracy, adaptability and robustness of the site selection results.

[0108] S6, Robust Location Optimization Solution

[0109] Specifically, this includes: the demand and revenue forecast value output from step S5. With risk disturbance forecast Combined with the comprehensive attainable cost output in step S3 And the construction cost output in step S4 A robust location optimization model is constructed with the objectives of maximizing demand coverage benefits, minimizing service cost dispersion, minimizing construction costs, and minimizing risk exposure. The comprehensive objective function can be expressed as:

[0110]

[0111] in, To comprehensively optimize the objective function, Candidate nodes The location decision variables, when candidate nodes When the site is selected ,otherwise ; Candidate nodes Demand and revenue forecasts; To serve fairness indicators; Candidate nodes Construction costs; Candidate nodes The predicted value of risk disturbances; To optimize the trade-off coefficients.

[0112] Among them, service fairness indicators It can be represented as:

[0113]

[0114] in, To serve fairness indicators, The total number of nodes required. For demand nodes The cost of service to the allocated charging facilities, The average cost of serving all demand nodes. To prevent tiny positive numbers with a denominator of zero.

[0115] The robust location optimization model also satisfies budget constraints:

[0116]

[0117] in, This is the upper limit of the total budget.

[0118] The robust location optimization model also satisfies the service constraint:

[0119]

[0120] in, For demand nodes The maximum acceptable service cost.

[0121] The robust location optimization model also satisfies the risk-opportunity constraint:

[0122]

[0123] in, The maximum permissible risk threshold, This represents the risk tolerance coefficient.

[0124] In one embodiment, a set of perturbation scenarios can be further constructed. Calculate the availability probability of candidate nodes under different extreme weather, road disruption, energy fluctuation, and communication disruption scenarios:

[0125]

[0126] in, Candidate nodes The probability of availability, The number of perturbation scenarios. Candidate nodes In the scene The risk value below, For the characteristic function, when candidate nodes When the risk value in scenario s does not exceed a preset threshold, the indicator function is set to 1; otherwise, it is set to 0. This serves as a criterion for selecting candidate nodes.

[0127] In terms of solution methods, genetic algorithms, particle swarm optimization, Lagrange relaxation algorithms, mixed integer programming algorithms, reinforcement learning algorithms, or combinations thereof can be used. When using reinforcement learning algorithms, the current set of selected nodes, remaining budget, current coverage rate, and risk state can be used as the state, selecting a candidate node can be used as the action, and a combination of coverage gain, fairness improvement, cost penalty, and risk penalty can be used as the reward. The algorithm is trained in an offline simulation environment, and then the nearest optimal addressing scheme is output.

[0128] The output data for this step includes: the location scheme for the final or current iteration. The outputs of the following data will be used as inputs to step S7: the distribution relationship between demand nodes and charging facilities, overall coverage, overall cost, and overall risk assessment results.

[0129] This step directly calls the output of step S5. and As an important component of the objective function and constraints, this step is coupled with steps S3 and S4 in the following way: this step simultaneously calls... and This process establishes fairness assessments, budget constraints, and service constraints; the site selection scheme and related indicators in this step serve as inputs for subsequent simulation verification.

[0130] This invention incorporates benefits, fairness, cost, and risk into a unified optimization framework and combines budget constraints, service constraints, and risk opportunity constraints for joint solution. This enables the formation of site selection schemes that take into account constructability, serviceability, and sustainability under complex rural conditions, thereby improving the overall performance and practical application value of the final deployment results.

[0131] S7, Simulation Verification and Feedback Update

[0132] Specifically, this includes: inputting the location scheme obtained in step S6 into the simulation environment, and performing load disturbance simulation, extreme environment simulation, and path redistribution simulation on the location scheme. The load disturbance simulation is used to verify the service stability of the location scheme under conditions of demand growth, demand shift, or local peak load; the extreme environment simulation is used to verify the availability and resilience of the location scheme under conditions of heavy rain, snow accumulation, landslides, communication interruptions, energy outages, or road blockages; and the path redistribution simulation is used to verify the ability of demand nodes to resume service when local nodes fail or paths fail.

[0133] The actual demand and benefit value of each candidate node or selected node is obtained based on the simulation results. and true risk value The predicted value is then compared with the value predicted in step S5; if the coverage, availability, or risk indicators do not meet the preset thresholds, the prediction results are then subject to rolling time-domain correction.

[0134] ,

[0135] ,

[0136] in, This is the updated demand and revenue forecast. This is the updated predicted value of risk disturbance. and These are the original demand and revenue forecasts and the original risk disturbance forecasts, respectively. and These represent the actual demand and benefit values ​​and the actual risk values ​​obtained from simulation feedback, respectively. To update the step size.

[0137] Furthermore, simulation feedback samples can be added to the training sample set in step S5 to perform online fine-tuning of the spatiotemporal prediction model; simultaneously, the tradeoff coefficients in step S6 can be adjusted based on the actual load overflow rate, actual service mismatch rate, and actual risk exceedance rate in the simulation feedback. Adjust the budget allocation strategy or risk-opportunity constraint boundary, and re-execute the robust site selection solution in step S6. If the simulation results meet the preset threshold, output the final charging facility site selection results.

[0138] The output results should include at least the location of the selected target node, the facility type, power level, power supply method, service radius, and phased deployment sequence of each target node.

[0139] This step directly calls the site selection scheme and its site configuration output in step S6 for simulation; this step sends the new samples generated by the simulation feedback back to step S5 to perform online correction of the prediction model; this step sends the feedback prediction results and constraint parameters back to step S6 to form a re-optimization.

[0140] This invention effectively avoids the failure of a one-time static site selection scheme in a real complex environment by adding simulation verification for disturbance scenarios after site selection is completed and using simulation feedback to iteratively correct the prediction model and optimization model. This results in site selection results having continuous correction capability, adaptive capability and higher long-term robustness.

[0141] This implementation method, through the coordinated execution of S1 to S7, elevates the rural charging facility site selection process from traditional static, experience-based planning to intelligent, closed-loop planning oriented towards complex environments. Specifically, it enhances multi-source data organization capabilities through unified graph modeling in S1, improves the ability to characterize dynamic charging demand through demand intensity modeling in S2, enhances the realism of service paths through reachability cost modeling in S3, improves engineering feasibility through deployment feasibility modeling in S4, enhances forward-looking decision-making capabilities through spatiotemporal prediction in S5, improves the overall site selection quality through robust optimization in S6, and enhances long-term adaptability and robustness through simulation feedback in S7. This allows for a more practical, safer, more economical, and more stable charging facility layout in complex rural environments.

[0142] Optionally, the present invention also provides a charging pile site selection system, which may consist of a processor, a memory, and program instructions stored in the memory and executable on the processor, or may consist of hardware modules and / or software modules with corresponding functions. Specifically, the data construction module is used to execute step S1, acquiring relevant data of the target rural area, performing grid-based discretization of the target rural area, and constructing a candidate node set, a node connection relationship set, and a rural spatiotemporal map; the demand modeling module is used to execute step S2, calculating the energy replenishment demand intensity of each candidate node based on the candidate node set and its corresponding historical trajectory, dwell information, energy replenishment behavior information, and task scenario information; the reachability cost modeling module is used to execute step S3, calculating the comprehensive reachability cost from the demand node to the candidate node; the deployment feasibility modeling module is used to execute step S4, calculating the deployment feasibility of each candidate node; and the prediction module... The first module executes step S5, feeding the combined energy demand intensity, overall reachability cost, and deployment feasibility as inputs into the spatiotemporal prediction model to obtain the predicted demand revenue and risk disturbance values ​​for candidate nodes. The second module executes step S6, constructing and solving the robust site selection optimization model to obtain the charging facility site selection scheme. The third module executes step S7, performing load disturbance simulation and extreme environment simulation on the charging facility site selection scheme, and updating the spatiotemporal prediction model and the robust site selection optimization model when the simulation results do not meet a preset threshold. The fourth module outputs the final charging facility site selection result. These modules are coupled according to the data flow sequence from steps S1 to S7 to achieve intelligent site selection for charging facilities in complex rural environments.

[0143] The above description is merely a preferred embodiment of the present invention, used to illustrate the technical solution of the present invention, and not to limit the scope of protection of the present invention. For those skilled in the art, various equivalent substitutions, modifications, improvements, or combinations can be made based on the content disclosed in the present invention without departing from the concept and technical essence of the present invention; any technical solution formed by adopting the technical features recorded in the specification and claims of the present invention, or by making equivalent substitutions thereof, should fall within the scope of protection of the present invention.

Claims

1. A method for selecting the location of a charging pile, characterized in that, include: S1. Obtain relevant data for the target rural area, perform grid-based discretization on the target rural area, and construct a candidate node set. Set of node connection relationships To form a rural time and space map ; S2, Based on the candidate node set And the corresponding historical trajectory, dwell information, refueling behavior information and task scenario information, to calculate each candidate node At any moment Energy replenishment demand intensity ; S3. Based on the aforementioned rural spatiotemporal map The required nodes are calculated based on road connectivity, terrain raster attributes, and meteorological environment attributes. to candidate node The overall attainable cost ; S4. Based on the energy attributes, construction and maintenance convenience, cost conditions, and risk conditions of the candidate nodes, calculate the parameters for each candidate node. Deployment feasibility ; S5, Intensity of Energy Replenishment Demand Comprehensive achievable cost and deployment feasibility As a joint input, it is fed into the spatiotemporal prediction model to obtain candidate nodes. Demand and revenue forecasts With risk disturbance forecast ; S6. Construct a robust addressing optimization model based on the following formula: in, To comprehensively optimize the objective function, Candidate nodes Location decision variables; Candidate nodes Demand and revenue forecasts; To serve fairness indicators; Candidate nodes Construction costs; Candidate nodes The predicted value of risk disturbances; , , To optimize the trade-off coefficients; And satisfy: in, This is the upper limit of the total budget; in, For demand nodes The maximum acceptable service cost; in, The maximum permissible risk threshold, This refers to the risk tolerance coefficient. S7. Perform load disturbance simulation and extreme environment simulation on the obtained charging facility location scheme. When the simulation results do not meet the preset threshold, update the spatiotemporal prediction model and the robust location optimization model according to the simulation feedback results, and re-execute step S6 until the final charging facility location result is output.

2. The charging pile site selection method according to claim 1, characterized in that, In step S2, a demand intensity sequence is constructed based on the energy replenishment demand intensity of each candidate node within a preset time window. in, Candidate nodes The energy replenishment demand intensity sequence, This represents the number of time steps.

3. The charging pile site selection method according to claim 1, characterized in that, In step S3, a comprehensive reachability cost matrix is ​​constructed between the demand node and the candidate node. in, To synthesize the reachability cost matrix, The total number of nodes required. This represents the total number of candidate nodes.

4. The charging pile site selection method according to claim 1, characterized in that, In step S4, the risk penalty item Represented as: According to the charging pile site selection method of claim 1, the risk penalty item in step S4 is represented as: in, Risk penalty item for candidate nodes, The geological hazard risk value of the candidate node. The extreme weather risk value for candidate nodes. This represents the risk value of communication failure for candidate nodes. The energy outage risk value for candidate nodes. , , , This is the risk weighting coefficient.

5. The charging pile site selection method according to claim 1, characterized in that, In step S4, the pre-screening score of the candidate nodes is calculated according to the following formula: in, Candidate nodes Pre-screening scores To normalize the demand intensity, To normalize deployment feasibility, To normalize the risk penalty item, These are the pre-screening weighting coefficients; and the top performers are selected from highest to lowest according to the pre-screening scores. These candidate nodes will be used as targets for subsequent optimization.

6. The charging pile site selection method according to claim 1, characterized in that, In step S5, the spatiotemporal prediction model adopts a fusion model of graph neural network and temporal network, and uses the feature vectors of candidate nodes. As input, where Candidate nodes At any moment The input feature vector, Candidate nodes The average comprehensive achievable cost for each demand node.

7. The charging pile site selection method according to claim 6, characterized in that, The spatiotemporal prediction model is trained using a joint loss function: in, For the joint loss function, Forecast losses for demand revenue. To predict losses due to risk disturbances, This is the loss weighting coefficient.

8. The charging pile site selection method according to claim 1, characterized in that, In step S6, the service fairness index Represented as: in, To serve fairness indicators, The total number of nodes required. For demand nodes The cost of service to the allocated charging facilities, The average cost of serving all demand nodes. To prevent tiny positive numbers with a denominator of zero.

9. The charging pile site selection method according to claim 1, characterized in that, In step S7, the predicted values ​​of demand and revenue and the predicted values ​​of risk disturbance are adjusted in the time domain based on the simulation feedback results: , , in, This is the updated demand and revenue forecast. This is the updated predicted value of risk disturbance. and These are the original demand and revenue forecasts and the original risk disturbance forecasts, respectively. and These represent the actual demand and benefit values ​​and the actual risk values ​​obtained from simulation feedback, respectively. To update the step size.

10. A charging pile site selection system, characterized in that, include: The data construction module is used to acquire terrain raster data, road trajectory data, historical stay and refueling behavior data, meteorological and environmental data, energy supply data, and safety risk data of the target rural area, and to perform grid-based discretization of the target rural area to construct a candidate node set. Node connection set and rural time and space map ; The requirement modeling module is used to model based on the candidate node set. And the corresponding historical trajectory, dwell information, refueling behavior information and task scenario information, to calculate each candidate node Energy replenishment demand intensity ; The reachability cost modeling module is used to calculate the comprehensive reachability cost from demand nodes to candidate nodes based on road connectivity, terrain grid attributes, and meteorological environment attributes. ; Deploy a feasibility modeling module to calculate the feasibility of each candidate node based on its energy attributes, ease of construction and maintenance, cost conditions, and risk conditions. Deployment feasibility ; The prediction module takes the energy replenishment demand intensity, overall reachability cost, and deployment feasibility as joint inputs and feeds them into the spatiotemporal prediction model to obtain the demand and revenue prediction values ​​for candidate nodes. With risk disturbance forecast ; A robust optimization module is used to optimize based on the predicted demand and revenue values. Risk disturbance prediction value Comprehensive achievable cost and construction costs A robust site selection optimization model with budget constraints, service constraints, and risk opportunity constraints is constructed and solved to obtain a site selection scheme for charging facilities. The feedback correction module is used to perform load disturbance simulation and extreme environment simulation on the charging facility site selection scheme, and update the spatiotemporal prediction model and the robust site selection optimization model when the simulation results do not meet the preset threshold. The output module is used to output the final location results of the charging facilities.

Images