A method and system for the layout of urban food facilities

Abstract

The application discloses a kind of layout method and system of urban food facilities, it is related to space planning technical field.The application includes: obtaining multi-period population heat map data, food facility data and road network data of urban area, and the urban area is divided into multiple spatial grids;Time series decomposition is carried out on multi-period population heat map, time-varying population characteristic component is obtained, and all spatial grids are divided into multiple types according to intraday amplitude ratio;Time-varying population characteristic component is differentially weighted, and time-varying demand value is obtained;The shortest path distance of each spatial grid to each food facility is determined, and the accessibility score of each spatial grid to each type of food facility in each period is determined;According to time-varying demand value and accessibility score, the supply-demand matching degree of each spatial grid in each period is determined;A multi-objective optimization model is constructed and iteratively solved, and the layout scheme of food facility is obtained.The application guarantees the applicability and accuracy of food facility layout scheme in dynamic urban environment.

Description

Technical Field

[0001] This invention relates to the field of spatial planning technology, and in particular to a method and system for the layout of urban food facilities. Background Technology

[0002] Urban food facilities include restaurants, farmers' markets, convenience stores, supermarkets, fresh food supermarkets, wholesale markets, and cold chain facilities. The layout of these facilities directly affects residents' access to food and the efficiency of urban operations. Optimizing the spatial distribution of these facilities is a key issue in urban planning, public health, and business decision-making.

[0003] In the prior art, Chinese invention patent application CN122222281A proposes a method and related products for optimizing the layout of public facilities based on crowd behavior analysis. The implementation process includes: using target detection and multi-target tracking algorithms to collect crowd behavior data and facility utilization rates in real time; gridding the monitoring area based on pedestrian dwell and walking data; statistically analyzing the cumulative dwell time and trajectory traversal times of each grid to generate dwell heat maps and walking path heat maps, and performing spatiotemporal dynamic analysis such as time period and seasonal differences; combining heat distribution and facility utilization rates, diagnosing problems such as inefficient facilities, blank areas, and uneven distribution through three-dimensional cross-analysis; constructing a reinforcement learning model (the state is the facility location and heat distribution, and the actions include adding, removing, and moving facilities) and a multi-objective optimization function to solve for the public facility adjustment strategy.

[0004] However, this existing technology homogenizes various facilities, ignoring the specific needs of food facilities in terms of service radius, usage time and user travel intentions, and fails to make full use of the time-varying information of multi-time period population heat maps to achieve dynamic matching of supply and demand, making it difficult to support scientific and accurate urban food facility layout decisions. Summary of the Invention

[0005] Therefore, it is necessary to provide a method and system for the layout of urban food facilities to address the aforementioned technical problems.

[0006] This invention provides a method for the layout of urban food facilities, comprising: Acquire multi-time-period population heat map data, food facility data, and road network data for urban areas, and divide the urban area into multiple spatial grids; The population heat map of each spatial grid is decomposed over time to obtain the time-varying population characteristic components of each spatial grid in different time periods. Based on the intraday amplitude ratio of the spatial grid, all spatial grids are divided into multiple types. For different types of spatial grids, the time-varying population characteristic components are differentially weighted to obtain the time-varying demand value of each spatial grid in each time period. Based on food facility data and road network data, the shortest path distance from each spatial grid to each food facility is determined, and the accessibility score of each spatial grid to each type of food facility is determined at each time period using the corresponding distance decay function according to the type of food facility. Based on the time-varying demand value and accessibility score of each spatial grid in each time period, the supply and demand matching degree of each spatial grid in each time period is determined. A multi-objective optimization model is constructed with the overall coverage rate of each time period not lower than a preset lower limit as a hard constraint, and the optimization objectives of maximizing the coverage rate during peak hours, minimizing the total road network travel distance, and achieving spatial balance as optimization goals. The model is then iteratively solved to obtain the layout scheme of food facilities within the urban area. Among them, the overall coverage rate is the proportion of spatial grids whose supply and demand matching degree meets the preset conditions within a single time period to all spatial grids, and the spatial balance is measured based on the degree of difference in supply and demand matching degree between each spatial grid.

[0007] Optionally, time-varying demographic components include: multi-day baseline components, diurnal fluctuation components, and weekend bias components; Among them, the multi-day baseline component characterizes the average population level of the spatial grid at different times of the day; The diurnal fluctuation component characterizes the degree of deviation of the spatial grid's real-time population from the multi-day baseline component on a specific date; The weekend bias component characterizes the systematic differences between the spatial grid on weekends and weekdays.

[0008] Optionally, the population heat values ​​of each spatial grid at multiple sampling times within a preset time period are obtained, and the population time change curve of each grid is constructed based on the population heat values; The daily population peak and daily population trough are extracted from the population time change curve, and the ratio of the daily population peak to the daily population trough is used as the daily amplitude ratio. The intraday amplitude ratios of all spatial grids are statistically sorted, and the classification thresholds are determined based on the ternary values ​​of the intraday amplitude ratios of all spatial grids. The classification thresholds include an upper threshold and a lower threshold. Spatial grids with intraday amplitude ratios exceeding the upper threshold are classified as commuter grids, indicating significant daytime population concentration and large intraday amplitude. Spatial grids with intraday amplitude ratios below the lower threshold are classified as residential, indicating significant nighttime population concentration and low intraday amplitude. The spatial grid with intraday amplitude ratios between the lower and upper thresholds is classified as steady-state, indicating that the population fluctuations throughout the day are relatively small.

[0009] Optionally, for different types of spatial grids, the time-varying population characteristic components are differentially weighted to obtain the time-varying demand values ​​for each spatial grid at each time period, specifically including: Obtain the basic time-varying population feature components of each spatial grid at each time period; Based on the type label of each spatial grid, the weighting coefficient of each spatial grid in each time period is determined from the preset weighting coefficient matrix. The row elements of the weighting coefficient matrix correspond to different grid types, and the column elements of the weighting coefficient matrix correspond to each time period. The weighting coefficient of the same type of grid is different in different time periods, and the weighting coefficient of different types of grids is different in the same time period. The weighting coefficient is used to characterize the degree of concentration or dispersion of food demand of the corresponding type of grid relative to the average level in the corresponding time period. The time-varying population characteristic components of each spatial grid in each time period are multiplied by the corresponding weighting coefficients to obtain the time-varying demand value of each spatial grid in each time period.

[0010] Optionally, the shortest path distance from each spatial grid to each food facility is determined based on food facility data and road network data, specifically including: A directed weighted graph is constructed based on the road network data of the urban area. The vertices of the directed weighted graph are the road intersections and the projection points of the centroid of the spatial grid onto the road network. The edges of the directed weighted graph are the road segments connecting adjacent vertices in the road network data. The weight of the edge is the road segment length or the estimated travel time. For each combination of spatial grid and each food facility, the shortest path distance along the directed weighted graph is determined by the shortest path algorithm; when the shortest path distance exceeds the preset search radius, the food facility is marked as unreachable.

[0011] Optionally, the accessibility score of each spatial grid to each type of food facility at each time period is determined by using the corresponding distance decay function based on the type of food facility. Specifically, this includes: The maximum acceptable distance and distance attenuation coefficient are determined according to the type of food facility, and the distance attenuation function is a piecewise Gaussian function. Based on the shortest path distance and distance decay function values ​​from each spatial grid to each food facility, the accessibility score of each spatial grid to each type of food facility at each time period is determined using a road network-constrained mobile search method, including: For each food facility, the ratio of the supply capacity of the food facility to the sum of the time-varying demand values ​​of all spatial grids within the reach of the food facility after distance attenuation weighting is determined to obtain the demand-weighted supply ratio of the food facility in each time period. For each spatial grid, the demand-weighted supply ratio of all food facilities within the reach of the spatial grid at each time period is accumulated after being weighted by distance attenuation to obtain the accessibility score of the spatial grid to each type of food facility at each time period.

[0012] Optionally, based on the time-varying demand value and accessibility score of each spatial grid in each time period, the supply-demand matching degree of each spatial grid in each time period is determined, specifically including: The accessibility scores of each spatial grid to each type of food facility are aggregated at each time period to obtain the total supply of each spatial grid at each time period. The total supply and the time-varying demand value of each spatial grid at each time period are used as the supply and demand matching degree of each spatial grid at each time period.

[0013] Optionally, it also includes: marking each spatial grid as a persistent gap, an intermittent gap, or covered based on the comparison results of the supply and demand matching degree of each spatial grid with a preset matching degree threshold at each time period; Specifically, for each spatial grid, if the supply and demand matching degree of the spatial grid is lower than the preset matching degree threshold in all time periods, the spatial grid is marked as a persistent gap; if the supply and demand matching degree of the spatial grid is lower than the preset matching degree threshold only in some time periods, the spatial grid is marked as an intermittent gap; if the supply and demand matching degree of the spatial grid is not lower than the preset matching degree threshold in all time periods, the spatial grid is marked as covered. The labeled results can be used as input constraints or optimization guidance information for multi-objective optimization models.

[0014] Optionally, a multi-objective optimization model is constructed with the overall coverage rate in each time period not lower than a preset lower limit as a hard constraint, and the optimization objectives being to maximize the coverage rate during peak hours, minimize the total road network travel distance, and achieve spatial balance. This model is then iteratively solved to obtain a layout scheme for food facilities within the urban area, specifically including: The location and type of candidate facilities are coded as decision variables. Candidate locations include: adjustment locations of existing facilities and candidate grids for new facilities selected from the persistent gap grids with the lowest supply-demand matching degree. A time-varying constrained multi-objective optimization model is constructed. The objective functions of the time-varying constrained multi-objective optimization model include: the overall coverage maximization objective, which is used to maximize the comprehensive level of coverage in each time period; the total road network travel distance minimization objective, which is used to minimize the sum of the time-varying demand value of each spatial grid and the shortest path distance to the nearest food facility in each time period; and the spatial balance objective, which is used to minimize the difference in the average supply and demand matching degree between each spatial grid, and the difference is measured in the form of Gini coefficient. The multiple objective functions of the time-varying constrained multi-objective optimization model are transformed into a single objective function by weighted summation. The single objective function is then iterated using a heuristic algorithm. The candidate solutions in the Pareto front obtained after the iteration are used as the layout schemes for food facilities in the urban area.

[0015] This invention provides a layout system for urban food facilities, comprising: The data acquisition module is used to acquire multi-time period population heat map data, food facility data, and road network data of urban areas, and divide the urban area into multiple spatial grids; The demand modeling module is used to perform time-series decomposition of the multi-period population heat map of each spatial grid, obtain the time-varying population characteristic components of each spatial grid in different time periods, and classify all spatial grids into multiple types according to the intraday amplitude ratio of the spatial grids; for different types of spatial grids, the time-varying population characteristic components are differentially weighted to obtain the time-varying demand value of each spatial grid in each time period. The accessibility determination module is used to determine the shortest path distance from each spatial grid to each food facility based on food facility data and road network data, and to determine the accessibility score of each spatial grid to each type of food facility at each time period using the corresponding distance decay function according to the type of food facility. The supply and demand matching module is used to determine the supply and demand matching degree of each spatial grid in each time period based on the time-varying demand value and accessibility score of each spatial grid in each time period. The layout optimization module is used to construct a multi-objective optimization model with the overall coverage rate of each time period not being lower than a preset lower limit as a hard constraint, and the optimization objectives being to maximize the coverage rate during peak periods, minimize the total road network travel distance, and achieve spatial balance. The model is then iteratively solved to obtain the layout scheme of food facilities within the urban area. Among them, the overall coverage rate is the proportion of spatial grids whose supply and demand matching degree meets the preset conditions within a single time period to all spatial grids, and the spatial balance is measured based on the degree of difference in supply and demand matching degree between each spatial grid.

[0016] The layout method and system for urban food facilities provided in this invention have the following advantages compared with the prior art: This invention performs time-series decomposition on multi-period population heat maps and divides spatial grid types according to intraday amplitude ratios. It reconstructs demand representation from static spatial snapshots into dynamic three-dimensional surfaces with phase evolution characteristics. The resulting time-varying demand values ​​fully mine multi-period population dynamic information, fundamentally breaking through the limitations of modeling based on single-period snapshots. It completely preserves the periodic structure, trend, and random fluctuation information of population distribution in the time dimension, ensuring the applicability and accuracy of food facility layout schemes in dynamic urban environments. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating a method for laying out urban food facilities in one embodiment; Figure 2 This is a schematic diagram of the layout system of an urban food facility provided in one embodiment. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0019] This invention provides a method for the layout of urban food facilities, the method comprising: Acquire multi-time-period population heat map data, food facility data, and road network data for urban areas, and divide the urban area into multiple spatial grids.

[0020] The population heatmaps of each spatial grid across multiple time periods are decomposed over time to obtain the time-varying population characteristic components of each spatial grid at different time periods. Based on the intraday amplitude ratio of the spatial grids, all spatial grids are classified into multiple types. For different types of spatial grids, the time-varying population characteristic components are differentially weighted to obtain the time-varying demand values ​​of each spatial grid at each time period.

[0021] Based on food facility data and road network data, the shortest path distance from each spatial grid to each food facility is determined, and the accessibility score of each spatial grid to each type of food facility is determined at each time period using the corresponding distance decay function according to the type of food facility.

[0022] Based on the time-varying demand value and accessibility score of each spatial grid in each time period, the supply and demand matching degree of each spatial grid in each time period is determined.

[0023] A multi-objective optimization model is constructed with the overall coverage rate of each time period not lower than a preset lower limit as a hard constraint, and the optimization objectives of maximizing the coverage rate during peak hours, minimizing the total road network travel distance, and achieving spatial balance as optimization goals. The model is then iteratively solved to obtain the layout scheme of food facilities within the urban area.

[0024] Among them, the overall coverage rate is the proportion of spatial grids whose supply and demand matching degree meets the preset conditions within a single time period to all spatial grids, and the spatial balance is measured based on the degree of difference in supply and demand matching degree between each spatial grid.

[0025] A specific embodiment of a method for the layout of urban food facilities is provided: First, multi-source spatiotemporal data are preprocessed to unify the spatial coordinate systems and temporal granularity of different sources, and food facility points of interest (POIs) are standardized by type coding. Based on this, multi-time-period (hourly × multi-day) population heatmaps are decomposed temporally, extracting multi-day baseline components, diurnal fluctuation components, and weekend deviation components. Accordingly, the spatial grid is automatically classified into commuter, residential, and steady-state types, constructing a demand surface with a temporal structure.

[0026] In terms of accessibility assessment, the shortest path distance between grids and facilities is calculated based on open street map road network data. Differentiated walking / cycling / driving distance decay functions are calibrated according to food facility type. A two-step movement search method with road network constraints is used to calculate the accessibility of each grid to each type of facility at each time period.

[0027] After completing demand modeling and accessibility calculation, dynamic supply and demand matching indexes are calculated on spatiotemporal grid cells to identify persistent and intermittent gaps.

[0028] Finally, a time-varying constrained multi-objective optimization model is constructed with the triple objectives of maximizing peak coverage, minimizing total road network travel distance, and achieving spatial balance. This model solves for the optimal location, type, and capacity of newly added or adjusted facilities, and outputs the layout scheme.

[0029] like Figure 1 As shown, the specific implementation process is as follows: S1 Multi-Source Data Preprocessing: Reads and parses food facility POI data, population heatmap data, OSM road network data, and population raster data. Automatically identifies the spatial coordinate systems of each dataset (WGS84, GCJ02, BD09, etc.) and uniformly transforms them to the target coordinate system (WGS84). Indexes the population heatmap data by date and time period to form... Population matrix N The total number of spatial grids, T The total number of time periods is 119 (7 consecutive days, 17 time periods per day). The type encoding of the POI data is mapped and standardized.

[0030] S2 Population Heatmap Time Series Decomposition: For each spatial grid i of T The population time-series vector is additively decomposed to extract multi-day baseline components, diurnal fluctuation components, and weekend deviation components. The decomposition method adopts classical time series decomposition, first extracting the trend period using moving averages, and then separating the seasonal term from the residuals.

[0031] For each spatial grid i Population time series decomposition is as follows: ; in, For spatial grid i During the period t Population thermal value, For spatial grid i Multi-day baseline components, For spatial grid i The diurnal fluctuation component, For spatial grid i The weekend deviation component, h ( t(Time period) t The corresponding hourly index, d ( t (Time period) t The corresponding weekday attribute (weekday / weekend). For the residual term, i For spatial grids, t For time periods.

[0032] S3 Grid Type Classification: Calculate the intraday amplitude ratio (daily average peak / daily average trough) of each grid and classify the grid into three categories based on preset upper and lower thresholds: commuter type (amplitude ratio higher than the upper threshold, daytime population concentration), residential type (amplitude ratio lower than the lower threshold, nighttime population concentration), and steady-state type (amplitude ratio in the middle, small fluctuation throughout the day). The classification threshold is adaptively determined through regional statistical distribution.

[0033] Implementation process of grid type classification: (1) For each spatial grid i, extract the population heat map values ​​for 17 time periods each day within a week, and take the median of each hour over 7 days according to the hour index h=0,1,…,16 (corresponding to 06:00–22:00) to form the typical daily population curve for that grid. .

[0034] (2) Calculate the intraday peak value on a typical daily curve. and intraday low Calculate the intraday volatility ratio .

[0035] (3) All N spatial grids Sort the values ​​in ascending order and calculate their tertiles: , As an adaptive threshold: lower threshold = Upper threshold = ; (4) Classification rules: like > Marked as commuter type, indicating significant daytime population concentration and large intraday fluctuations; like < Marked as residential, this indicates significant nighttime population concentration and small intraday fluctuations; like ≤ ≤ The value is marked as steady-state, indicating that the population fluctuates relatively little throughout the day.

[0036] S4 Time-Varying Demand Surface Generation: This involves overlaying the multi-day baseline components of the time-series decomposition with the intraday fluctuation components and weekend deviation components, and combining this with grid type labels to generate... The time-varying demand matrix. The demand value of each grid can be weighted and adjusted according to its type to reflect the differences in food consumption rhythms in different functional areas.

[0037] Implementation process: (1) Construct a weighted matrix of time period types The rows correspond to three grid types (commuter, residential, and steady-state), and the columns correspond to T time periods. The weighting coefficients are designed with the following logic: a certain type of grid receives a weight higher than 1 during periods of high population density (reflecting the concentration of food demand during that period) and a weight lower than 1 during periods of low population density.

[0038] Example: Commuter-type grid: Weekdays 08:00–12:00 and 13:00–18:00 weight w=1.2, lunch break 12:00–13:00 weight w=1.3, night 20:00–22:00 weight w=0.7; Residential grid: Weight w=1.1 for weekday mornings 06:00–08:00, weight w=1.2 for nighttimes 18:00–22:00, weight w=0.6 for midday 12:00–13:00; weight w=0.6 for 08:00–12:00 and 13:00–18:00. Steady-state grid: The weight is always w=1.0 throughout the entire time period.

[0039] (2) For each spatial grid i Its type label is known. {Commuting, Residence, Steady State} represents the sum of time-varying demographic components (multi-day baseline components + daytime fluctuation components + weekend deviation components). With the corresponding weighting coefficients Multiplying them together gives the time-varying demand value: in (That is, the sum of the three characteristic components).

[0040] (3) The weighting of weekends and weekdays uses an independent time period type weighting matrix, and the weekend weighting coefficient reflects the unique consumption pattern of weekends.

[0041] S5 Road Network Shortest Path Calculation: Constructing a Directed Weighted Graph Based on Road Network Data , where the vertex V For each spatial grid, the road network projection points are the intersections and the centroids of the grid, with edge weights equal to the road segment length or estimated travel time. i With each facility objectj The combination of these factors is used to calculate the shortest path distance of the road network using Dijkstra's shortest path algorithm. To control the computational scale, an upper limit (5km) can be set for the search radius; facilities exceeding this radius will be marked as unreachable.

[0042] S6 Typed Accessibility Decay Calculation: Based on Food Facility Type k Select the corresponding distance decay function . Let this be the distance decay function, when the road network distance... d Exceeding the maximum acceptable distance for this type When the attenuation value is set to 0, the attenuation value is set to 0.

[0043] ; ; in, Food facility type k Distance decay function, Food facility type k The maximum acceptable distance, Food facility type k The distance attenuation coefficient makes It decays to approximately 0.5 over time. k This refers to a type of food facility.

[0044] S7 network constraint two-step moving search: Step 1, for each facility object j (Type of food facility) k The first step is to calculate the demand-weighted supply ratio within the weighted coverage area of ​​the given time period; the second step is to calculate the demand-weighted supply ratio for each spatial grid. i The accessibility score for that type of facility is obtained by summing the attenuated demand-weighted supply ratios of all facilities within its reachable range. .

[0045] Two-step mobile search constrained by road network – Step 1 (facilities supply ratio) ; Two-step movement search with road network constraints – Step 2 (grid reachability) ; in, j For facility objects, For spatial grid i to facilities j The shortest path distance on the road network, For type k facilities j Supply capacity For facility objectsj (Type of food facility) k During the time period t The demand-weighted supply ratio For spatial grid i During the period t Type of food facility k The accessibility score.

[0046] S8 spatiotemporal grid supply and demand matching: for each spatiotemporal unit Aggregate the accessibility scores of all facility types into the total supply, and divide by the food demand for that grid during that time period. To obtain the matching index .

[0047] ; in, For spatial grid i During the period t The degree of supply and demand matching For spatial grid i During the period t Food needs.

[0048] S9 Gap Classification and Recognition: For each spatial grid i ,statistics exist T Performance within each time period: If all time periods are below the threshold, it is marked as a persistent gap; if only some time periods are below the threshold, it is marked as an intermittent gap; if all time periods meet the requirements, it is marked as covered.

[0049] S10 Time-Varying Constrained Multi-Objective Optimization: A three-objective optimization model is constructed, with the constraint that the overall coverage rate in each time period is not lower than a preset lower limit. After transforming the multi-objective model into a scalar weighted objective function using a weighted summation method, a heuristic algorithm (genetic algorithm) is used to solve the model, outputting candidate solutions in the Pareto front.

[0050] Implementation process: (1) Decision variable coding: The location and type of candidate facilities are coded as decision variables. , where p is the candidate location index and k is the facility type. The candidate location set includes: adjustment locations for existing facilities and candidate grids for new facilities (selected from the persistent gap grids with the lowest supply-demand matching degree).

[0051] (2) Time-varying constrained multi-objective optimization—three objective functions: Maximize overall coverage: Minimize total road network travel distance: Spatial equilibrium objective (Gini coefficient form): in, The minimum matching threshold. For decision variables (whether to add a new type of food facility) k The facility is located at candidate location p).

[0052] (3) Time-varying minimum coverage constraint: ; in, For time period t Overall coverage Minimum coverage time constraint, This is an indicator function; it takes the value 1 if the condition is true, and 0 otherwise.

[0053] (4) Weighted summation scalarization: After normalizing the three objectives, perform a weighted summation to transform it into a single objective function: in Normalized values ​​for each objective (normalized using the maximum and minimum values ​​of each objective in the feasible region). For target weight and The weights are traversed through the Pareto front by systematically changing the weight combinations.

[0054] (5) Solving using genetic algorithm: A. Population initialization: Randomly generate P feasible solutions (each solution is a binary vector, indicating whether each candidate location is equipped with each type of facility), ensuring that each solution satisfies the time-varying coverage constraint (if not, regenerate or correct using the penalty function method); B. Fitness function design: Since the genetic algorithm performs maximization optimization by default, we take the negative value of F to impose a penalty term on individuals that violate the time-varying constraints; C. Selection operator: Roulette wheel selection, which assigns the probability of being selected based on the individual's fitness. Individuals with higher fitness have a greater probability of being passed on to the next generation. D. Crossover operator: Single-point crossover, using single-point crossover to achieve parental gene recombination. A random split point is selected on the binary encoded vector, and gene segments on one side of the split point of two parent individuals are swapped to generate offspring;

[0055] E. Mutation operator: Bit flip mutation, iterates through each bit of the individual binary code, generates a random number, and if the value of the bit is less than the mutation probability, then the value of that bit is flipped to mutate the probability. F. Termination of iteration: The maximum number of iterations is reached or the optimal solution is not significantly improved after Δ iterations.

[0056] (6) Pareto front output: through systematically changing By combining these solutions and repeatedly solving the single-objective problem, a Pareto front is obtained, consisting of a set of non-dominated solutions. The final layout scheme is then selected from the Pareto front according to the decision-maker's preferences.

[0057] S11 Layout Scheme Generation Output: Select the final scheme from the candidate solutions based on the decision-maker's preferences. The output includes: location coordinates (latitude and longitude) of the added or adjusted facilities, facility type codes, suggested service capacity, and supporting visualizations such as gap coverage heatmaps and matching degree comparison charts before and after optimization.

[0058] This invention provides a layout system for urban food facilities, the system comprising: The data acquisition module is used to acquire multi-time period population heat map data, food facility data, and road network data of urban areas, and divide the urban area into multiple spatial grids.

[0059] The demand modeling module performs time-series decomposition of the multi-period population heatmap for each spatial grid, obtaining the time-varying population characteristic components of each spatial grid at different time periods. It then classifies all spatial grids into various types based on their intraday amplitude ratio. For different types of spatial grids, the time-varying population characteristic components are weighted differently to obtain the time-varying demand values ​​for each spatial grid at each time period.

[0060] The accessibility determination module is used to determine the shortest path distance from each spatial grid to each food facility based on food facility data and road network data, and to determine the accessibility score of each spatial grid to each type of food facility at each time period using the corresponding distance decay function according to the type of food facility.

[0061] The supply and demand matching module is used to determine the supply and demand matching degree of each spatial grid in each time period based on the time-varying demand value and accessibility score of each spatial grid in each time period.

[0062] The layout optimization module is used to construct a multi-objective optimization model with the overall coverage rate of each time period not being lower than a preset lower limit as a hard constraint, and the optimization objectives being to maximize the coverage rate during peak periods, minimize the total road network travel distance, and achieve spatial balance. The model is then iteratively solved to obtain the layout scheme of food facilities within the urban area.

[0063] Among them, the overall coverage rate is the proportion of spatial grids whose supply and demand matching degree meets the preset conditions within a single time period to all spatial grids, and the spatial balance is measured based on the degree of difference in supply and demand matching degree between each spatial grid.

[0064] like Figure 2 As shown, a specific embodiment of a layout system for urban food facilities is provided: 1. Data layer.

[0065] Food facility POI data: Includes coordinates (supports multiple coordinate systems), type codes, names, addresses, and other information for various food facilities within the target city. Facility types include, but are not limited to: restaurants, farmers' markets, convenience stores, supermarkets, fresh food supermarkets, wholesale markets, cold chain facilities, and flexible distribution points.

[0066] Population heatmap data: population heat map data with hourly granularity over multiple consecutive days. Each record contains grid coordinates and the corresponding population estimate.

[0067] Road network data: Road network data in open street map format, including attributes such as road segment geometry, length, and road class.

[0068] Population raster data: Medium to high resolution annual population distribution raster, used as a static baseline reference.

[0069] 2. Data preprocessing module.

[0070] The above-mentioned multi-source heterogeneous data are standardized to provide a unified spatiotemporal reference and data structure for subsequent modules.

[0071] Coordinate system cleaning: Identify the coordinate systems of different data (such as WGS84, GCJ02, BD09) and convert them to the target coordinate system; remove coordinate outliers and duplicate records.

[0072] Time series alignment: Align population heat map data by time period (hour) to form a neat time series matrix (spatial grid × time period).

[0073] POI type coding standardization: Mapping the classification codes of food facilities from different sources to a unified type system to ensure the consistency of classification in subsequent typed accessibility calculations.

[0074] Relationship: This module provides cleaned and standardized input data for the requirements modeling module and the accessibility calculation module.

[0075] 3. Requirements Modeling Module.

[0076] The population heat map of multiple time periods is transformed into an information demand surface with a time structure, and the grid is classified based on the time series characteristics.

[0077] The time-series decomposition submodule uses an additive decomposition method to decompose the population heat map time series of each grid into a baseline component (multi-day averages for each time period), a diurnal fluctuation component (the deviation of the current day from the baseline), and a weekend deviation component (systematic differences between weekdays and weekends). The decomposition uses the moving average method to extract trend, periodic, and residual components.

[0078] Grid Classification Submodule: Based on the intraday amplitude ratio (the ratio of daytime peak to nighttime trough) and weekend deviation of each grid, automatically label the grid type as commuter type (daytime population is significantly higher than nighttime population), residential type (nighttime population is significantly higher than daytime population), and steady-state type (daily population fluctuation is relatively small).

[0079] Time-varying demand surface generation: Combining time-series decomposition results with grid classification labels, a multi-dimensional time-varying demand surface is generated as input to the supply and demand matching module.

[0080] Relationship: Obtain unified heat map data from the data preprocessing module and output time-varying demand surface to the supply and demand matching module.

[0081] 4. Reachability calculation module.

[0082] The shortest path distance from each grid to each food facility is calculated based on the real road network, and an accessibility score is calculated by introducing a distance decay function specific to facility type.

[0083] Road network shortest path calculation: A directed weighted graph is constructed based on OSM road network data. Dijkstra's algorithm is used to calculate the shortest path distance along the road network from the centroid of each spatial grid to each food facility.

[0084] Typed attenuation function: Different distance attenuation parameters are set for different types of food facilities. Convenience stores use a 5-minute walk (approximately 400m) as the attenuation radius; farmers' markets use a 15-minute walk (approximately 1200m); fresh food supermarkets use a 10-minute walk (approximately 800m); and wholesale markets use a 20-minute drive. The attenuation function is a piecewise Gaussian function.

[0085] The road network-constrained two-step mobile search is as follows: First, for each facility, calculate the ratio of its supply capacity to the weighted population of the surrounding accessible grids (facility-grid supply ratio). Second, for each grid, sum the attenuated weighted supply ratios of all facilities within its reach to obtain the grid's accessibility score for that type of facility. Both steps are constrained by the shortest path distance in the road network.

[0086] Relationship: Obtain POI and road network data from the data preprocessing module, and output the accessibility matrix of each grid to each facility type for each time period to the supply and demand matching module.

[0087] 5. Supply and demand matching module.

[0088] Quantify the degree of matching between supply and demand on a spatiotemporal grid unit, and identify and classify coverage gaps.

[0089] Dynamic matching index calculation: The supply and demand matching index is calculated for each spatiotemporal unit—the total supply weighted by accessibility in the current time period of the grid is divided by the population demand in the current time period of that grid. An index greater than 1 indicates ample supply, while a value less than 1 indicates a supply gap.

[0090] Gap classification and identification: The matching degree sequence of each grid across time periods is analyzed: if the matching degree of all time periods is lower than the threshold, it is marked as "persistent gap"; if the matching degree is lower than the threshold only during peak time periods and is satisfied during off-peak time periods, it is marked as "intermittent gap"; if it is satisfied in all time periods, it is marked as "covered".

[0091] The time-varying demand surface is obtained from the demand modeling module, the reachability matrix is ​​obtained from the reachability calculation module, and the gap type map and matching degree matrix are output to the layout optimization module.

[0092] 6. Layout optimization module.

[0093] Based on the supply and demand matching and gap analysis results, the optimal solution for adding or adjusting food facilities is found.

[0094] Time-varying constraint construction: "The overall coverage rate in each time period shall not be lower than the preset minimum level" is used as a hard constraint to ensure that the optimization results do not sacrifice the coverage requirements of any time period.

[0095] Multi-objective optimization solution: Construct an optimization model with three objectives—maximizing peak-hour coverage, minimizing the total travel distance of residents to the nearest facility, and maximizing spatial equilibrium (minimizing the Gini coefficient between coverage areas of different grids). The Pareto front is solved using either the weighted summation method or the ε-constraint method.

[0096] Solution generation and output: Combine the optimization results (location coordinates of new facilities, facility type, and suggested capacity) with visualization outputs such as gap maps and matching degree heat maps to form a complete layout decision recommendation.

[0097] Relationship: Receives the analysis results from the supply and demand matching module and outputs the final facility layout plan.

[0098] The embodiments described above are merely examples of several implementations of the present invention, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention.

Claims

1. A method for arranging urban food facilities, characterized in that, include: Acquire multi-time period population heat map data, food facility data, and road network data for urban areas, and divide the urban area into multiple spatial grids; The population heatmap of each spatial grid is decomposed over time to obtain the time-varying population characteristic components of each spatial grid in different time periods. Based on the intraday amplitude ratio of the spatial grid, all spatial grids are divided into multiple types. For different types of spatial grids, the time-varying population characteristic components are differentially weighted to obtain the time-varying demand value of each spatial grid in each time period. Based on the food facility data and the road network data, the shortest path distance from each spatial grid to each food facility is determined, and the accessibility score of each spatial grid to each type of food facility is determined at each time period using the corresponding distance decay function according to the type of food facility. Based on the time-varying demand value and the accessibility score of each spatial grid in each time period, the supply and demand matching degree of each spatial grid in each time period is determined. A multi-objective optimization model is constructed with the overall coverage rate of each time period not lower than a preset lower limit as a hard constraint, and the optimization objectives of maximizing the coverage rate during peak periods, minimizing the total road network travel distance, and achieving spatial balance as optimization goals. The model is then iteratively solved to obtain the layout scheme of food facilities within the urban area. The overall coverage rate is the proportion of spatial grids whose supply and demand matching degree meets the preset conditions within a single time period to all spatial grids, and the spatial balance is measured based on the degree of difference in supply and demand matching degree between each spatial grid.

2. The method for arranging urban food facilities as described in claim 1, characterized in that, The time-varying population characteristic components include: multi-day baseline component, daytime fluctuation component, and weekend deviation component; The multi-day baseline component represents the average population level of the spatial grid at different times of the day; The diurnal fluctuation component characterizes the degree of deviation of the real-time population of the spatial grid on a specific date from the multi-day baseline component; The weekend deviation component characterizes the systematic difference between the spatial grid on weekends and weekdays.

3. The method for arranging urban food facilities as described in claim 1, characterized in that, The method of classifying all spatial grids into multiple types based on the intraday amplitude ratio of the spatial grids specifically includes: Obtain population heat values ​​for each spatial grid at multiple sampling times within a preset time period, and construct population temporal change curves for each grid based on the population heat values; The daily population peak and daily population trough are extracted based on the population time change curve, and the ratio of the daily population peak to the daily population trough is used as the daily amplitude ratio. The intraday amplitude ratios of all spatial grids are statistically sorted, and the classification thresholds are determined based on the ternary values ​​of the intraday amplitude ratios of all spatial grids. The classification thresholds include an upper threshold and a lower threshold. Spatial grids with intraday amplitude ratios higher than the upper threshold are classified as commuter grids, indicating significant daytime population concentration and large intraday amplitude. Spatial grids with intraday amplitude ratios below the lower threshold are classified as residential, indicating significant nighttime population concentration and low intraday amplitude. Dividing the spatial grid where the intraday amplitude ratio falls between the lower threshold and the upper threshold into a steady-state type indicates that the population fluctuations throughout the day are relatively small.

4. The method for arranging urban food facilities as described in claim 1, characterized in that, The method involves differentially weighting the time-varying population feature components for different types of spatial grids to obtain the time-varying demand values ​​for each spatial grid at each time period, specifically including: Obtain the basic time-varying population feature components of each spatial grid at each time period; Based on the type label of each spatial grid, the weighting coefficient of each spatial grid in each time period is determined from the preset weighting coefficient matrix. The row elements of the weighting coefficient matrix correspond to different grid types, and the column elements of the weighting coefficient matrix correspond to each time period. The weighting coefficient of the same type of grid is different in different time periods, and the weighting coefficient of different types of grids is different in the same time period. The weighting coefficient is used to characterize the degree of concentration or dispersion of food demand of the corresponding type of grid relative to the average level in the corresponding time period. The time-varying population characteristic components of each spatial grid in each time period are multiplied by the corresponding weighting coefficients to obtain the time-varying demand value of each spatial grid in each time period.

5. The method for arranging urban food facilities as described in claim 1, characterized in that, The step of determining the shortest path distance from each spatial grid to each food facility based on the food facility data and the road network data specifically includes: A directed weighted graph is constructed based on the road network data of the urban area. The vertices of the directed weighted graph are the road intersections and the projection points of the centroid of the spatial grid onto the road network. The edges of the directed weighted graph are the road segments connecting adjacent vertices in the road network data. The weight of the edge is the road segment length or the estimated travel time. For each combination of spatial grid and each food facility, the shortest path distance along the directed weighted graph is determined by the shortest path algorithm; when the shortest path distance exceeds a preset search radius, the food facility is marked as unreachable.

6. The method for arranging urban food facilities as described in claim 1, characterized in that, The process of determining the accessibility score of each spatial grid to each type of food facility at each time period using a corresponding distance decay function based on the type of food facility specifically includes: The maximum acceptable distance and distance attenuation coefficient are determined according to the type of food facility, wherein the distance attenuation function is a piecewise Gaussian function; Based on the shortest path distance from each spatial grid to each food facility and the distance decay function value, the accessibility score of each spatial grid to each type of food facility at each time period is determined by a road network-constrained mobile search method, including: For each food facility, the ratio of the supply capacity of the food facility to the sum of the time-varying demand values ​​of all spatial grids within the reach of the food facility after distance attenuation weighting is determined to obtain the demand-weighted supply ratio of the food facility in each time period. For each spatial grid, the demand-weighted supply ratio of all food facilities within the reach of the spatial grid in each time period is summed after being weighted by distance attenuation to obtain the accessibility score of the spatial grid for each type of food facility in each time period.

7. The method for arranging urban food facilities as described in claim 1, characterized in that, The determination of the supply-demand matching degree of each spatial grid in each time period based on the time-varying demand value and the accessibility score of each spatial grid in each time period specifically includes: The accessibility scores of each spatial grid to each type of food facility are aggregated at each time period to obtain the total supply of each spatial grid at each time period. The total supply and the time-varying demand value of each spatial grid at each time period are used as the supply and demand matching degree of each spatial grid at each time period.

8. The method for arranging urban food facilities as described in claim 1, characterized in that, Also includes: Based on the comparison results of the supply and demand matching degree of each spatial grid at each time period with the preset matching degree threshold, each spatial grid is marked as a persistent gap, an intermittent gap, or covered. Specifically, for each spatial grid, if the supply and demand matching degree of the spatial grid is lower than the preset matching degree threshold in all time periods, the spatial grid is marked as a persistent gap; if the supply and demand matching degree of the spatial grid is lower than the preset matching degree threshold only in some time periods, the spatial grid is marked as an intermittent gap; if the supply and demand matching degree of the spatial grid is not lower than the preset matching degree threshold in all time periods, the spatial grid is marked as covered. The labeling results are used as input constraints or optimization guidance information for the multi-objective optimization model.

9. A method for arranging urban food facilities as described in claim 8, characterized in that, The process involves constructing a multi-objective optimization model with the hard constraint that the overall coverage rate in each time period is not lower than a preset lower limit, and with the optimization objectives of maximizing the coverage rate during peak periods, minimizing the total road network travel distance, and achieving spatial balance. This model is then iteratively solved to obtain the layout scheme of food facilities within the urban area. Specifically, this includes: The location and type of candidate facilities are encoded as decision variables, and the candidate locations include: the adjustment locations of existing facilities and candidate grids for new facilities selected from the persistent gap grids with the lowest supply-demand matching degree; A time-varying constrained multi-objective optimization model is constructed. The objective functions of the time-varying constrained multi-objective optimization model include: an overall coverage maximization objective, which maximizes the comprehensive level of coverage in each time period; a total road network travel distance minimization objective, which minimizes the sum of the time-varying demand value of each spatial grid and the shortest path distance to the nearest food facility in each time period; and a spatial balance objective, which minimizes the degree of difference in the average supply and demand matching degree between each spatial grid, wherein the degree of difference is measured in the form of the Gini coefficient. The multiple objective functions of the time-varying constrained multi-objective optimization model are transformed into a single objective function by weighted summation. The single objective function is then iterated using a heuristic algorithm. The candidate solutions in the Pareto front obtained after the iteration are used as the layout schemes of food facilities in the urban area.

10. A layout system for urban food facilities, characterized in that, include: The data acquisition module is used to acquire multi-time period population heat map data, food facility data, and road network data of the urban area, and divide the urban area into multiple spatial grids; The demand modeling module is used to perform time-series decomposition of the multi-period population heat map of each spatial grid to obtain the time-varying population characteristic components of each spatial grid in different time periods, and to classify all spatial grids into multiple types according to the intraday amplitude ratio of the spatial grids; for different types of spatial grids, the time-varying population characteristic components are differentially weighted to obtain the time-varying demand value of each spatial grid in each time period. The accessibility determination module is used to determine the shortest path distance from each spatial grid to each food facility based on the food facility data and the road network data, and to determine the accessibility score of each spatial grid to each type of food facility in each time period using the corresponding distance decay function according to the type of food facility. The supply and demand matching module is used to determine the supply and demand matching degree of each spatial grid in each time period based on the time-varying demand value and the accessibility score of each spatial grid in each time period. The layout optimization module is used to construct a multi-objective optimization model with the overall coverage rate of each time period not being lower than a preset lower limit as a hard constraint, and the optimization objectives being to maximize the coverage rate during peak periods, minimize the total road network travel distance, and achieve spatial balance. The model is then iteratively solved to obtain the layout scheme of food facilities within the urban area. The overall coverage rate is the proportion of spatial grids whose supply and demand matching degree meets the preset conditions within a single time period to all spatial grids, and the spatial balance is measured based on the degree of difference in supply and demand matching degree between each spatial grid.

Images