Online car-hailing supply and demand prediction method fused with real-time anomaly detection

Abstract

The invention discloses an online car-hailing supply and demand prediction method fused with real-time anomaly detection, and relates to the technical field of supply and demand prediction management, and the method comprises the steps: determining the total length of a curb according to a line segment set of the curb of an urban road; generating a grid feasible region set according to the maximum closure capacity and the peak incoming water flow of the adjacent gutter inlet corresponding to the curb line segment; and calculating the reachable dry curb proportion based on the total dry curb length and the total curb length of the grid feasible region set. According to the method, the situation of virtual high prediction value or service failure caused by supply and demand mismatching in a traditional method is avoided.

Description

Technical Field

[0001] This invention relates to the field of supply and demand forecasting and management technology, and in particular to a method for forecasting the supply and demand of ride-hailing services that integrates real-time anomaly detection. Background Technology

[0002] Currently, with the rapid growth of urban travel demand, ride-hailing services play a crucial role in improving urban traffic efficiency and meeting diverse travel needs. However, in complex urban environments, especially affected by sudden weather changes such as rainfall, urban road conditions often fluctuate drastically, leading to frequent problems such as curb flooding, road closures, and passenger waiting inconvenience. This uncertainty significantly increases the difficulty of ride-hailing operation scheduling, posing challenges to order completion rates, driver response efficiency, and user satisfaction. Traditional static scheduling and fixed-site supply and demand forecasting strategies are no longer suitable for dynamically changing supply and demand environments, especially lacking the ability to effectively integrate real-time road conditions, road accessibility, and meteorological factors. In extreme scenarios such as rainy days, some roads, although geographically accessible, may be unusable for passenger pick-up and drop-off due to poor drainage or short-term flooding. This urgently requires a supply and demand forecasting and control method that can dynamically identify dry, usable curb areas, match available drivers and passenger requests in real time, and has scheduling guidance capabilities.

[0003] Traditional solutions to urban transportation problems typically rely on historical order data or fixed-area modeling, which struggles to effectively reflect changes in roadside availability and service capacity under current weather conditions. This leads to significant prediction bias, delayed dispatch response, and inefficient resource allocation. Existing methods fail to adequately consider the relationship between storm drain capacity and inflow intensity in urban roads, cannot precisely label the availability of specific road segments, and lack a dynamic supply-demand constraint model. Consequently, they cannot support efficient and reliable dispatch decisions under abnormal conditions such as rainfall. Summary of the Invention

[0004] The purpose of this invention is to address the shortcomings of existing technologies that fail to establish a dynamic constraint model between supply and demand, and cannot support efficient and reliable scheduling decisions under abnormal conditions such as rainfall. The proposed invention is a ride-hailing supply and demand forecasting method that integrates real-time anomaly detection.

[0005] To address the problems existing in the prior art, the present invention adopts the following technical solution: A ride-hailing supply and demand forecasting method integrating real-time anomaly detection includes: S1. Determine the total length of the curb based on the set of line segments of the urban road curb; S2. Generate a set of feasible grid regions based on the maximum interception capacity and peak inflow of the adjacent storm drains corresponding to the curb segment; S3. Calculate the proportion of reachable dry curbs based on the total length of dry curbs and the total length of curbs in the grid feasible region set; S4. Based on the proportion of accessible dry curbs and the spatial area grid in urban road curbs, calculate the feasible supply, and calculate the achievable passenger pick-up intensity based on the set of feasible regions in the grid and the set of passenger requests for all currently pending orders. S5. Based on achievable passenger boarding intensity, feasible supply and the proportion of accessible dry curbs, determine the predicted demand gap for the next time window; S6. Dynamically adjust and optimize ride-hailing dispatch and order assignment strategies based on the number of orders that can be completed in the next time window and the predicted demand gap.

[0006] Preferably, the total length of the curb is determined based on the set of line segments of the urban road curb, including: The urban road curb is divided according to the road shape to obtain a set of line segments, which includes the number and length of the curb line segments. Associate the effective catchment area of ​​the curb segment with that of the adjacent storm drain inlet of the curb segment; The total curb length is obtained by summing the lengths of the line segments within the same grid.

[0007] Preferably, a set of feasible grid regions is generated based on the maximum interception capacity and peak inflow of adjacent storm drains corresponding to the curb segment, including: The inflow rate per unit time is obtained by multiplying the rainfall intensity provided by the external meteorological system with the effective catchment area corresponding to the curb segment. The peak inflow rate is obtained by multiplying the inflow rate per unit time with the preset runoff coefficient. The difference between the maximum interception capacity and the peak inflow of the adjacent storm drain corresponding to the curb segment is calculated to obtain the capacity-water requirement difference. The curb segments are marked based on the relationship between the capacity-water-demand difference and 0, and the status marking results of the curb segments are obtained. The curb segments whose status marking results are dry and usable are filtered and grouped by grid to obtain the grid feasible region set.

[0008] Preferably, based on the total length of dry curb and the total curb length of the grid feasible region set, the proportion of reachable dry curb is calculated, including: The total length of dry curb is obtained by summing the lengths of all curb segments in the feasible region set of the grid. The ratio of the total length of dry curb to the total length of curb is calculated to obtain the proportion of achievable dry curb.

[0009] Preferably, based on the proportion of accessible dry curbs and the spatial area grid in urban road curbs, the feasible supply is calculated, including: Supply intensity is calculated for spatial area grids of urban road curbs based on distance weighting and driver online status, thus obtaining the visible supply of the spatial area grids; The executable supply is obtained by linearly scaling the visible supply and the reachable dry curb ratio.

[0010] Preferably, the calculation of passenger boarding intensity based on the set of feasible grid regions and the set of all passenger requests currently awaiting dispatch includes: Spatial encapsulation and attribute encoding are performed on the set of feasible grid regions to obtain a dry curb layer; Get the set of all passenger requests currently pending dispatch; Based on the dry curb layer, each request in the passenger request set is redirected to the boarding point to obtain a successfully matched subset of requests; The successful matching request subsets are categorized and statistically analyzed by grid to obtain the achievable passenger load strength.

[0011] Preferably, the predicted demand gap for the next time window is determined based on achievable passenger boarding intensity, feasible supply, and the proportion of accessible dry curbs, including: The available capacity is obtained by linearly multiplying the proportion of accessible dry curbs and the baseline capacity of curb segments under normal weather conditions. Minimize the achievable passenger load intensity, feasible supply, and available capacity to obtain the number of orders that can be completed in the next time window; The difference between the achievable passenger load intensity and the achievable order volume in the next time window is calculated to obtain the predicted demand gap.

[0012] Preferably, the ride-hailing dispatch and order assignment strategy is dynamically adjusted and optimized based on the number of orders that can be completed in the next time window and the predicted demand gap, including: The driver scheduling and order dispatch priority are dynamically allocated and optimized based on the number of orders that can be completed; The grid quota and vehicle guidance strategy are adjusted and supplemented in real time based on the predicted demand gap.

[0013] Compared with the prior art, the beneficial effects of the present invention are: 1. In this invention, by introducing the capacity-to-water-demand difference calculation, the maximum interception capacity and peak inflow per unit time of urban curb segments and adjacent storm drains are analyzed to establish a dynamic measurement relationship between surface rainfall runoff intensity and the actual carrying capacity of the drainage system. Combined with spatial grid screening, dry and usable curb segments are accurately assigned to the grid feasible domain set, and a dry curb layer is constructed. Through big data analysis, the real-time usable road resources are finely marked. Compared with the traditional method of judging passable areas based solely on historical trajectories or static maps, this method improves the dynamic response capability and spatial accuracy of road accessibility mapping, and provides reliable spatial boundary constraints for downstream supply and demand matching and scheduling optimization.

[0014] 2. In this invention, not only is the number of online drivers considered, but the spatial supply intensity is also calculated by combining the real-time status of drivers with the weighted coefficients of their distance from each grid. The proportion of accessible dry curbs is then introduced for linear scaling to obtain the executable supply. On the demand side, a passenger pick-up point redirection mechanism is introduced to map the original request to a dry, available actual parking area, ensuring that the demand input has the possibility of execution. Furthermore, through a three-dimensional constraint model, the minimum value of achievable pick-up intensity, executable supply, and available capacity is used to derive the number of orders that can be completed in the next time window, avoiding the situation of inflated predictions or service failures caused by supply-demand mismatch in traditional methods.

[0015] 3. In this invention, by calculating the difference between the achievable passenger load intensity and the achievable order volume, a predictive demand gap index is constructed. Based on this gap, a dynamic supplementary mechanism of grid quota redistribution and vehicle guidance strategy is implemented, thereby achieving systematic supply and demand self-balancing regulation. In areas with high gaps, drivers are guided to enter first and the order dispatch priority is increased to ensure service capacity in areas with high demand. In areas with low gaps, redundant scheduling is appropriately suppressed to save transportation resources. Attached Figure Description

[0016] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this application, illustrate exemplary embodiments of the invention and, together with their description, serve to explain the invention and do not constitute an undue limitation thereof. In the drawings: Figure 1 This is a flowchart illustrating a ride-hailing supply and demand prediction method that integrates real-time anomaly detection, as provided in an embodiment of the present invention. Detailed Implementation

[0017] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0018] Example: This example provides a ride-hailing supply and demand forecasting method that integrates real-time anomaly detection. See [link / reference]. Figure 1 Specifically, including: S1. Determine the total length of the curb based on the set of line segments of the urban road curb; In an embodiment of the present invention, determining the total length of the curb based on the set of line segments of the urban road curb includes: The urban road curb is divided according to the road shape to obtain a set of line segments, which includes the number and length of the curb line segments. Specifically, the process of segmenting urban road curbs according to road morphology includes the following steps: First, acquire spatial geographic information data of urban roads, including the geometric morphology information of road centerlines, road boundary lines, and road intersections. Then, extract curb lines along both sides of the road based on the road boundary lines. By identifying the degree of road curvature, intersections, entrances and exits, and the location of road grade changes, the continuous curb lines are segmented according to geometric characteristics to ensure that each curb segment has a relatively consistent slope, width, and drainage direction. Next, assign a unique number to each segmented curb line to ensure that different road segments can be independently indexed and retrieved in the database. Subsequently, use a geographic information system to calculate the actual length of each line segment and store the length value in correspondence with the line segment number to form a line segment set data structure. The final line segment set contains the number and corresponding length information of each curb line segment, providing basic data for subsequent steps such as establishing storm drain associations, calculating catchment areas, and grid aggregation.

[0019] Associate the effective catchment area of ​​the curb segment with that of the adjacent storm drain inlet of the curb segment; The total curb length is obtained by summing the lengths of the line segments within the same grid.

[0020] Specifically, the process of associating curb segments with the effective catchment areas of adjacent storm drains and summarizing the lengths of segments within the same grid includes the following steps: First, based on geographic information data of urban drainage networks and road ancillary facilities, the spatial coordinates, model dimensions, and catchment direction information of storm drains are obtained; then, using spatial proximity analysis, the shortest distance between each storm drain and the curb segment is calculated in the geographic coordinate system, and storm drains adjacent to the curb segment within a certain threshold range are identified; next, based on the road's cross slope, longitudinal slope, and catchment direction, the effective catchment area covered by each storm drain is identified. The system first defines the area and calculates the corresponding effective catchment area. Then, it establishes a one-to-one correspondence between the obtained effective catchment area values ​​and the corresponding curb segment numbers, forming a table linking curb segments and storm drain attributes. Following urban spatial division rules, all curb segments are assigned to their respective spatial grids, and the length data of all curb segments contained within each grid are extracted. Finally, the lengths of all curb segments within the same grid are numerically summarized to calculate the total curb length for that grid. This total length, along with the grid number, is recorded as the basic input data for subsequent calculations of the reachable dry curb ratio and supply-demand forecasting.

[0021] Specifically, urban road curbs refer to the linear areas along the edge of a road in a city, used for temporary vehicle parking, passenger pick-up / drop-off, or waiting. Their physical attributes include length, location, and relative relationship to the carriageway and drainage facilities. Road morphology refers to the geometric structure and orientation of a road in space, used to guide the spatial division of curbs. A curb segment is the smallest linear unit obtained by dividing a continuous curb according to its road morphology. Each segment has a unique number and a defined length, describing the spatially identifiable range of the curb. A storm drain is a drainage facility located at the road edge, used to collect and channel surface runoff into the drainage system. Its catchment capacity determines the water accumulation in the curb segment during rainfall. The effective catchment area refers to the actual surface area that can collect rainwater into the storm drain and adjacent curb segments during rainfall; its size reflects the potential water volume that the curb segment can withstand. A grid divides urban space into several adjacent, equal-scale, or custom-sized spatial units. Each grid is used to collect and statistically analyze the curb segments and their attributes contained within it. Total curb length refers to the sum of the lengths of all curb segments within the same grid, used to represent the total physical resource scale available for parking or passenger pick-up and drop-off in that area.

[0022] S2. Generate a set of feasible grid regions based on the maximum interception capacity and peak inflow of the adjacent storm drains corresponding to the curb segment; In an embodiment of the present invention, a set of feasible grid regions is generated based on the maximum interception capacity and peak inflow of adjacent storm drains corresponding to the curb segment, including: The inflow rate per unit time is obtained by multiplying the rainfall intensity provided by the external meteorological system with the effective catchment area corresponding to the curb segment. The peak inflow rate is obtained by multiplying the inflow rate per unit time with the preset runoff coefficient. The difference between the maximum interception capacity and the peak inflow of the adjacent storm drain corresponding to the curb segment is calculated to obtain the capacity-water requirement difference. Specifically, the capacity-requirement water difference calculation is based on the hydraulic balance principle of surface runoff and drainage systems. During rainfall, the product of rainfall intensity and catchment area represents the volumetric flow rate of water converted from atmospheric precipitation into surface runoff per unit time. This flow rate is collected at adjacent storm drains by the surface slope. Due to differences in ground materials, slope, and the resistance of the runoff path, some rainwater is lost through infiltration or retention. Therefore, a runoff coefficient is introduced in engineering calculations to correct the theoretical inflow, thereby obtaining a peak inflow rate that truly reflects the surface runoff conditions. Meanwhile, the maximum interception capacity of a storm drain is determined by its geometric cross-section, the opening ratio of the grid, and the hydraulic gradient of the stormwater pipe, representing the maximum flow rate that the facility can discharge under specific head conditions. When the difference between the peak inflow rate and the maximum interception capacity of the storm drain is calculated, the result reflects the system's drainage margin or overload level. A positive difference indicates that the drainage system still has a capacity; a zero or negative difference indicates that the runoff caused by rainfall exceeds the drainage limit, easily leading to localized waterlogging. Therefore, this difference can be used to quantify the dynamic balance between surface water inflow and drainage capacity, thus obtaining the capacity-water demand difference.

[0023] Specifically, rainfall intensity refers to the average amount of rainfall falling to the ground surface per unit time. Its magnitude reflects the intensity change of the rainfall process and is used to characterize the rate of precipitation input to the surface at a specific moment. Effective catchment area refers to the area of ​​the ground surface that can collect rainwater into a specific storm drain or flow along the road edge towards that storm drain under rainfall conditions. The larger the area, the more rainwater is collected. Inflow per unit time refers to the amount of water collected from the effective catchment area to a specific curb segment or storm drain within a specific time interval. It is usually expressed in volumetric flow rate and is used to characterize the instantaneous water input scale of that curb segment. Runoff coefficient is a proportionality coefficient describing the conversion of surface rainfall into surface runoff. It is determined by factors such as ground material, slope, and vegetation cover. A higher value indicates that rainwater is more likely to form runoff. Peak flow rate refers to the maximum flow rate that enters a specific curb segment per unit time during rainfall, and is used to reflect the most unfavorable drainage load under short-term heavy rainfall conditions. Maximum interception capacity refers to the maximum flow rate that a storm drain inlet can withstand and effectively discharge under specific hydraulic conditions. It is a key indicator for evaluating the drainage performance of storm drains. Capacity-Required Water Difference refers to the difference between the maximum interception capacity of a storm drain inlet and the actual peak inflow rate. A positive difference indicates sufficient drainage capacity, while a zero or negative difference indicates insufficient drainage capacity and may lead to roadside water accumulation.

[0024] The curb segments are marked based on the relationship between the capacity-water-demand difference and 0, and the status marking results of the curb segments are obtained. The curb segments whose status marking results are dry and usable are filtered and grouped by grid to obtain the grid feasible region set.

[0025] Specifically, the implementation steps for marking curb segments and obtaining a set of feasible grid regions based on the relationship between the capacity water requirement difference and zero include the following: First, import the calculated capacity water requirement difference data for each curb segment into the system database and create a unique index for each segment for batch judgment; then, the system compares the capacity water requirement difference of each curb segment according to the set logical judgment conditions. When the difference is greater than zero, the curb segment is marked as dry and usable; when the difference is less than or equal to zero, the curb segment is marked as waterlogged or unusable; next, the system generates... A state label table containing all curb segment numbers and corresponding state attributes is generated for subsequent spatial filtering. Then, the curb segments in the dry and usable state are extracted, and the spatial grid number to which each segment belongs is calculated according to its geographic coordinates and spatial grid division rules. These dry and usable segments are then clustered and grouped according to the grid number, and the dry segments in the same grid are stored as a set object. Finally, a set of dry and usable curb segments in each grid is formed. All these sets constitute the grid feasible domain set at the current moment, providing a spatial input basis for subsequent dryness ratio calculation, supply and demand forecasting, and dynamic scheduling.

[0026] Specifically, the status label is an operational status attribute assigned to each curb segment based on the numerical result of the water demand difference. This attribute distinguishes whether the segment is usable; segments with a positive difference are marked as dry and usable, while those with a negative difference are marked as waterlogged or unusable. A dry and usable curb segment refers to a section of the curb that, under current rainfall and drainage conditions, remains free of surface water and is suitable for vehicle parking or passenger drop-off / pick-up. Grid filtering involves assigning all curb segments marked as dry and usable to corresponding urban grid areas according to spatial division rules for regional statistics and spatial management. The grid feasible region set refers to the set of all curb segments judged as dry and usable and successfully assigned to the grid at the same time. This set represents the range of physically usable space in the city under given rainfall conditions.

[0027] Specifically, the determination of the feasible region set is based on the principles of continuous surface drainage and balanced spatial distribution. Under rainfall, surface runoff concentrates in low-lying areas along the road slope and catchment path. The drainage status of curb segments is affected by both the local storm drain capacity and the incoming water load. By calculating the difference in water demand, it can be determined whether each curb segment has the conditions for continuous drainage and dryness. When the difference is greater than zero, it indicates that rainwater can be drained away in time, the surface remains free of water accumulation, and it is passable. When the difference is less than or equal to zero, it indicates that the drainage capacity at that location is fully occupied or insufficient, the surface has water accumulation or water blockage, and it is not passable. Since road drainage is a spatially continuous process, dry and usable curb segments constitute a connectable feasible region in space, while waterlogged sections form a barrier region. By filtering and grouping line segments marked as dry according to the urban spatial grid, a set of multiple dry line segments can be obtained. This set corresponds to the area where the surface can still maintain normal function and vehicles can safely park or pick up and drop off passengers, i.e., the grid feasible domain set.

[0028] Specifically, the reason for generating a grid feasible region set based on the maximum interception capacity and peak inflow of adjacent storm drains corresponding to curb segments is that the usability of urban roads is directly constrained by the balance between surface runoff and drainage capacity. The maximum interception capacity of a storm drain represents the maximum flow that the area can discharge per unit time, while the peak inflow reflects the maximum actual water load during rainfall. When the inflow exceeds the interception capacity, rainwater cannot be drained in time, and water will accumulate on the surface, causing the curb to lose its parking or passenger pick-up / drop-off function. Conversely, when the interception capacity is greater than the inflow load, the surface remains dry and passable. By comparing the difference between the two, it is possible to accurately identify which curb segments are still usable and which are ineffective due to water accumulation. By grouping these dry and usable curb segments according to a spatial grid, a physical area that is still practically passable under the current rainfall conditions can be formed, i.e., the grid feasible region set, thus providing a real and dynamic spatial constraint basis for subsequent supply and demand forecasting and scheduling control.

[0029] S3. Calculate the proportion of reachable dry curbs based on the total length of dry curbs and the total length of curbs in the grid feasible region set; In an embodiment of the present invention, the proportion of reachable dry curbs is calculated based on the total length of dry curbs and the total curb length of the grid feasible region set, including: The total length of dry curb is obtained by summing the lengths of all curb segments in the feasible region set of the grid. The ratio of the total length of dry curb to the total length of curb is calculated to obtain the proportion of achievable dry curb.

[0030] Specifically, the length of a curb segment refers to the actual linear distance of a single curb segment in geographic space, representing the physical scale at which that curb segment is usable in urban roads. The total length of dry curbs refers to the cumulative length of all curb segments determined to be dry and usable within a grid's feasible domain set at the same time, representing the scale of curb resources in that grid that can still maintain normal usability. The total curb length refers to the sum of the lengths of all legal curb segments within the same grid, representing the theoretical maximum usable curb length of that grid under design conditions. The reachable dry curb ratio is a quantitative indicator calculated by comparing the total dry curb length with the total curb length. It describes the proportion of the actual usable curb length to the theoretical total length in an urban grid under specific meteorological conditions at the current moment. The higher this ratio, the better the surface drainage in the area and the closer the curb usability is to normal levels.

[0031] Specifically, the calculation of the reachable dry curb ratio is based on the principle of continuity and proportional quantification of spatial availability distribution. The usability of urban curbs under rainfall conditions depends on the drainage capacity and catchment load balance of their location. Dry, usable curb segments spatially represent well-drained, passable areas, while all curb segments constitute the complete supply range of that area under design conditions. By summing the lengths of dry curb segments in the grid feasible domain set, the total length maintaining normal usability at a specific time can be obtained. The ratio of this length to the total length of all curb segments in the same area reflects the effective utilization ratio of the curb system at the current time. Since length directly represents the linear reachability range in geometric space, this ratio accurately reflects the proportion of urban curb space occupied by dry conditions, thus quantifying the passability and supply availability of urban roads under rainfall conditions. Therefore, this calculation result is the reachable dry curb ratio.

[0032] S4. Based on the proportion of accessible dry curbs and the spatial area grid in urban road curbs, calculate the feasible supply, and calculate the achievable passenger pick-up intensity based on the set of feasible regions in the grid and the set of passenger requests for all currently pending orders. In embodiments of the present invention, based on the proportion of accessible dry curbs and the spatial region grid in urban road curbs, an executable supply is calculated, including: Supply intensity is calculated for spatial area grids of urban road curbs based on distance weighting and driver online status, thus obtaining the visible supply of the spatial area grids; Specifically, the implementation steps for calculating the supply intensity of urban road curb spatial area grids based on distance weighting and driver online status include the following: First, obtain the location and working status information of online drivers from the platform's real-time location data, and filter out drivers who are currently idle or available for taking orders as effective supply sources; then, call urban road and curb spatial data, divide the urban area into several spatial grid units according to a uniform scale, and calculate the center point coordinates of each grid as the basis for calculating supply intensity; next, the system calculates the spatial distance between each online driver and the center of each grid, and determines the weighting coefficient according to the distance function, with higher weight for closer distances and lower weight for farther distances, thus forming a supply influence weight matrix; subsequently, the weighted supply values ​​from different drivers in each grid are superimposed and summed to obtain the supply intensity of that grid at the current moment; then, combined with driver online status information, the weights of drivers in an executable state are retained, while the weights of drivers in a service or offline state are removed; finally, the comprehensive supply intensity value corresponding to each spatial grid is output, which is the visible supply of the spatial area grid, providing input data for subsequent executable supply correction based on road accessibility ratios.

[0033] The executable supply is obtained by linearly scaling the visible supply and the reachable dry curb ratio.

[0034] Specifically, the implementation steps for linearly scaling the visible supply and the reachable dry curb ratio to obtain the executable supply include the following: First, obtain the visible supply data corresponding to each spatial grid from the previous steps. This data reflects the theoretical supply intensity calculated based on driver online status and distance weighting at the current time. Then, call the real-time updated reachable dry curb ratio data. This ratio represents the proportion of curb length that remains dry and usable for parking or passenger pick-up / drop-off within the grid, representing the road availability of the area. Next, the system performs grid-by-grid scaling of the visible supply value and the reachable dry curb ratio. The system performs a matching process to ensure that the two sets of data are consistent in spatial resolution and timestamps. Then, a linear scaling operation is performed, which multiplies the visible supply value of each grid with the corresponding proportion of accessible dry curbs to proportionally reduce supply losses caused by water accumulation, congestion, or unusable surfaces. All the results are then aggregated into a new grid supply dataset, and this value is recorded as executable supply in the database. Finally, the executable supply result for each spatial grid is output to reflect the effective capacity that the platform can actually deploy under the current weather and surface conditions, providing a physical constraint basis for subsequent supply and demand matching, dispatch strategies, and prediction models.

[0035] Specifically, distance-weighted calculation refers to adjusting the contribution weight of drivers based on their spatial distance from the target grid center during the supply intensity calculation process. Drivers closer to the grid have a greater impact on its supply intensity, while those farther away have a smaller impact, reflecting the spatial decay law of supply accessibility. Driver online status refers to the driver's work status information on the ride-hailing platform, indicating their ability to respond to dispatch orders immediately. Spatial area grid refers to dividing urban road space into several adjacent calculation units according to a uniform scale, used to statistically analyze driver distribution and service capacity within each area. Visible supply refers to the theoretically available travel service scale within a spatial area, calculated based on the number of online drivers and distance-weighted calculations, reflecting the perceptible potential capacity in the current time period. The reachable dry curb ratio is an indicator describing the proportion of urban curbs that can be parked under rainfall conditions; a higher value indicates stronger accessibility. Linear scaling operation refers to the calculation process of proportionally adjusting the visible supply according to the reachable dry curb ratio, ensuring that the supply intensity remains consistent with the actual road availability. The executable supply is a modified result obtained by multiplying the visible supply by the proportion of accessible dry curbs, used to characterize the effective transport capacity that can actually carry out passenger pick-up or service operations on urban roads under current environmental conditions.

[0036] In embodiments of the present invention, the passenger pick-up intensity can be calculated based on the set of feasible grid regions and the set of all passenger requests currently awaiting dispatch, including: Spatial encapsulation and attribute encoding are performed on the set of feasible grid regions to obtain a dry curb layer; Specifically, the steps for spatially encapsulating and attribute-encoding the grid feasible region set to obtain the dry curb layer include the following: First, the system reads the grid feasible region set data obtained through drainage calculation and state determination, which includes the spatial coordinates, length, and grid number of each dry available curb segment; then, in the geographic information system environment, the dry curb segments within the same grid are spatially aggregated to generate a continuous spatial boundary outline, forming the geometric boundary object of the feasible area; next, the generated spatial boundary object is geometrically topologically corrected to ensure that each dry curb segment does not overlap or break in space, and maintains consistency with... Spatial consistency of the urban road base map is achieved. Subsequently, a unique identifier is assigned to each spatial object, and an attribute table is established. The attribute table records information such as curb segment number, length, road segment type, catchment area number, grid number, and timestamp for subsequent data query and dynamic updates. Then, spatial geometric information and attribute information are bound together, and a vectorized data file containing spatial coordinates and related attributes is generated through spatial encapsulation operations. Finally, this vector file is output as a dry curb layer and stored in the geographic information database. This layer can be called and visualized in real time in subsequent passenger pick-up redirection, supply and demand analysis, and scheduling control.

[0037] Get the set of all passenger requests currently pending dispatch; Based on the dry curb layer, each request in the passenger request set is redirected to the boarding point to obtain a successfully matched subset of requests; Specifically, the implementation steps for obtaining the set of all passenger requests awaiting dispatch and redirecting passenger requests to pick-up points based on the dry curb layer include the following: First, extract all passenger request data currently awaiting dispatch from the ride-hailing platform's real-time dispatch system. This data includes passenger identifiers, original pick-up location coordinates, call time, and destination information. Then, perform spatial preprocessing on the extracted data, uniformly transforming the passenger pick-up locations to the same coordinate system as the urban road and curb layer to ensure the accuracy of subsequent spatial matching. Next, the system calls the newly generated dry curb layer, which records... The system first determined the spatial location, length, and grid number of each available dry curb segment. Then, using a spatial nearest neighbor search algorithm, it calculated the shortest Euclidean distance from each passenger's requested location to a dry curb segment and determined candidate pick-up points based on the principle of minimum distance. If the distance was within the feasible walking range threshold, the passenger's pick-up location was redirected to the corresponding dry curb segment, and the matching result was recorded in the system. If the distance exceeded the threshold, the request was determined to be temporarily unfeasible. Finally, all successfully redirected request information was summarized into a successfully matched request subset, and the corresponding dry curb segment number and grid number were recorded.

[0038] The successful matching request subsets are categorized and statistically analyzed by grid to obtain the achievable passenger load strength.

[0039] Specifically, the steps for classifying and statistically analyzing successfully matched request subsets by grid to obtain the achievable passenger pick-up intensity include the following: First, passenger request data that has been successfully redirected to dry curb sections is imported into the statistics module. This data includes the passenger number corresponding to each request, the number of the redirected dry curb section, and the spatial grid number to which the curb section belongs. Then, the system groups all successfully matched requests according to the grid number, grouping all requests within the same grid into the same statistical unit. Next, the number of requests within each grid is counted to obtain the number of passenger pick-up requests that can be achieved in that grid at the current time. Subsequently, the statistical results are time-stamped to track the changing trend in subsequent time series forecasts. Then, the number of requests in each grid is bound to the corresponding spatial coordinates to form a geographically attributed achievable passenger pick-up intensity dataset. Finally, the achievable passenger pick-up intensity value for each grid is output. This value reflects the scale of passenger requests that can be successfully projected to dry and available curb sections in each spatial area under the current rainfall and road conditions, providing accurate demand-side input for subsequent supply and demand balance analysis, forecast calculation, and dynamic dispatching decisions.

[0040] Specifically, spatial encapsulation refers to encoding and storing the geometric boundaries of these feasible areas in a geographic information system (GIS), enabling spatial query and attribute retrieval capabilities. Attribute encoding assigns a unique identifier and related parameter information to each spatial unit for subsequent matching and statistical calculations. The dry curb layer is a GIS layer generated based on the spatial encapsulation results. It records the location, length, and grid number of each dry, usable curb segment, providing a visual representation of the passable area. The passenger request set refers to the collection of boarding requests from all passengers awaiting dispatch in the current system. Each request includes a timestamp, boarding coordinates, and passenger identification information. Boarding point redirection is the process of reprojecting the passenger's original boarding location to the nearest dry, usable curb segment to ensure the safety and accessibility of the boarding location. The successfully matched request subset refers to the set of passenger requests that can find corresponding dry curb segments during the redirection process, indicating that these requests can successfully complete the boarding operation under the current road conditions. Grid-based statistics is the process of aggregating and summing these successfully matched requests according to their spatial grids, reflecting the number of passengers that can be boarded within each grid. The achievable passenger pick-up intensity refers to the number of passenger requests successfully projected onto the dry curb section in each spatial grid at the current moment, indicating the degree to which passenger demand can be met under existing rainfall and road conditions.

[0041] S5. Based on achievable passenger boarding intensity, feasible supply and the proportion of accessible dry curbs, determine the predicted demand gap for the next time window; In embodiments of the present invention, determining the predicted demand gap for the next time window based on achievable passenger boarding intensity, feasible supply, and the proportion of available dry curbs includes: The available capacity is obtained by linearly multiplying the proportion of accessible dry curbs and the baseline capacity of curb segments under normal weather conditions. Specifically, the steps for obtaining available capacity by linearly multiplying the reachable dry curb ratio and the baseline capacity of curb segments under normal weather conditions include the following: First, the system reads the baseline capacity data of each spatial grid under normal weather conditions from the historical operation database. This baseline capacity is calculated from the order completion rate, average dwell time, and total usable curb length during past stable operations, reflecting the maximum service capacity of the area under conditions without meteorological interference. Then, the system obtains the reachable dry curb ratio of each grid from the real-time monitoring module. This ratio is calculated by comprehensively considering rainfall intensity, drainage capacity of storm drains, and water accumulation monitoring results, representing the proportion of curb length in the area that is still dry and usable under the current environmental conditions. Next, the reachable dry curb ratio is matched one-to-one with the baseline capacity of the corresponding grid to ensure that the two sets of data are consistent in terms of timestamp and spatial resolution. Subsequently, a linear product operation is performed on each grid, multiplying the baseline capacity by the reachable dry curb ratio to obtain the actual available capacity after weather and road condition correction.

[0042] Minimize the achievable passenger load intensity, feasible supply, and available capacity to obtain the number of orders that can be completed in the next time window; Specifically, the steps for minimizing the achievable passenger pick-up intensity, executable supply, and available capacity to obtain the number of orders that can be completed in the next time window include the following: First, obtain the achievable passenger pick-up intensity data for each spatial grid at the current time from the system database. This data represents the number of passenger requests that can be successfully redirected and meet the pick-up conditions within the dry, available curb area. Second, retrieve the executable supply data after adjusting for the proportion of available dry curbs. This data reflects the number of drivers who can actually complete order-taking operations under the current road and weather conditions. Then, read the available capacity data obtained through the previous calculation step. This data represents the number of orders that can be completed within the current time window within the grid. The system first determines the maximum passenger pick-up and drop-off capacity that can be supported at any given time. Then, it aligns these three sets of data in time and space to ensure they correspond to the same time and grid. Next, it performs a minimum value calculation, comparing the achievable passenger pick-up intensity, available supply, and available capacity for each grid, and taking the minimum value as the actual number of orders that grid can complete in the next time window. The result is then marked as the number of orders that can be completed and written into the database as a prediction result, reflecting the actual service capacity of each region in the future time period. Finally, the system aggregates the number of orders that can be completed from all grids to generate a global order completion distribution map for the next time window.

[0043] The difference between the achievable passenger load intensity and the achievable order volume in the next time window is calculated to obtain the predicted demand gap.

[0044] Specifically, the reason for calculating the difference between achievable passenger pick-up intensity and the number of orders that can be completed in the next time window to obtain the predicted demand gap is that this difference reflects the dynamic imbalance between passenger achievable demand and system-executable supply within a specific time and space range. Achievable passenger pick-up intensity represents the number of passenger requests that are physically accessible and have service potential under current environmental conditions, while the number of orders that can be completed is the number of orders that the system can actually complete after being limited by executable capacity and available curb capacity. The difference between the two reflects the degree to which some demand cannot be responded to in a timely manner due to insufficient capacity, decreased curb availability, or localized flooding. By calculating this difference, the potential service gap in each grid or region can be quantified, providing the system with a basis for identifying supply and demand imbalances in advance, thereby enabling the prediction of future short-term demand pressure and proactive adjustment of scheduling resources.

[0045] Specifically, the baseline capacity of a curb segment under normal weather conditions refers to the maximum number of vehicles that can simultaneously complete passenger pick-up / drop-off or parking operations per unit time under conditions of no rainfall or external interference. It serves as a reference service capacity of the road under ideal operating conditions. Available capacity is a dynamic capacity value obtained by multiplying the proportion of accessible dry curbs by the baseline capacity. It represents the effective service capacity that urban curbs can actually provide under current weather and drainage conditions. The number of orders that can be completed refers to the number of orders that can actually be completed within a given time window, considering the constraints of achievable passenger pick-up intensity, executable supply, and available capacity. It represents the upper limit of the system's achievable service level. The predicted demand gap is the result of calculating the difference between achievable passenger pick-up intensity and the number of orders that can be completed. It represents the portion of passenger demand that exceeds the system's executable capacity under current operating conditions. It is used to reflect the degree of regional supply-demand imbalance and provides a basis for subsequent scheduling and resource compensation.

[0046] S6. Dynamically adjust and optimize ride-hailing dispatch and order assignment strategies based on the number of orders that can be completed in the next time window and the predicted demand gap.

[0047] In embodiments of the present invention, the ride-hailing dispatch and order assignment strategy is dynamically adjusted and optimized based on the number of orders that can be completed in the next time window and the predicted demand gap, including: The driver scheduling and order dispatch priority are dynamically allocated and optimized based on the number of orders that can be completed; Specifically, the implementation steps for dynamically allocating and optimizing driver scheduling and order dispatch priorities based on the number of orders that can be completed include the following: First, the system reads the number of orders that can be completed from each spatial grid from the prediction module. This data reflects the number of orders that can actually be completed in each region within the current or next time window, used to measure the supply saturation of different regions. Then, the number of orders that can be completed is spatially matched with the real-time monitored driver distribution data to identify regions with surplus supply and regions with potential supply shortages. Next, the system calculates the scheduling priority weight based on the proportion of orders that can be completed in each grid and the differences between adjacent regions. Regions with higher weights indicate greater demand pressure and higher service completion potential. Subsequently, the system allocates priority according to the weight. The system generates driver dispatch instructions and dynamically plans routes for drivers who are idle or whose trips are about to end, guiding them to areas with a higher number of completed orders to increase the probability of order matching. Simultaneously, at the order dispatch level, the system introduces a priority factor into the order allocation algorithm, assigning higher matching weights to orders in high-priority areas to ensure that available driver resources prioritize passenger needs in these areas. Priority parameters are then dynamically adjusted based on real-time feedback data, enabling the dispatch strategy to continuously respond to changes in road feasibility and demand fluctuations. Finally, the optimized dispatch and scheduling results are output to the execution end, forming an adaptive capacity allocation mechanism, thereby achieving dynamic supply and demand balance and maximizing service efficiency across different regions of the platform.

[0048] The grid quota and vehicle guidance strategy are adjusted and supplemented in real time based on the predicted demand gap.

[0049] Specifically, the implementation steps for real-time adjustment and supplementation of grid quotas and vehicle guidance strategies based on predicted demand gaps include the following: First, the system retrieves predicted demand gap data for each spatial grid from the prediction module. This data represents the portion of passenger demand exceeding available supply within a future time window, serving as a core indicator of the degree of supply-demand imbalance. Next, the system normalizes the predicted demand gaps for all grids and sets tiered thresholds based on the gap magnitude, dividing the system into high-demand gap zones, mild-gap zones, and supply-demand balance zones to achieve tiered control. Then, dynamic capacity quotas are prioritized for high-demand gap zones. The system calculates the set of dispatchable vehicles based on the driver's current location, driving status, and remaining mileage, and generates the optimal guidance route using a route planning algorithm, placing idle or soon-to-be-completed vehicles in the high-demand gap zones. The system guides drivers to the target grid to shorten response time. Simultaneously, for areas with slight shortages, the system attracts surrounding vehicles to move to that area by adjusting price incentive parameters or extending dwell time, achieving flexible supply replenishment. Subsequently, the system updates the quota distribution of each grid in real time and generates a dynamic supply-demand heatmap on the monitoring interface, ensuring the dispatch center can intuitively grasp the resource matching status of each area. Furthermore, it dynamically corrects vehicle guidance routes based on traffic congestion information and weather changes to prevent vehicles from concentrating in impassable or restricted areas. Finally, the updated grid quota table and vehicle guidance instructions are simultaneously pushed to the driver's terminal, forming a set of real-time executable dispatch instructions. This enables adaptive capacity allocation and regional supply control based on predicted demand gaps, effectively alleviating the supply-demand imbalance in high-demand areas.

[0050] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A ride-hailing supply and demand forecasting method integrating real-time anomaly detection, characterized in that, Includes the following steps: S1. Determine the total length of the curb based on the set of line segments of the urban road curb; S2. Generate a set of feasible grid regions based on the maximum interception capacity and peak inflow of the adjacent storm drains corresponding to the curb segment; S3. Calculate the proportion of reachable dry curbs based on the total length of dry curbs and the total length of curbs in the grid feasible region set; S4. Based on the proportion of accessible dry curbs and the spatial area grid in urban road curbs, calculate the feasible supply, and calculate the achievable passenger pick-up intensity based on the set of feasible regions in the grid and the set of passenger requests for all currently pending orders. S5. Based on achievable passenger boarding intensity, feasible supply and the proportion of accessible dry curbs, determine the predicted demand gap for the next time window; S6. Dynamically adjust and optimize ride-hailing dispatch and order assignment strategies based on the number of orders that can be completed in the next time window and the predicted demand gap.

2. The ride-hailing supply and demand forecasting method integrating real-time anomaly detection according to claim 1, characterized in that, Based on the set of line segments of the urban road curb, determine the total length of the curb, including: The urban road curb is divided according to the road shape to obtain a set of line segments, which includes the number and length of the curb line segments. Associate the effective catchment area of ​​the curb segment with that of the adjacent storm drain inlet of the curb segment; The total curb length is obtained by summing the lengths of the line segments within the same grid.

3. The ride-hailing supply and demand forecasting method integrating real-time anomaly detection according to claim 2, characterized in that, Based on the maximum interception capacity and peak inflow of adjacent storm drains corresponding to the curb segment, a set of feasible grid regions is generated, including: The inflow rate per unit time is obtained by multiplying the rainfall intensity provided by the external meteorological system with the effective catchment area corresponding to the curb segment. The peak inflow rate is obtained by multiplying the inflow rate per unit time with the preset runoff coefficient. The difference between the maximum interception capacity and the peak inflow of the adjacent storm drain corresponding to the curb segment is calculated to obtain the capacity-water requirement difference. The curb segments are marked based on the relationship between the capacity-water-demand difference and 0, and the status marking results of the curb segments are obtained. The curb segments whose status marking results are dry and usable are filtered and grouped by grid to obtain the grid feasible region set.

4. The ride-hailing supply and demand forecasting method integrating real-time anomaly detection according to claim 1, characterized in that, Based on the total length of dry curbs and the total curb length of the grid feasible region set, the proportion of reachable dry curbs is calculated, including: The total length of dry curb is obtained by summing the lengths of all curb segments in the feasible region set of the grid. The ratio of the total length of dry curb to the total length of curb is calculated to obtain the proportion of achievable dry curb.

5. The ride-hailing supply and demand forecasting method integrating real-time anomaly detection according to claim 1, characterized in that, Based on the proportion of accessible dry curbs and the spatial area grid in urban road curbs, the feasible supply is calculated, including: Supply intensity is calculated for spatial area grids of urban road curbs based on distance weighting and driver online status, thus obtaining the visible supply of the spatial area grids; The executable supply is obtained by linearly scaling the visible supply and the reachable dry curb ratio.

6. The ride-hailing supply and demand forecasting method integrating real-time anomaly detection according to claim 1, characterized in that, The boarding intensity can be calculated based on the set of feasible regions in the grid and the set of passenger requests for all pending orders, including: Spatial encapsulation and attribute encoding are performed on the set of feasible grid regions to obtain a dry curb layer; Get the set of all passenger requests currently pending dispatch; Based on the dry curb layer, each request in the passenger request set is redirected to the boarding point to obtain a successfully matched subset of requests; The successful matching request subsets are categorized and statistically analyzed by grid to obtain the achievable passenger load strength.

7. The ride-hailing supply and demand forecasting method integrating real-time anomaly detection according to claim 1, characterized in that, Based on achievable passenger pick-up intensity, feasible supply, and the proportion of accessible dry curbs, the projected demand gap for the next time window is determined, including: The available capacity is obtained by linearly multiplying the proportion of accessible dry curbs and the baseline capacity of curb segments under normal weather conditions. Minimize the achievable passenger load intensity, feasible supply, and available capacity to obtain the number of orders that can be completed in the next time window; The difference between the achievable passenger load intensity and the achievable order volume in the next time window is calculated to obtain the predicted demand gap.

8. The ride-hailing supply and demand forecasting method integrating real-time anomaly detection according to claim 1, characterized in that, The ride-hailing dispatch and order assignment strategies are dynamically adjusted and optimized based on the number of orders that can be completed in the next time window and the predicted demand gap, including: The driver scheduling and order dispatch priority are dynamically allocated and optimized based on the number of orders that can be completed; The grid quota and vehicle guidance strategy are adjusted and supplemented in real time based on the predicted demand gap.

Images