Intelligent site selection base station management system and method

Abstract

The invention relates to the technical field of intelligent site selection of new energy automobile charging facilities, and discloses an intelligent site selection base station management system and method, and the system comprises a city multi-dimensional data collection and preprocessing module, a knowledge base construction module, a city evolution dynamics simulation module, a future scene generation module, and a charging network optimization model construction module. A robust optimization solving module; the method comprises the following steps: collecting and constructing a knowledge base of multi-dimensional city data; simulating city dynamics by using multi-agent modeling and a space-time diagram neural network to generate a plurality of future scene sets; and finally, constructing a multi-target robust optimization model based on the scenes, and solving to generate a group of Pareto optimal robust charging network planning schemes. According to the method, the overall scientificity, foresight and robustness of charging base station site selection planning are improved, so that the investment decision risk of the charging base station site selection planning is effectively reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent site selection for new energy vehicle charging facilities, and in particular to a management system and method for a smart site selection base station. Background Art

[0002] With the global energy transition and the advancement of the "dual carbon" goals, the new energy vehicle industry is rapidly developing. The planning and construction of charging networks, as a key infrastructure, are crucial. The strategic location of charging base stations directly impacts user experience and return on investment. Therefore, the precise, forward-looking, and intelligent site selection of charging base stations has become a key issue in the industry.

[0003] Existing technology often relies on static data from geographic information systems (GIS) and expert experience to determine charging base station sites. This method, which overlays layers of points of interest (POIs) and road networks for visual analysis to screen candidate locations, is a widely used technique.

[0004] However, this type of technology has significant limitations. It relies on static data, making it difficult to integrate dynamic socioeconomic information such as population and transportation, and lacks a deep understanding of demand drivers. Furthermore, this current-state analysis approach lacks the ability to predict future urban dynamics, resulting in insufficient foresight in planning solutions. More importantly, traditional approaches tend to seek a single optimal solution, ignoring future uncertainties such as market conditions and policies. This makes planning solutions less adaptable in the long term and carries a high investment risk.

[0005] Therefore, the present invention proposes a smart site selection base station management system and method to address the deficiencies of the prior art. Summary of the Invention

[0006] The purpose of the present invention is to provide an intelligent site selection base station management system and method, which solves the problem that traditional site selection methods rely on static data and lack dynamic prediction capabilities, resulting in insufficient foresight and robustness of planning schemes and high investment risks.

[0007] To achieve the above objectives, the present invention is implemented through the following technical solutions: a smart site selection base station management system, comprising: Urban multi-dimensional data collection and preprocessing module, used to collect and preprocess multi-dimensional urban data; A knowledge base construction module, which constructs a structured industry data knowledge base based on the multi-dimensional city data; An urban evolution dynamics simulation module, which is used to simulate the evolution of urban dynamics and future charging demand based on data from the industry data knowledge base through multi-agent modeling and spatiotemporal graph neural network-assisted calibration, and generate corresponding simulation results; A future scenario generation module, configured to generate a scenario set comprising a variety of future urban evolution scenarios by setting different parameters and disturbances based on the simulation results; a charging network optimization model construction module, configured to analyze scenario characteristics based on the scenario set and define optimization objective functions and constraints for charging base station network planning, so as to construct a charging base station network optimization model; The robust optimization solution module is used to solve the charging base station network optimization model using a multi-objective robust optimization algorithm and combine it with a robustness metric to generate a Pareto optimal robust charging base station network planning solution set.

[0008] Preferably, the city multi-dimensional data collection and preprocessing module includes: Collecting multi-dimensional urban data including at least geospatial data, demographic and socioeconomic data, traffic dynamics data, energy and facilities data, and urban planning and policy data; Perform data cleaning, data fusion and data transformation processing on the collected multi-dimensional urban data.

[0009] Preferably, the knowledge base construction module includes: Receiving the multi-dimensional city data pre-processed by the city multi-dimensional data acquisition and pre-processing module; Performing structural processing and spatiotemporal alignment on the multi-dimensional urban data; The processed structured data is integrated and stored to form the structured industry data knowledge base.

[0010] Preferably, the urban evolution dynamics simulation module includes: constructing a virtual city environment containing geospatial information based on data obtained from the industry data knowledge base; defining the types, attributes, behavior rules, and decision logic of agents in the virtual city environment to perform multi-agent modeling; The multi-agent modeling is used to simulate the evolution of urban dynamics and future charging demand, and generate corresponding simulation results.

[0011] Preferably, when the urban evolution dynamics simulation module simulates the multi-agent modeling, a spatiotemporal graph neural network is used to assist in the calibration of the multi-agent modeling. The spatiotemporal graph neural network updates the node feature representation by the following formula: : ; Where, For the The node feature representation matrix of the layer; is the adjacency matrix with self-loop added; for The diagonal matrix of For the Layer trainable weight matrix; is the activation function.

[0012] Preferably, the future scenario generation module includes: Receiving the simulation result generated by the urban evolution dynamics simulation module; By setting multiple groups of parameter combinations and disturbance factors based on the simulation results; Multiple rounds of urban evolution simulation are performed to generate the scenario set comprising multiple future urban evolution scenarios.

[0013] Preferably, when setting the multiple sets of parameter combinations and disturbance factors, the future scenario generation module sets the parameters in the first In the future urban evolution scenario The setting value of each parameter One of the following: or ; Where, Representative In the future urban evolution scenario The setting value of each parameter; Representative The baseline value of each parameter; Representative for the The parameter in Multiplicative disturbance factors under each scenario; Representative for the The parameter in The additive disturbance factor under each scenario; The disturbance factor is sampled from a preset probability distribution or determined according to a predefined set of rules.

[0014] Preferably, the charging network optimization model building module includes: Based on the scenario set, analyzing the charging demand distribution, traffic flow, and temporal and spatial evolution trends under each scenario to extract scenario characteristics; defining an optimization objective function with the goal of minimizing at least one of the total network cost and maximizing the service coverage; Define constraints including capacity limits of candidate sites, total budget limits, and service level requirements; The charging base station network optimization model is constructed by combining the optimization objective function and the constraint conditions.

[0015] Preferably, the robust optimization solution module includes: solving the charging base station network optimization model using a multi-objective robust optimization algorithm selected from the group consisting of an improved non-dominated sorting genetic algorithm and a multi-objective particle swarm optimization algorithm; During the solution process, a robustness metric is used to evaluate the performance stability of each planning scheme under different scenarios. The robustness metric includes a regret minimization metric or a scenario worst performance optimization metric. Solve to obtain a set of solutions that trade off between the objective function and the robustness measure; The Pareto optimal robust charging base station network planning solution set is identified and output from the solution.

[0016] The present invention also provides a method for managing a smart site selection base station, comprising the following steps: Collect and preprocess multi-dimensional urban data; Building a structured industry data knowledge base based on the multi-dimensional city data; Based on the data in the industry data knowledge base, through multi-agent modeling and spatiotemporal graph neural network-assisted calibration, the evolution of urban dynamics and future charging demand is simulated and corresponding simulation results are generated; Based on the simulation results, a scenario set containing various future urban evolution scenarios is generated by setting different parameters and disturbances; Based on the scenario set, the scenario characteristics are analyzed, and the optimization objective function and constraint conditions of the charging base station network planning are defined to construct a charging base station network optimization model; Based on the charging base station network optimization model, a multi-objective robust optimization algorithm is adopted to solve it, and combined with a robustness metric to generate a Pareto optimal robust charging base station network planning solution set.

[0017] In summary, the present invention includes at least one of the following beneficial technical effects: 1. This invention overcomes the information incompleteness inherent in traditional site selection methods, which rely solely on static geographic information and expert experience, by constructing a structured industry knowledge base encompassing multi-dimensional urban data such as population, transportation, and commerce. This solution integrates discrete, heterogeneous data into a machine-readable knowledge network, providing a comprehensive and timely digital foundation for subsequent dynamic simulation and intelligent decision-making. This significantly enhances the data breadth and depth of smart site selection analysis, making the assessment of a site's commercial potential more accurate and reliable.

[0018] 2. This invention addresses the difficulty of traditional site selection tools in effectively capturing dynamic urban changes and predicting future demand by introducing an urban evolution dynamics simulation module, combined with multi-agent modeling and spatiotemporal graph neural network calibration techniques. This approach no longer relies on static data analysis, but instead simulates the micro-behavior and macro-emergence of various entities in the city to deduce the dynamic evolution of future traffic flow and charging demand. This enables forward-looking assessment of site selection decisions, significantly improving the timeliness and accuracy of site selection proposals.

[0019] 3. By constructing and solving a multi-objective robust optimization model, this invention addresses the lack of stability of traditional single-planning schemes in the face of future uncertainty, effectively reducing investment risk and decision-making costs. This approach does not generate a single "optimal solution," but rather systematically considers multiple future evolution scenarios to generate a set of Pareto-optimal robust planning schemes that strike a balance between cost, coverage, and robustness. This enables decision makers to select the most appropriate solution based on their strategic preferences, significantly improving the scientific nature and decision-making flexibility of charging network planning. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 This is a system architecture diagram of the present invention; Figure 2 Flow chart of the method of the present invention. DETAILED DESCRIPTION

[0021] The following is combined with Figure 1 -Attached Figure 2 , the present invention is described in further detail.

[0022] An embodiment of the present invention provides a smart site selection base station management system, including: Urban multi-dimensional data collection and preprocessing module, used to collect and preprocess multi-dimensional urban data; In this embodiment, the operational quality of the urban multidimensional data acquisition and preprocessing module directly determines the authenticity of subsequent urban evolution simulations, the diversity of scenario generation, and the reliability and robustness of the final site selection plan. Its internal workflow can be primarily divided into two closely connected phases: data acquisition and data preprocessing.

[0023] During the data collection phase, the module is configured to systematically acquire the multidimensional urban data required to build a city digital twin from multiple heterogeneous sources. The term "multidimensional" here refers not only to the diversity of data types but also to the temporal and spatial attributes. Preferably, the collected multidimensional urban data types include at least: Geospatial data: This is the framework for building a virtual urban environment. It encompasses not only the macroscopic urban road network topology and administrative boundaries, but also the microscopic land use types (such as commercial, residential, and industrial land) and the precise geographic locations of points of interest (POIs). The purpose of collecting this data is to provide a physical, spatial constraint for subsequent agent activities and to provide a foundation for calculating key metrics such as accessibility and service radius.

[0024] Demographic and socioeconomic data: This is the core driver of the charging demand simulation. The module collects population density, age structure, income level, and vehicle ownership for each Transportation Analysis Zone (TAZ) or finer grid, specifically the current penetration rate and growth forecast for electric vehicles (EVs). This data is crucial for initializing agent attributes, as residents from different socioeconomic backgrounds have significant differences in travel patterns, charging habits, and willingness to pay.

[0025] Traffic dynamics data: This is key to replicating the pulse of a city's operations. By accessing the city's traffic information center or parsing floating vehicle data (FCD), the module obtains historical and real-time traffic flow, speed fluctuations, and congestion indices for key road sections. This data not only calibrates traffic flows within the simulation module but also identifies potential congestion bottlenecks, preventing the placement of time-sensitive facilities like fast-charging stations there, thereby improving the user experience.

[0026] Energy and facility data: This data serves as a physical constraint to ensure the feasibility of planning solutions. This module collects the location, rated capacity, and real-time load of each substation in the existing power grid. This is essential because charging stations, especially DC fast-charging stations, impose significant loads on the power grid, and their locations must ensure sufficient grid margin to avoid impacts. Furthermore, collecting information on the location and utilization of existing charging stations facilitates competitive landscape analysis and identifies service gaps.

[0027] Urban planning and policy data: This is key information that empowers the model's forward-looking capabilities. The module collects and structuredly analyzes text or documents such as future city master plans, regional development strategies, and new energy subsidy policies. Unlike other types of data that rely on historical data, this type of data reveals future development trends and is essential input for generating robust and forward-looking planning solutions. It also guides the parameter setting of subsequent future scenario generation modules.

[0028] During the data preprocessing stage, the module performs a series of deep processing operations on the raw, heterogeneous, multi-dimensional urban data collected above, aiming to "purify" it into regular data assets that can be directly used by the model.

[0029] First, perform data cleaning. Professional statistical methods are used to address outliers, missing values, and noise in multi-dimensional urban data caused by sensor failures, network transmission errors, or human data entry errors. For example, time series-based interpolation algorithms (such as linear interpolation and spline interpolation) can be used to fill short-term gaps in traffic flow data; or outlier detection algorithms such as isolation forests can be used to identify and remove obviously unreasonable data points.

[0030] Secondly, perform data fusion operations. Since multi-dimensional urban data comes from a wide range of sources, their spatial references, temporal granularity, and data formats vary. The core task of this step is to achieve spatiotemporal alignment and format unification. Spatially, through the spatial join technology of the geographic information system (GIS), data from different layers (such as point-like POIs, linear road networks, and surface population areas) are mapped to a unified geographic grid or traffic analysis area. Temporally, through resampling or aggregation, data of different frequencies (such as vehicle trajectories at the second level and daily population statistics) are unified to a time granularity suitable for analysis, preferably at the hourly level.

[0031] Finally, data transformation is performed. To meet the input data requirements of subsequent machine learning models (especially spatiotemporal graph neural networks), multi-dimensional urban data needs to be normalized. For example, for numerical features such as population and traffic flow, min-max scaling or Z-score normalization is used to scale them to a fixed interval (such as [0, 1] or a mean of 0 and a variance of 1) to eliminate dimensional differences and prevent gradient vanishing or exploding during model training. For categorical features such as land use type, one-hot encoding is used to convert them into numerical vectors.

[0032] Through the above collection and preprocessing process, the module finally outputs a set of clean, complete, dimensionally rich and uniformly formatted structured multi-dimensional urban data.

[0033] A knowledge base construction module, which constructs a structured industry data knowledge base based on the multi-dimensional city data; In this embodiment, the main task of the knowledge base construction module is to receive the multi-dimensional city data processed by the previous module and organize it into a structured industry data knowledge base that contains internal connections.

[0034] This module is first configured to receive multidimensional urban data preprocessed by the Urban Multidimensional Data Acquisition and Preprocessing Module. This means that the module's input data already has a high degree of consistency and regularity, allowing the module to focus on building logical relationships and organizing knowledge between data.

[0035] After receiving the data, the key functions of the module—structuring and spatial-temporal alignment of multi-dimensional urban data—begin to execute.

[0036] In terms of structured processing, given that urban systems contain multiple elements and the complex interactions between them, simple table structures have limitations in expressing this inherent correlation. Therefore, in a preferred embodiment, a heterogeneous information network (i.e., a graph structure) is used to organize urban data. Specifically, the module maps different types of urban entities (e.g., a point of interest (POI), a plot of land, a road network intersection) as "nodes" in the network, and abstracts the relationships between entities (e.g., the physical connection of the road network, the spatial adjacency between plots of land, and the commuting relationship from residential areas to work areas) as "edges" in the network. In this way, the originally relatively discrete multi-dimensional urban data is integrated into a knowledge graph that reflects the internal logic.

[0037] In terms of spatiotemporal alignment, this step is further processed based on the preliminary alignment of the previous module. Its purpose is to ensure that in the above-mentioned graph structure, the attributes of all nodes and edges are attached with clear and unified spatiotemporal labels. For example, the attributes of the "edge" of the road network, in addition to static information such as length and level, can also include a time series in hours to record its traffic flow at different times. Similarly, the attributes of the "node" of the population area can include population forecast data for different years. This is intended to ensure that the knowledge base can provide a logically self-consistent and consistent snapshot of the city at any time slice, providing a data consistency basis for subsequent dynamic simulations.

[0038] Finally, the module integrates and stores the processed structured data to form a structured industry data knowledge base. This step instantiates and persistently stores the previously defined graph structure and its associated spatiotemporal attribute data. When selecting storage technology, graph databases (such as Neo4j) are preferred for storing the city's topological network and semantic relationships to support efficient graph-related queries. Furthermore, for geographic information, spatial databases with spatiotemporal indexing capabilities (such as PostGIS) can be used to handle operations such as spatial range queries.

[0039] Ultimately, the module outputs a structured, queryable, and analyzable industry data knowledge base.

[0040] An urban evolution dynamics simulation module, which is used to simulate the evolution of urban dynamics and future charging demand based on data from the industry data knowledge base through multi-agent modeling and spatiotemporal graph neural network-assisted calibration, and generate corresponding simulation results; In this embodiment, the main task of the urban evolution dynamics simulation module is to use the constructed industry data knowledge base to deduce the long-term dynamic evolution process of the future city through a hybrid method that integrates micro-behavior simulation and macro-data calibration, and generate the corresponding future charging demand spatiotemporal distribution results.

[0041] The module first constructs a virtual city environment containing geospatial information based on data obtained from the industry data repository. This virtual city environment is not generated out of thin air, but rather a three-dimensional mapping of the structured data in the repository. Specifically, the module reads geospatial information from the repository, including road network topology, land parcel functional divisions, and point of interest (POI) distribution, to construct a digital city "sandbox" in computer memory. This sandbox forms the underlying spatial foundation for all simulation activities.

[0042] Then, within the virtual city environment, the module defines the types, attributes, behavioral rules, and decision-making logic of the agents to perform multi-agent modeling. Here, agents are computational abstractions of various decision-making entities in the city.

[0043] In one specific implementation, the types of agents may include residents, commuter vehicles, commercial entities, etc.

[0044] Its attributes are initialized from the corresponding demographic and socioeconomic data in the knowledge base. For example, the resident intelligent body is given attributes such as age, income, place of residence, place of work, and whether it owns an electric car.

[0045] Its behavioral rules define the daily activity patterns of the agent, such as commuting, shopping, leisure and other travel chains set based on historical travel survey (OD) data.

[0046] Its decision-making logic is even more critical. It stipulates how the intelligent agent behaves when faced with choices. For example, when the battery level of an electric car owner is below a certain threshold, the intelligent agent will decide which charging station to go to based on factors such as the distance to the destination, the expected charging cost, and the waiting time at the charging station.

[0047] After completing these definitions, the module simulates urban dynamics and the evolution of future charging demand through multi-agent modeling. Tens of thousands of agents interact in parallel within a virtual environment. Their micro-level behaviors (such as individual travel and charging choices) are aggregated to reveal the dynamics of the entire city at a macro level, such as the spatiotemporal distribution of traffic flow, the cyclical prosperity of commercial districts, and most crucially, the future charging demand of different areas at different times.

[0048] However, due to the numerous parameters of simple multi-agent models, their simulation results may deviate from the real world. To improve the fidelity of the simulation, a key technical feature of this module is the use of a spatiotemporal graph neural network (ST-GNN) to assist in the calibration of the multi-agent model. This is essential because ST-GNN can learn the complex, nonlinear spatiotemporal dependencies of urban systems from the vast amount of historical spatiotemporal data (such as traffic flow and charging records) stored in the knowledge base. These learned macroscopic laws are used to constrain and calibrate the microscopic behavioral parameters in the multi-agent model, ensuring that the simulated macroscopic phenomena match the statistical characteristics of the real data.

[0049] In a preferred embodiment, the spatiotemporal graph neural network updates the node feature representation by the following formula: : ; This formula describes a core operation in a graph convolutional network, which aggregates feature information of neighboring nodes in the graph.

[0050] Where, For the The node feature representation matrix of the layer, where each row can represent the feature vector (such as population, traffic flow, etc.) of an area at a specific time; The adjacency matrix with self-loops added is used to represent the spatial connectivity between regions; for The diagonal matrix is ​​used for normalization to keep the data scale stable; For the The layer’s trainable weight matrix encodes the patterns of how spatiotemporal features are transformed and transferred; The activation function (such as ReLU) enables the model to learn and express complex nonlinear relationships.

[0051] Through this mechanism, the simulation results generated by this module are not only derived from the deduction of microscopic behavior, but are also calibrated by the laws of macroscopic data.

[0052] A future scenario generation module, configured to generate a scenario set comprising a variety of future urban evolution scenarios by setting different parameters and disturbances based on the simulation results; In this embodiment, the future scenario generation module is a core component for addressing the profound uncertainty of the future. Its primary purpose is not to make a single precise prediction, but rather to systematically explore multiple possible future development paths, thereby constructing a set of scenarios for robustness assessment and optimization.

[0053] The module begins by receiving simulation results generated by the Urban Evolution Dynamics Simulation module. This simulation typically represents a "baseline" or "business-as-usual" future development path, representing the most likely scenario based on current knowledge. However, long-term planning that relies solely on this single scenario carries significant risks.

[0054] Therefore, the core function of the module is to explore the vast space of possibilities beyond the baseline scenario by setting multiple parameter combinations and perturbation factors based on simulation results. This process first requires identifying key uncertainties that significantly influence the long-term evolution of cities and charging demand. These parameters can preferably include macroeconomic growth rates, the rate of technological advancement and cost reduction curves of electric vehicles, long-term fluctuations in fuel prices, the continuity and intensity of relevant subsidy policies, and urban spatial expansion patterns.

[0055] When setting parameters, this module adopts a structured perturbation method. In a preferred embodiment, at least for some parameters, In the future urban evolution scenario The setting value of each parameter One of the following: or ; The formulas here provide a standardized way to generate new parameter values ​​from baseline values.

[0056] Where, Representative In the future urban evolution scenario The setting value of each parameter; Representative The baseline value of each parameter; Representative for the The parameter in Multiplicative disturbance factors under each scenario; Representative for the The parameter in The additive disturbance factor under each scenario.

[0057] The first form (multiplicative perturbation) is achieved by the perturbation factor Achievement is applicable to parameters that describe relative rates of change, such as "economic growth rate is 20% higher than the benchmark."

[0058] The second form (additive perturbation) is achieved by the perturbation factor Implementation, suitable for describing changes in absolute values.

[0059] Perturbation Factor and The source of is the key to achieving scenario diversity in this module. In a specific implementation, the perturbation factor is sampled from a preset probability distribution or determined according to a predefined set of rules.

[0060] When sampling from a probability distribution, for example, one can assume that the perturbation factor for economic growth follows a normal distribution with mean 0. This means that in most scenarios, economic growth will fluctuate slightly around the baseline value, but there is also a small probability of significant deviation. This approach allows for the systematic generation of statistically significant scenarios on a large scale.

[0061] When determined by a rule set, scenarios with clear narrative logic can be constructed based on expert knowledge or policy analysis. For example, a "radical transition" scenario could be defined, assuming a combination of parameters such as strong electric vehicle subsidies, high fuel prices, and rapid technological progress. This approach ensures that certain key, structural future possibilities are fully assessed.

[0062] After setting a complete set of perturbations for all key parameters, the module executes multiple rounds of urban evolution simulations. Each parameter combination is fed into the previously described "Urban Evolution Dynamics Simulation Module" and a complete simulation is run from the present to the future. The output of each round constitutes a separate future urban evolution scenario, detailing the spatiotemporal distribution of future charging demand under that set of parameters.

[0063] By repeating this process, the module eventually generates a scenario set containing various future urban evolution scenarios.

[0064] a charging network optimization model construction module, configured to analyze scenario characteristics based on the scenario set and define optimization objective functions and constraints for charging base station network planning, so as to construct a charging base station network optimization model; In this embodiment, the core function of the charging network optimization model construction module is to transform the complex, multi-dimensional charging base station network planning problem into a structured, solvable mathematical optimization model. This module bridges the gap between scenario generation and optimization solution, formalizing the various future scenarios generated by the previous modules into the optimization problem's objectives and constraints.

[0065] This module begins by analyzing the distribution of charging demand, traffic flow, and temporal and spatial evolution trends within each scenario based on a set of scenarios to extract scenario characteristics. It receives the set of scenarios output by the "Future Scenario Generation Module" and analyzes each of them. This analysis aims to quantify the performance potential of each candidate site under different future scenarios. For example, the module calculates key indicators such as the potential charging demand within the service area of ​​each candidate site under each scenario, as well as peak-hour traffic flow, providing specific, quantitative data input for the subsequent definition of the objective function and constraints.

[0066] Next, the module enters the core mathematical modeling phase, which begins by defining an optimization objective function that aims to minimize at least one of the following: total network cost and maximize service coverage. In a preferred embodiment, this is a multi-objective optimization problem that balances the economic efficiency of the planning solution with the service effectiveness.

[0067] Goal 1: Minimize total network cost. This goal focuses on the economic feasibility of the solution and aims to minimize the sum of the costs associated with all candidate sites selected for construction. Its mathematical form can be expressed as: ; Where, Represents the set of all candidate sites; is a binary decision variable. When its value is 1, it means that Build a charging station, if it is 0, it will not be built; Representative in position The total costs associated with establishing the site, which may include construction, land, and discounted operation and maintenance costs; is the objective function value of the total network cost to be optimized.

[0068] Goal 2: Maximize service coverage. This goal focuses on the service quality and robustness of the solution, aiming to improve the performance in the worst case scenario among all possible scenarios. Its mathematical form can be expressed as: ; Where, represents the set of all future scenarios; Represents the set of all charging demand points; Representative demand points In the scenario Charging demand under It is an auxiliary binary variable, and when its value is 1, it indicates a demand point In the scenario The network can be covered; is the objective function value of the robust service coverage to be optimized. This formula reflects the consideration of the performance stability of the planning scheme by maximizing the minimum value of the service coverage in all scenarios.

[0069] The module then defines constraints, including capacity limits for candidate sites, budget limits, and service level requirements. These constraints ensure that the resulting solution is feasible in the real world.

[0070] Total budget constraint: This constraint ensures that the total cost of the plan does not exceed the preset total investment budget . ; Capacity constraints of candidate sites: This constraint ensures that the total demand served by each planned charging station does not exceed its own design capacity in any future scenario. .

[0071] ; Where, For the scenario Next assigned to site The total charging demand is . This constraint must hold for every candidate site and every future scenario.

[0072] Service level requirement: This is the bottom line constraint of service quality, requiring the network to meet a minimum service coverage level percentage in any future scenario .

[0073] ; This constraint also needs to be satisfied for every future scenario.

[0074] Finally, the module comprehensively optimizes the objective function and constraints to construct a charging base station network optimization model. This model mathematically describes the complex decision-making problem of finding a series of charging network layout solutions that achieve different trade-offs between cost and service coverage, while meeting realistic constraints such as budget, capacity, and service level, and ensuring that these trade-offs remain as robust as possible under multiple future scenarios.

[0075] The robust optimization solution module is used to solve the charging base station network optimization model using a multi-objective robust optimization algorithm and combine it with a robustness metric to generate a Pareto optimal robust charging base station network planning solution set.

[0076] The core task of the robust optimization solution module in this embodiment is to efficiently solve the charging base station network optimization model constructed by the previous module, which contains multiple objectives and future uncertainties, and ultimately generate a series of planning schemes that strike a balance between economy, serviceability, and robustness.

[0077] This module first accepts the mathematical model defined by the "Charging Network Optimization Model Construction Module" that includes the complete objective function and constraints. It then solves the charging base station network optimization model using a multi-objective robust optimization algorithm selected from the group consisting of an improved non-dominated sorting genetic algorithm and a multi-objective particle swarm optimization algorithm. This type of heuristic algorithm is preferred because charging network site selection problems are typically large-scale, multi-constrained, and multi-objective, and the objective function may be non-convex, making it difficult for traditional exact algorithms to solve them within a reasonable time. Methods such as the improved non-dominated sorting genetic algorithm (NSGA-II) are particularly suitable for finding Pareto optimal solutions to multi-objective problems by simulating the selection, crossover, and mutation processes of biological evolution and enabling global searches in complex solution spaces.

[0078] During the solution process, a key technical feature of this module is the incorporation of robustness metrics to assess the performance stability of each planning solution under different scenarios. This ensures that the generated solution not only performs well under a desired scenario but is also resilient to the negative impacts of future uncertainty. Preferably, the robustness metric includes a regret minimization metric or a worst-case scenario performance optimization metric.

[0079] Worst-case performance optimization metric: This metric has been formally defined as the robust service coverage objective function in the previous "Charging Network Optimization Model Construction Module". When executing the algorithm, this solution module directly minimizes the total network cost and maximizes the As the two core optimization goals, the search process of the driving algorithm converges towards lower cost and higher bottom-line performance.

[0080] Regret Minimization Metric: This is an alternative robustness assessment method that can be used in conjunction with the previous one. It measures the performance gap between a solution's performance in a given scenario and the optimal solution if that scenario were known to occur and designed specifically for it. This gap is the "regret value." The goal of this metric is to find a solution that minimizes the "maximum regret value" across all possible scenarios. Its mathematical form can be defined as: ; Where, Represents planning options currently under evaluation; Represents a specific future scenario; It's a plan In the scenario Performance values ​​under the circumstance (such as service coverage); Represents a known situation Under the premise that this will happen, the theoretically optimal solution that can achieve the best performance in this scenario, This is the maximum regret value of the solution. When this metric is used, it can be incorporated into the algorithm as the third optimization objective.

[0081] In each iteration of the algorithm, for each candidate solution in the population (i.e. a specific charging network planning scheme), the module calculates its cost target value. , and one or more of the above robustness metrics, and perform non-dominated sorting and selection operations based on them.

[0082] Through multi-generational evolutionary calculations, the module ultimately solves a set of solutions that balance the objective function and robustness measure. These solutions do not dominate each other in the solution space.

[0083] Finally, the module identifies and outputs a Pareto-optimal robust charging base station network planning solution set from the solutions. This set of solutions constitutes the final decision support output. Each solution in the set is a Pareto-optimal solution, meaning it is difficult to improve at least one objective without sacrificing another (e.g., increasing cost or reducing robustness). This set of solutions provides decision makers with a range of high-quality alternatives, allowing them to select the most appropriate charging network planning solution based on their risk appetite and strategic priorities.

[0084] The present invention also provides a method for managing a smart site selection base station, comprising the following steps: S1. Collect and preprocess multi-dimensional urban data; S2. Building a structured industry data knowledge base based on the multi-dimensional city data; S3. Based on the data in the industry data knowledge base, simulate the evolution of urban dynamics and future charging demand through multi-agent modeling and spatiotemporal graph neural network-assisted calibration, and generate corresponding simulation results; S4. Based on the simulation results, generating a scenario set including multiple future urban evolution scenarios by setting different parameters and disturbances; S5. Based on the scenario set, analyze scenario characteristics and define optimization objective functions and constraints for charging base station network planning to construct a charging base station network optimization model; S6. Based on the charging base station network optimization model, a multi-objective robust optimization algorithm is used to solve the problem, and a robustness metric is combined to generate a Pareto optimal robust charging base station network planning solution set.

[0085] The method of this embodiment can be used to execute the above system embodiment. Its principles and technical effects are similar and will not be described in detail here.

[0086] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. Intelligent site selection base station management system, characterized by: include: Urban multi-dimensional data collection and preprocessing module, used to collect and preprocess multi-dimensional urban data; A knowledge base construction module, which constructs a structured industry data knowledge base based on the multi-dimensional city data; An urban evolution dynamics simulation module, which is used to simulate the evolution of urban dynamics and future charging demand based on data from the industry data knowledge base through multi-agent modeling and spatiotemporal graph neural network-assisted calibration, and generate corresponding simulation results; A future scenario generation module, configured to generate a scenario set comprising a variety of future urban evolution scenarios by setting different parameters and disturbances based on the simulation results; a charging network optimization model construction module, configured to analyze scenario characteristics based on the scenario set and define optimization objective functions and constraints for charging base station network planning, so as to construct a charging base station network optimization model; The robust optimization solution module is used to solve the charging base station network optimization model using a multi-objective robust optimization algorithm and combine it with a robustness metric to generate a Pareto optimal robust charging base station network planning solution set.

2. The intelligent site selection base station management system according to claim 1, characterized in that: The urban multi-dimensional data collection and preprocessing module includes: Collecting multi-dimensional urban data including at least geospatial data, demographic and socioeconomic data, traffic dynamics data, energy and facilities data, and urban planning and policy data; Perform data cleaning, data fusion and data transformation processing on the collected multi-dimensional urban data.

3. The intelligent site selection base station management system according to claim 1, characterized in that: The knowledge base building module includes: Receiving the multi-dimensional city data pre-processed by the city multi-dimensional data acquisition and pre-processing module; Performing structural processing and spatiotemporal alignment on the multi-dimensional urban data; The processed structured data is integrated and stored to form the structured industry data knowledge base.

4. The intelligent site selection base station management system according to claim 1, characterized in that: The urban evolution dynamics simulation module includes: constructing a virtual city environment containing geospatial information based on data obtained from the industry data knowledge base; defining the types, attributes, behavior rules, and decision logic of agents in the virtual city environment to perform multi-agent modeling; The multi-agent modeling is used to simulate the evolution of urban dynamics and future charging demand, and generate corresponding simulation results.

5. The intelligent site selection base station management system according to claim 4, characterized in that: When the urban evolution dynamics simulation module simulates the multi-agent model, a spatiotemporal graph neural network is used to assist in the calibration of the multi-agent model. The spatiotemporal graph neural network updates the node feature representation by the following formula: : ； Where, For the The node feature representation matrix of the layer; is the adjacency matrix with self-loop added; for The diagonal matrix of For the Layer trainable weight matrix; is the activation function.

6. The intelligent site selection base station management system according to claim 1, characterized in that: The future scenario generation module includes: Receiving the simulation result generated by the urban evolution dynamics simulation module; By setting multiple sets of parameter combinations and disturbance factors based on the simulation results; Multiple rounds of urban evolution simulation are performed to generate the scenario set comprising multiple future urban evolution scenarios.

7. The intelligent site selection base station management system according to claim 6, characterized in that: When the future scenario generation module sets the multiple sets of parameter combinations and disturbance factors, at least for some parameters, in the first In the future urban evolution scenario The setting value of each parameter One of the following: or ; Where, Representative In the future urban evolution scenario The setting value of each parameter; Representative The baseline value of each parameter; Representative for the The parameter in Multiplicative disturbance factors under each scenario; Representative for the The parameter in The additive disturbance factor under each scenario; The disturbance factor is sampled from a preset probability distribution or determined according to a predefined set of rules.

8. The intelligent site selection base station management system according to claim 1, characterized in that: The charging network optimization model building module includes: Based on the scenario set, analyzing the charging demand distribution, traffic flow, and temporal and spatial evolution trends under each scenario to extract scenario characteristics; defining an optimization objective function with the goal of minimizing at least one of the total network cost and maximizing the service coverage; Define constraints including capacity limits, budget limits, and service level requirements for candidate sites; The optimization objective function and the constraint conditions are combined to construct the charging base station network optimization model.

9. The intelligent site selection base station management system according to claim 1, characterized in that: The robust optimization solution module includes: solving the charging base station network optimization model using a multi-objective robust optimization algorithm selected from the group consisting of an improved non-dominated sorting genetic algorithm and a multi-objective particle swarm optimization algorithm; During the solution process, a robustness metric is used to evaluate the performance stability of each planning scheme under different scenarios. The robustness metric includes a regret minimization metric or a scenario worst performance optimization metric. Solve to obtain a set of solutions that trade off between the objective function and the robustness measure; The Pareto optimal robust charging base station network planning solution set is identified and output from the solution.

10. A method for managing a smart site selection base station, applied to the smart site selection base station management system according to any one of claims 1 to 9, characterized in that: The following steps are involved: Collect and preprocess multi-dimensional urban data; Building a structured industry data knowledge base based on the multi-dimensional city data; Based on the data in the industry data knowledge base, through multi-agent modeling and spatiotemporal graph neural network-assisted calibration, the evolution of urban dynamics and future charging demand is simulated and corresponding simulation results are generated; Based on the simulation results, a scenario set containing various future urban evolution scenarios is generated by setting different parameters and disturbances; Based on the scenario set, the scenario characteristics are analyzed, and the optimization objective function and constraint conditions of the charging base station network planning are defined to construct a charging base station network optimization model; Based on the charging base station network optimization model, a multi-objective robust optimization algorithm is adopted to solve it, and combined with a robustness metric to generate a Pareto optimal robust charging base station network planning solution set.

Images