Urban area supply and demand imbalance short-term prediction method and system for online car-hailing platform

Abstract

The invention discloses an urban area supply and demand imbalance short-time prediction method and system for an online car-hailing platform, and relates to the technical field of supply and demand prediction. Comprising the following steps: acquiring intersection steering limitation, terrain gradient, residual endurance mileage of a transport capacity carrier, accumulated online duration of the transport capacity carrier, spatial distribution data of the transport capacity carrier, real-time order demand quantity and regional transaction deviation of a current period; generating a path resistance variable based on the intersection steering limitation and the terrain gradient, and constructing an efficiency attenuation reference table by using the residual endurance mileage of the transport capacity carrier and the accumulated online duration of the transport capacity carrier; the method has the beneficial effects that the service capability collapse of the transport capacity carrier caused by physical power consumption increment in a complex geographical environment can be accurately anchored, so that the predicted unbalance strength index is closer to the real physical supply and demand level.

Description

Technical Field

[0001] This invention relates to the field of supply and demand forecasting technology, and in particular to a method and system for short-term forecasting of supply and demand imbalances in urban areas for ride-hailing platforms. Background Technology

[0002] With the deep integration of mobile internet and intelligent transportation technologies, ride-hailing platforms have become an important mode of transportation for urban residents. To improve the platform's dispatch efficiency and service quality, accurately predicting the supply and demand balance within specific urban areas over short periods has become a core task of ride-hailing dispatch systems. Traditional forecasting methods typically rely on historical order data, real-time location information, and macro-level characteristics such as weather and time of day. By establishing statistical or machine learning models, they assess the supply and demand of transportation capacity in future periods, thereby calculating the supply-demand gap. This forecasting mechanism aims to identify potential areas of capacity depletion or surplus in advance, providing data support for the platform's dynamic pricing, capacity scheduling, and incentive policies. This is of great significance for alleviating urban traffic congestion and optimizing resource allocation.

[0003] Existing supply and demand forecasting technologies often focus on static quantitative statistics of transport vehicles or macroscopic estimations based on ideal efficiency when assessing transport capacity. This approach primarily focuses on the spatial distribution and quantity of transport vehicles, while neglecting the complexity of the urban geographical topology and the coupled impact of the dynamic physical attributes of the transport vehicles themselves on actual service capacity. In actual operating environments, due to the lack of in-depth consideration of physical obstacles in the road network and the real-time energy efficiency degradation of the vehicles, the system struggles to quantify the actual losses of transport capacity during task execution. This results in a significant discrepancy between the calculated supply and the actual available capacity efficiency. This "nominal" modeling of the transport supply side makes it impossible for the forecast results to accurately anchor the effective supply level under complex physical constraints, limiting the platform's resource scheduling accuracy and system robustness in variable urban environments. Summary of the Invention

[0004] In view of the above-mentioned prior art, this application is hereby made. Embodiments of this application provide a method and system for short-term prediction of supply and demand imbalance in urban areas for ride-hailing platforms. This method can accurately pinpoint the service capacity collapse caused by increased physical power consumption of transportation vehicles in complex geographical environments, making the predicted imbalance intensity index closer to the actual physical supply and demand levels.

[0005] According to one aspect of this application, a method for short-term prediction of supply and demand imbalance in urban areas for ride-hailing platforms is provided, including: Obtain the current period's intersection turning restrictions, terrain slope, remaining driving range of the transport vehicles, cumulative online time of the transport vehicles, spatial distribution data of the transport vehicles, real-time order demand, and regional transaction deviation; Based on the intersection turning restrictions and the terrain slope, path resistance variables are generated, and an efficiency decay benchmark table is constructed using the remaining driving range of the transport vehicle and the cumulative online time of the transport vehicle. The path resistance variable is injected into the efficiency decay benchmark table, the physical power consumption increment of the transport vehicle is calculated, and the spatial distribution data of the transport vehicle is converted into an effective supply energy value with physical energy loss attributes. The difference between the effective supply capacity and the real-time order demand is calculated to generate an imbalance prediction intensity index that indicates the intensity of the supply and demand contradiction in the current period. Based on the fluctuation frequency of the imbalance prediction intensity index, the data collection granularity for the remaining driving range and cumulative online time of the transport vehicle in the next cycle is adjusted in reverse; and The influence operator weights in the effective supply energy value conversion process are corrected by using the regional transaction deviation. The corrected influence operator weights are then used as input parameters for the next cycle to perform cyclical compensation for the effective supply energy value.

[0006] According to another aspect of this application, a short-term prediction system for urban area supply and demand imbalances for ride-hailing platforms is provided, comprising: Multi-dimensional data acquisition module: used to acquire the intersection turning restrictions, terrain slope, remaining driving range of the transport vehicle, cumulative online time of the transport vehicle, spatial distribution data of the transport vehicle, real-time order demand, and regional transaction deviation for the current period; Energy efficiency benchmark construction module: used to generate path resistance variables based on the intersection turning restrictions and the terrain slope, and to construct an efficiency decay benchmark table using the remaining driving range of the transport vehicle and the cumulative online time of the transport vehicle; Effective supply conversion module: used to inject the path resistance variable into the efficiency decay benchmark table, calculate the physical power consumption increment of the transport vehicle, and convert the spatial distribution data of the transport vehicle into an effective supply energy value with physical energy loss attributes; Imbalance Intensity Prediction Module: Used to calculate the difference between the effective supply energy value and the real-time order demand, and generate an imbalance prediction intensity index to indicate the intensity of the supply and demand contradiction in the current period. Feedback correction module: used to adjust the data collection granularity of the remaining driving range and cumulative online time of the transport vehicle in the next cycle in reverse, based on the fluctuation frequency of the imbalance prediction intensity index; and The influence operator weights in the effective supply energy value conversion process are corrected by using the regional transaction deviation. The corrected influence operator weights are then used as input parameters for the next cycle to perform cyclical compensation for the effective supply energy value.

[0007] According to another aspect of this application, an electronic device is provided, including a memory and a processor, the memory being used to store computer-executable instructions, and the processor being used to execute the computer-executable instructions, which, when executed by the processor, implement the steps of the method described above.

[0008] According to another aspect of this application, a computer storage medium is provided that stores computer-executable instructions thereon, which, when executed by a processor, implement the steps of the method described above.

[0009] Compared with the prior art, the method and system for short-term prediction of urban area supply and demand imbalance for ride-hailing platforms according to the embodiments of this application can transform traditional static spatial distribution data into effective supply energy values ​​with physical energy efficiency attributes. This solves the drawbacks of the "nominal" modeling of the transportation capacity in the prior art, and can accurately anchor the service capacity collapse caused by the increase in physical power consumption of transportation carriers in complex geographical environments, making the predicted imbalance intensity index closer to the real physical supply and demand level. During periods or regions of drastic supply and demand fluctuations, the system can automatically improve the accuracy of key parameter acquisition, while reducing the sensing load during stable periods. This adaptive feedback control not only ensures the robustness of predictions under extreme conditions, but also effectively filters out environmental noise and prediction drift, ensuring that the system can automatically adjust with changes in regional characteristics and maintain a high level of prediction consistency and resource scheduling accuracy over the long term, thereby effectively solving the problem of prediction failure in volatile urban environments. Attached Figure Description

[0010] The above and other objects, features, and advantages of this application will become more apparent from the more detailed description of the embodiments of this application in conjunction with the accompanying drawings. The drawings are provided to further illustrate the embodiments of this application and form part of the specification. They are used together with the embodiments of this application to explain this application and do not constitute a limitation thereof. In the drawings, the same reference numerals generally represent the same components or steps.

[0011] Figure 1 This is a schematic diagram of the overall process of the short-term prediction method for supply and demand imbalance in urban areas for ride-hailing platforms according to the present invention.

[0012] Figure 2 This is a schematic diagram of the logical framework of the short-term prediction method for urban area supply and demand imbalance in ride-hailing platforms according to the present invention.

[0013] Figure 3 This is an extended schematic diagram of the method for short-term prediction of supply and demand imbalance in urban areas for ride-hailing platforms, as described in this invention. Detailed Implementation

[0014] Hereinafter, exemplary embodiments according to this application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this application, and not all embodiments of this application. It should be understood that this application is not limited to the exemplary embodiments described herein.

[0015] Example 1: In existing technologies, the supply and demand forecasting technology for ride-hailing platforms has evolved from simple historical average statistics to spatiotemporal feature modeling based on deep learning. However, accurately anchoring the true supply level in the complex and ever-changing urban physical environment still faces significant challenges. Traditional assessment logic often focuses only on the spatial distribution of vehicles or macroscopic estimates based on ideal efficiency, ignoring the physical constraints of the urban geographical topology and the actual loss of service capacity due to the dynamic physical attributes of the transport vehicles themselves. This results in the calculated supply being only a "nominal supply," which cannot reflect the true effective carrying capacity of vehicles under complex operating conditions, thus limiting the prediction accuracy and system robustness of the dispatching system in extreme or complex scenarios.

[0016] To address the aforementioned issues, this invention proposes a technical solution based on physical energy efficiency modeling and adaptive sensing adjustment. Analysis reveals that the core of inaccurate capacity assessment lies in the coupled reduction of energy efficiency caused by environmental impedance and vehicle status. Therefore, a technical approach combining path resistance quantification, efficiency decay modeling, and closed-loop feedback correction is established. By dynamically injecting path resistance variables and efficiency decay benchmark tables, vehicle spatial distribution data is transformed into an "effective supply energy value" with physical energy consumption attributes. Furthermore, the data collection granularity is adjusted inversely based on the fluctuation frequency of supply-demand imbalance prediction intensity, while simultaneously utilizing regional transaction deviations to achieve cyclical compensation of the influence operator weights. This solution achieves supply-side modeling from "nominal" to "physical," effectively eliminating assessment biases caused by the physical environment and significantly improving the accuracy and predictive stability of resource scheduling in variable urban environments.

[0017] Reference Figures 1-3 As an embodiment of the present invention, a short-term prediction method for supply and demand imbalance in urban areas for ride-hailing platforms is provided: Figure 1 The illustration shows a short-term prediction method for supply and demand imbalance in urban areas for ride-hailing platforms according to an embodiment of this application, including: S1-S5.

[0018] S1: Obtain the intersection turning restrictions, terrain slope, remaining driving range of the transport vehicles, cumulative online time of the transport vehicles, spatial distribution data of the transport vehicles, real-time order demand, and regional transaction deviation for the current period; Specifically, by retrieving pre-stored high-precision urban map data, the system obtains the intersection turning restrictions and terrain slope within the current prediction area. The intersection turning restrictions characterize the connectivity cost of nodes in the road network topology, and the terrain slope records the physical undulations of road segments in vertical space. These environmental features serve as static boundary conditions for subsequent path resistance calculations. Through asynchronous data interaction with the on-board diagnostic (OBD) system and battery management system (BMS) of the transport vehicles, the system obtains the remaining driving range of each transport vehicle. Simultaneously, by combining the platform's check-in records for each vehicle within the billing cycle, the system calculates the cumulative online time of each transport vehicle. Furthermore, the system uses a satellite positioning system (GPS / BeiDou) to track and report the spatial distribution data of each transport vehicle in real time to determine the instantaneous physical coordinates of the vehicles in the road network. The platform gathers real-time order demand in the current forecast area. This data is aggregated from all instant car-hailing requests initiated by users. At the same time, the system retrieves the operating logs from the previous statistical period to obtain the regional transaction deviation. This regional transaction deviation serves as a key feedback indicator for measuring the accuracy of historical forecasts and the consistency between actual transactions. Furthermore, for extreme cases where the previous statistical period is unavailable, such as when the system is first launched, for newly opened prediction areas, or when historical operating logs are missing, regional transaction deviation can be handled in the following ways: If no historical data is available, the regional transaction deviation is initialized to a preset default value (e.g., 0, indicating no deviation baseline), or a proxy value is assigned based on the historical average transaction value of similar areas. Specifically, the system can query the platform's global database for reference areas that match the geographical features (such as road network density and terrain complexity) of the current area, and extract their average transaction deviation as the initial proxy value; this proxy value is iteratively updated in subsequent periods based on actual transaction feedback. After acquiring the aforementioned multidimensional raw data, the system performs necessary normalization preprocessing on the data. For example, it converts the terrain slope into the corresponding slope angle in radians and the remaining driving range into a unified energy scalar, thereby providing standardized data input for the subsequent construction of the performance degradation benchmark table.

[0019] S2: Generate path resistance variables based on intersection turning restrictions and terrain slope; Specifically, by deconstructing the physical topology loss and gravitational potential energy compensation of the road network, the additional power consumption that a transport vehicle needs to overcome to travel a unit distance within a specific area is quantified, and the path resistance variable is calculated using the following formula: ; in, Here, is the path resistance variable, representing the expected additional resistance generated per unit distance traveled by the transport vehicle within the current prediction area, and N is the number of intersection turning restrictions within the current prediction area. Let L be the equivalent energy loss per unit turn at the i-th turn-restricted intersection, whose value is preset based on the turn angle (left turn, U-turn, etc.), and L be the total length of the road network in the current prediction area. This represents the average mass density (constant) of the transport vehicle within the current region. It is the acceleration due to gravity. This represents the average sine value of the terrain slope in the direction of travel; A performance degradation benchmark table is constructed using the remaining driving range and the cumulative online time of the transport vehicle. Specifically, based on the remaining driving range, the nominal service range of the transport vehicle under ideal operating conditions is determined. Then, the cumulative online time is introduced as a fatigue correction factor. In practice, as the cumulative online time increases, the order-taking efficiency score of the transport vehicle is gradually reduced through a preset decay curve (e.g., reflecting the driver's physiological fatigue or the thermal decay of vehicle components). Finally, with the remaining driving range as the horizontal dimension and the cumulative online time as the vertical dimension, a matrix containing multiple efficiency evaluation nodes is generated through cross-indexing, namely the efficiency decay benchmark table. This not only quantifies the "resistance" of the external environment to vehicle driving, but also quantifies the "internal friction" of the vehicle's own state on service capabilities, thus logically completing the initial coupling between the physical environment and the vehicle attributes.

[0020] S3: Inject the path resistance variable into the efficiency degradation benchmark table and calculate the physical power consumption increment of the transport vehicle. This step quantifies the additional energy consumption of the transport vehicle under specific physical constraints compared to the ideal operating state. The specific formula is as follows: ; in, This represents the increase in physical power consumption of the transport vehicle, reflecting the additional power loss per unit time caused by path resistance within the current forecast period. This represents the total length of the road network within the current prediction area. The cumulative online time of the transportation capacity. The real-time performance factor is obtained by cross-indexing the remaining driving range and cumulative online time in the performance degradation benchmark table. This characterizes the current energy conversion efficiency of the transport vehicle. Transform the spatial distribution data of transportation capacity into an effective supply energy value with physical energy consumption and loss attributes; Specifically, the conversion of effective energy supply includes: Based on the number of turning restrictions at intersections and the frequency of terrain slope fluctuations, the expected power consumption of the transport vehicle per unit distance is calculated, quantifying the average expected power required for the transport vehicle to travel a unit distance within the current area. The specific formula is as follows: ; in, For expected power consumption variables, Here, N is the index number of the intersection turning restriction, and N is the number of intersection turning restrictions in the current prediction area. Let i be the average energy loss per unit turning action at the i-th turning-restricted intersection. This represents the total length of the road network within the current prediction area. The frequency of the slope fluctuations (dimensionless). To overcome the equivalent resistance required to overcome unit slope fluctuation; The remaining service range of the vehicle is obtained by applying physical loss correction to the remaining range of the vehicle using the expected power consumption variable. The remaining service range of the transport vehicle includes: Based on the expected power consumption variable, match the energy consumption sensitivity coefficient corresponding to the power source type of the transportation vehicle. The energy consumption sensitivity coefficient is coupled with the expected power consumption variable in a timely manner to determine the real-time energy consumption redundancy of the transport vehicle in the current road segment. The specific formula is as follows: ; in, This is the real-time energy redundancy. Energy sensitivity coefficient For expected power consumption variables, This is a preset safety redundancy distance; The remaining service range is obtained by dynamically thresholding the remaining driving range using real-time energy consumption redundancy, as shown in the following formula: ; in, For the remaining service range, For the remaining driving range, This represents the average driving resistance of the transport vehicle under ideal operating conditions. Based on the cumulative online time of the transport vehicle, the order-accepting efficiency weight of the transport vehicle in the current period is determined; using the order-accepting efficiency weight, the remaining service range located at different coordinate positions is spatiotemporally aggregated to generate an effective supply energy value, as shown in the following formula: ; in, To effectively supply energy, For each individual transport vehicle, an individual index number is assigned. This represents the total number of transport capacity vehicles in the current region. Let j be the order-receiving efficiency weight. For the remaining service range of the j-th transport vehicle, This is the spatial position attenuation factor corresponding to the j-th transport vehicle; In the embodiments of this application, the energy consumption sensitivity coefficient The system can obtain the energy efficiency characteristics of a pre-established power source through matching. Specifically, based on the power source type of the transport vehicle (such as pure electric, plug-in hybrid, or fuel-powered) and its battery / engine rated power curve, the system uses experimental test data or factory energy consumption parameters to calibrate the response of different power sources to changes in path resistance, thereby finding the energy consumption sensitivity coefficient most suitable for the current vehicle's hardware characteristics; a preset safety redundancy distance is also included. The efficiency weight of the j-th order can be obtained by analyzing abnormal energy consumption samples in historical driving trajectories and the average impact length of unforeseen factors in the road network (such as temporary construction, traffic control, etc.). The spatial location attenuation factor for the j-th capacity carrier can be calculated using historical service response data of the carrier within the current billing cycle, combined with its cumulative online duration, through a nonlinear attenuation function. The distance between the real-time coordinates of the carrier and the core hotspot (high-frequency order generation area) can be calculated and mapped using a Gaussian kernel function or inverse distance weighting (IDW).

[0021] By injecting path resistance variables into the performance degradation benchmark table and calculating the physical power consumption increment of the transport vehicle, the actual energy loss of the transport vehicle during task execution can be quantified. This allows for precise targeting of the service capacity collapse caused by the physical power consumption increment in complex geographical environments, making the generated effective supply energy value closer to the actual physical supply and demand level. This solves the technical problem of discrepancies between the supply and actual release capacity efficiency caused by "nominal" modeling. When dealing with complex urban conditions, compared to macro-estimates based solely on historical order data or ideal efficiency, this application can respond in real time to intersection turning restrictions, terrain slope, and the vehicle's remaining range and cumulative online time. This physical attribute-based conversion method ensures that the system has higher resource scheduling accuracy and robustness in variable urban environments. For example, assume the total length of the road network in the current prediction area is... (m), including the number of intersection turning restrictions. ; First, calculate the expected power consumption variable. If the average energy loss per unit turning action at the i-th turning-restricted intersection is... (J), Frequency of slope undulation The equivalent resistance required to overcome unit slope fluctuation (N); ; Secondly, determine the real-time energy redundancy. For a specific type of power source in a transportation vehicle, the energy consumption sensitivity coefficient obtained by matching is... Preset safety redundancy distance (m); ; Next, calculate the remaining service range. If the remaining range (m), the average running resistance of the transport vehicle under ideal operating conditions. (N); ; The calculation results show that after dynamically removing the remaining range by adjusting the real-time energy consumption redundancy, the remaining service range is corrected from the nominal 40,000m to 37,998m. Finally, the effective energy supply value is generated. If the j-th order-receiving efficiency weight corresponding to the transportation capacity carrier The spatial location attenuation factor corresponding to the j-th transport vehicle Then the effective energy contribution of this individual is: ; The system generates effective supply energy value by aggregating the spatiotemporal efficiency of the energy values ​​of various transportation carriers within the region.

[0022] S4: Calculate the difference between effective supply capacity and real-time order demand to generate an imbalance prediction intensity index that indicates the strength of the supply-demand imbalance in the current period. The specific formula is as follows: ; in, As an index for predicting the intensity of imbalance, Individual index number for real-time order demand. This represents the total real-time order demand within the current forecast region. The expected service mileage corresponds to the k-th order demand, reflecting the theoretical driving distance required to complete the order.

[0023] S5: Based on the fluctuation frequency of the imbalance prediction intensity index, adjust the data collection granularity of the remaining driving range and cumulative online time of the transport vehicle in the next cycle in reverse. Specifically, the reverse adjustment of data acquisition granularity includes: By utilizing the time-domain distribution of the imbalance prediction intensity index within the current period, the trend term is stripped using the least squares method or the moving average method to obtain a high-frequency residual sequence that reflects instantaneous fluctuations. Among them, determining the transient disturbance characteristic values ​​includes: The trend term of the imbalance prediction intensity index is stripped in the time domain to obtain the high-frequency residual sequence of the imbalance prediction intensity index. The energy distribution entropy of the high-frequency residual sequence in the current period is calculated using the following formula: ; ; in, Energy distribution entropy (dimensionless) reflects the degree of energy dispersion in the supply-demand imbalance fluctuations within the current cycle. This serves as the sampling time index for the high-frequency residual sequence within the current period. This represents the total number of sampling points within the current period. The energy proportion factor at time t represents the proportion of instantaneous fluctuation energy to the total fluctuation energy of the current period. Let be the instantaneous fluctuation energy value at time t, used to characterize the squared disturbance intensity of the imbalance prediction intensity index deviating from the trend term at a specific time. For the auxiliary index variable of the cumulative summation, Let be the instantaneous fluctuation energy value at the r-th sampling point; The intensity of fluctuations in the imbalance prediction intensity index is quantified using energy distribution entropy to generate transient disturbance characteristic values. The specific formula is as follows: ; in, is the base of the natural logarithm. The preset steady-state entropy threshold; The transient perturbation eigenvalues ​​are compared with the preset sampling confidence interval. The system performs mapping and matching to determine the scheduling weights of the sensing resources in the next cycle. Specifically, the system uses piecewise linear interpolation to determine the scheduling weights of the sensing resources in the next cycle, as shown below: ; in, The weighting of sensing resources for the next cycle reflects the multiplication adjustment ratio of the baseline acquisition frequency. The preset maximum sampling gain coefficient ( ); The remaining driving range and cumulative online time of the transportation vehicle are adjusted inversely using the perceived resource scheduling weights. The specific formula is as follows: ; in, This is the actual sampling frequency for the next cycle. The baseline sampling frequency reflects the standard data sampling frequency preset by the system for the remaining driving range and cumulative online time of the transport vehicle when the imbalance prediction intensity index is in a stable state (i.e., the transient disturbance characteristic value has not triggered the upper and lower limit trigger thresholds of the sampling confidence interval).

[0024] In this embodiment of the application, the preset steady-state entropy threshold The mean energy distribution entropy of the high-frequency residual sequence can be calculated by retrieving the imbalance prediction intensity index of the region during historical off-peak periods or periods of relative supply-demand balance; a preset sampling confidence interval is also available. The baseline acquisition frequency can be determined by utilizing the probability density distribution of historical transient perturbation characteristic values ​​and selecting a specific confidence level interval (such as a 95% confidence interval) that conforms to the characteristics of a normal distribution or a long-tailed distribution. The maximum sampling gain coefficient can be set based on the hardware power consumption limitations of the vehicle's onboard terminal and the average carrying capacity of mobile communication bandwidth, combined with the minimum accuracy requirements for supply and demand imbalance prediction over a long historical period; This can be obtained by analyzing the hardware power consumption tolerance limit of the vehicle-mounted terminal under the highest load conditions and the peak throughput of the communication link.

[0025] In the actual operation of ride-hailing platforms, the supply-demand imbalance is not a constant static distribution, but rather involves a large number of instantaneous and sudden disturbances (such as sudden weather changes or order surges caused by large gatherings). If a fixed data collection frequency is used, it will lead to a waste of terminal energy consumption and bandwidth resources during stable periods, while during periods of violent fluctuations, it will lead to the omission of key status information, thus affecting the accuracy of prediction. This application aims to quantify the dispersion of such fluctuations in real time by stripping the trend term from the imbalance prediction intensity index and introducing energy distribution entropy, thereby realizing the on-demand allocation of sensing resources. That is, by capturing the transient disturbances of supply and demand contradictions, the collection resources are dynamically allocated, minimizing the unnecessary power consumption of the transportation carrier terminal while ensuring the sensitivity of imbalance prediction. Compared to common fixed-step sampling techniques or simple threshold-triggered sampling schemes in the field, this application utilizes information entropy theory for refined measurement. Fixed-step schemes lack flexibility and cannot adapt to complex urban environments; while simple threshold-triggered schemes are prone to frequent oscillations in sampling frequency due to noise interference. This application adopts piecewise linear interpolation mapping based on energy distribution entropy, which can smoothly and non-linearly generate sensing resource scheduling weights, ensuring the stability and accuracy of acquisition frequency adjustment. It can transform abstract predictive index fluctuations into quantifiable frequency adjustment commands, realizing closed-loop coupling between the prediction system and the underlying sensing layer. For example, suppose the system sets the total number of sampling points within the current prediction period. After stripping away the trend term, the instantaneous fluctuation energy value at time t is obtained. The values ​​are: 2.0, 1.5, 4.0, 1.0, and 1.5. First, calculate the total energy. Energy percentage factor at time t They are respectively: We can obtain: ; ; ; ; ; Summing yields the energy distribution entropy. ; Assuming a preset steady-state entropy threshold ,but ; Assuming a preset lower limit for the sampling confidence interval upper limit Preset maximum sampling gain coefficient ,because Perform piecewise linear interpolation calculations: ; If the reference acquisition frequency The actual sampling frequency in the next cycle The calculation results show that, since the supply and demand imbalance fluctuation in the current cycle is slightly higher than the stability threshold, the system automatically increases the data collection frequency of the remaining driving range and cumulative online time of the transport vehicle in the next cycle from 0.1Hz to about 0.1417Hz (that is, the sampling interval is shortened to about 7 seconds) by reverse adjustment, thereby enhancing the perception granularity of subsequent fluctuations.

[0026] In step S5, the weights of the impact operators in the effective supply energy conversion process are corrected using the regional transaction deviation. The specific formula is as follows: ; in, The corrected influence operator weights, The weights of the impact operators for the current period. To adjust the gain coefficient for feedback, Due to regional transaction deviations, The effective supply energy value of the previous cycle; The corrected influence operator weights are used as input parameters for the next cycle to perform cyclic compensation on the effective supply energy value. Specifically, the system utilizes The sensitivity coefficient is adjusted on the incremental physical power consumption of the transport vehicle calculated in subsequent cycles, thereby achieving cyclic compensation of the effective supply energy value during the dynamic threshold stripping process. It should be noted that adjusting the sensitivity coefficient means that in the calculation process of the next cycle, the weight of the influencing operator is applied as a proportional operator to the expected power consumption variable in order to adjust the contribution weight of environmental resistance to the final power consumption increment. Specifically, when the regional transaction deviation indicates that the historical forecast supply is too high, by increasing the weight of the influencing operator, the expected power consumption variable calculated under the same path resistance is made more sensitive to the energy loss feedback generated by the transport carrier, thereby achieving a refined correction of the supply-side energy value. The cyclical compensation of effective supply energy value in the dynamic threshold stripping process refers to the system using the real-time energy consumption redundancy adjusted by the influence operator weights to deduct the original remaining range of the capacity vehicle when calculating the remaining service range. The threshold of this stripping process is not a fixed constant, but a physical limit that dynamically evolves with the iteration of the influence operator weights. By stripping the non-service energy consumption part that is lost due to environmental resistance from the remaining range, the final effective supply energy value can be adaptively reduced or compensated based on historical transaction feedback, ensuring that the prediction results are highly aligned with the actual carrying capacity.

[0027] In this embodiment of the application, the feedback adjustment gain coefficient This can be obtained by analyzing the correlation between the convergence speed of the influence operator weights and the fluctuation of the system prediction error within the historical period. Specifically, the system can use an adaptive control algorithm to dynamically adjust the value of the influence operator weights based on the prediction residuals of the previous statistical period. This ensures that when correcting the influence operator weights, it can quickly compensate for the evaluation deviation caused by physical environment losses and avoid system oscillations caused by over-adjustment. As a result, the corrected influence operator weights can be smoothly used as input parameters for the next period to compensate for the effective supply energy value. This embodiment does not impose specific limitations on this.

[0028] In the actual operation of ride-hailing platforms, due to the complexity of urban road network environments (such as sudden traffic control) and the performance degradation of the vehicle's own power system over time, the initially set physical conversion parameters are often difficult to maintain absolute accuracy in the long term. If only static physical formulas are relied upon, the prediction system will not be able to perceive the cumulative errors caused by model simplification or sudden environmental changes, resulting in the predicted imbalance intensity index gradually deviating from the actual business results. This application aims to establish a parameter correction chain with self-evolutionary capabilities by retrieving known regional transaction deviations as feedback signals and correcting the weights of influencing operators in real time. This achieves a business closed loop at the algorithm level. By feeding back historical transaction results to the underlying physical energy consumption calculation process, the system can dynamically align physical loss prediction with actual transaction feedback, greatly enhancing the system's adaptability and prediction stability under different urban conditions. Compared to the residual correction techniques based on moving averages or pure data-driven neural network fitting schemes commonly used in the field, this application retains the interpretability of physical modeling. Moving average techniques often have significant lag effects and cannot be precisely adjusted for specific physical influencing factors (such as path resistance). While pure data-driven schemes have high fitting accuracy, they are prone to black-box failures when data is sparse or distributions change abruptly. This application chooses to use the influence operator weight as a proportional operator applied to the expected power consumption variable. This not only enables rapid convergence of errors through feedback adjustment of the gain coefficient, but also clearly reflects the true proportion of environmental resistance to service capacity loss through the dynamic threshold stripping process. It not only provides a dimensionally consistent physical correction path, but also ensures that in subsequent cycles, the system can specifically adjust the sensitivity of environmental resistance to energy efficiency loss, thereby fundamentally eliminating the technical defects of overestimating or underestimating the supply energy value caused by environmental modeling bias.

[0029] like Figure 3 As shown, this application further proposes that, before generating the imbalance prediction intensity index, the method also includes: Obtain historical supply and demand distribution data for the current cycle; Specifically, the system retrieves historical supply and demand distribution data corresponding to the current statistical period (such as Monday morning rush hour) from the cloud database. The historical supply and demand distribution data includes the average order density and average effective capacity distribution in the same area during the same period in history. Based on historical supply and demand distribution data and real-time order demand, a dynamic weighting factor is generated to correct the effective supply capacity. The specific formula is as follows: ; in, As a dynamic weighting factor, For the index number of the historical samples, This represents the total number of historical samples corresponding to the current period. This represents the sum of the historical average service mileage within the region of the m-th historical sample. This represents the real-time order demand (dimensionless) for the current period. Average service mileage per order; The weighted effective supply energy value is obtained based on the dynamic weighting factor, and the specific formula is as follows: ; in, The weighted effective supply energy value, To effectively supply energy; A revised imbalance forecast intensity index is generated based on the weighted effective supply energy and real-time order demand, using the following formula: ; in, This is the revised imbalance prediction intensity index. This represents the total real-time order demand within the current forecast region. This represents the expected service mileage corresponding to the demand for the kth order.

[0030] In the context of ride-hailing services, while real-time physical energy consumption modeling can capture the individual supply capacity of vehicles, it often overlooks the macro-background of regional supply and demand structure during specific time periods (such as morning and evening rush hours). This application introduces historical supply and demand distribution data to compare the current real-time order demand with the historical context of the same period. The aim is to generate a proportional operator that reflects the degree to which the current demand intensity deviates from the historical norm, i.e., a dynamic weighting factor. This injects business prior knowledge into the purely physical supply assessment, so that the corrected imbalance prediction intensity index not only includes the physical constraints of individual vehicles, but also reflects the overall business load pressure of the region. This effectively corrects the problem of diminishing marginal utility of transport capacity caused by sudden surges in traffic, and greatly improves the robustness of the prediction model under special time nodes. Compared to trend forecasting models based on time series analysis or simple supply-demand ratio calculation schemes commonly used in the field, this application has stronger physical meaning and real-time response capabilities. Traditional trend forecasting schemes (such as the ARIMA model) often lag in capturing nonlinear fluctuations and cannot be directly correlated with the remaining service range of the vehicles at the underlying level. Simple supply-demand ratio schemes ignore the energy value matching of orders and capacity in the mileage dimension. This application chooses to use dynamic weighting factors to weight the effective supply energy value, which is essentially performing a nonlinear alignment based on business experience in the energy value dimension. Without changing the underlying physical energy consumption algorithm, it can dynamically adjust the contribution weight of the supply side by calculating the ratio of the sum of historical average service mileage to the current real-time total mileage demand, ensuring that the forecast index maintains a high degree of accuracy in complex urban spatiotemporal environments. For example, assuming we are currently in the Monday morning rush hour forecast period, the system retrieves historical supply and demand distribution data, and the total number of historical samples corresponding to the current period. The sum of historical average service mileage within each historical sample region They are: 12000m, 13500m, 11500m, and 13000m respectively; First, calculate the sum of the historical average service mileage within each historical sample region: Get the real-time order demand for the current period. (units), preset average service mileage per order ; ; The results show that the real-time order demand mileage in the current cycle (60,000m) exceeds the historical average supply capacity (50,000m) for the same period, thus generating a weighting factor of less than 1 to lower the assessment energy value of the supply side. Assuming the effective energy supply value obtained through the aforementioned calculations Weighted effective supply energy value The calculation is as follows: ; Assuming the total real-time order demand in the current forecast region is D=200, the sum of the expected service mileage for each order is... The measured value is 60,000m. The corrected imbalance prediction intensity index is calculated as follows: ; After dynamic weighting factor correction, the degree of supply shortage in the region (negative value direction) is more accurately reflected. Without this dynamic weighting correction, the imbalance prediction intensity index would be calculated based on 55000m, leading to an overly optimistic assessment of risk. This application achieves cyclical compensation and precise alignment of effective supply energy value.

[0031] This application further proposes that, after generating the modified imbalance prediction intensity index, the method also includes: Based on the corrected imbalance prediction intensity index and regional transaction deviation, the linear deviation value is obtained, and the specific formula is as follows: ; in, This is a linear deviation value, reflecting the residual between the corrected imbalance strength and the actual transaction deviation on the normalized dimension. The regional transaction deviation is (m). The weighted effective energy supply value (m); The dynamic weighting factor for the next period is adjusted in real time using the linear deviation value. The specific formula is as follows: ; in, The corrected dynamic weighting factor for the next cycle. The dynamic weighting factor for the current period. For learning rate weighting operators; The system adjusts the contribution weight of the effective supply energy value in generating the corrected imbalance prediction intensity index based on the linear deviation value. Specifically, the system adjusts the contribution weight based on the linear deviation value. The contribution weight of the adjusted effective supply emergy in generating the adjusted imbalance prediction intensity index Contribution weight Used to adjust the weighted effective supply energy value in the numerator of the generation index. The proportional relationship between the order demand mileage and the exponential generation formula evolved as follows: ; Among them, contribution weight The correction logic is as follows: ; in, This is the proportional correction factor, when the linear deviation value An increase in the value indicates that the predicted supply is too high relative to the transaction feedback. In this case, the contribution weight should be reduced. It can directly suppress the weighted effective supply energy value in the imbalance prediction intensity index generation process.

[0032] In this embodiment of the application, the learning rate weighting operator By analyzing the evolution curves of dynamic weighting factors over historical periods and the rate of elimination of regional transaction deviations, the most suitable iteration step size for the current regional characteristics can be found; proportional correction coefficient. The sensitivity mapping relationship between the deviation value and the contribution weight of the effective supply energy value can be determined by using linear regression analysis of the historical imbalance prediction intensity index and the actual transaction results.

[0033] By calculating the linear deviation value, the system can quantify the normalized residual between the predicted imbalance intensity and the actual transaction deviation, thereby transforming the macro transaction results into correction instructions that can be used for parameter evolution. This constructs a dynamically evolving closed-loop system that can not only optimize the dynamic weight factor of the next cycle online through the learning rate weight operator, making it more versatile in time and space, but also directly increase or suppress the supply energy value in the imbalance prediction intensity index generation stage through real-time adjustment of contribution weight. This dual correction mechanism ensures that the imbalance prediction intensity index can still maintain extremely high predictive sensitivity and accuracy when facing non-steady-state business fluctuations. Compared to error compensation techniques or offline parameter recalibration schemes based on fixed proportional coefficients commonly used in the field, this application has superior real-time response capabilities and physical-logical consistency. Fixed proportional coefficient schemes cannot dynamically adjust the correction intensity according to the measured strength of transaction deviations, which can easily lead to overshooting or adjustment lag in the system. Offline recalibration schemes cannot cope with real-time operational drift. This application chooses to use the linear deviation value to simultaneously apply to the dynamic weight factor of the next cycle and the currently generated contribution weight. In fact, it performs error attribution compensation synchronously in the time and logical dimensions. Through the coupling of the proportional correction coefficient and the learning rate weight operator, the originally abstract business transaction residual is transformed into precise control of the contribution of physical energy value. This realizes the technical leap of the prediction model from "static mapping" to "dynamic adaptive evolution", fundamentally solving the problem of perception strategy failure caused by the disconnect between prediction and transaction.

[0034] Example 2: This is one embodiment of the present invention, which differs from the previous embodiment in that: A short-term forecasting system for urban supply and demand imbalances used by ride-hailing platforms includes: Multi-dimensional data acquisition module: used to acquire the intersection turning restrictions, terrain slope, remaining driving range of the transport vehicle, cumulative online time of the transport vehicle, spatial distribution data of the transport vehicle, real-time order demand, and regional transaction deviation for the current period; Energy efficiency benchmark construction module: used to generate path resistance variables based on intersection turning restrictions and terrain slope, and to construct an efficiency degradation benchmark table using the remaining driving range and cumulative online time of the vehicle. Effective supply conversion module: used to inject path resistance variables into the efficiency decay benchmark table, calculate the physical power consumption increment of the transport vehicle, and convert the spatial distribution data of the transport vehicle into an effective supply energy value with physical energy loss attributes; Imbalance Intensity Prediction Module: This module calculates the difference between effective supply capacity and real-time order demand to generate an imbalance prediction intensity index that indicates the intensity of supply-demand imbalance in the current period. Feedback correction module: used to adjust the data collection granularity of the remaining range and cumulative online time of the transport vehicles in the next cycle in reverse, based on the fluctuation frequency of the imbalance prediction intensity index; and By using regional transaction deviations to correct the influence operator weights in the effective supply energy value conversion process, the corrected influence operator weights are used as input parameters for the next cycle to perform cyclical compensation for the effective supply energy value.

[0035] Example 3: In one embodiment of the present invention, which differs from the previous embodiment, the electronic device includes one or more processors and a memory.

[0036] A processor can be a central processing unit (CPU) or other form of processing unit with data processing and / or instruction execution capabilities, and can control other components in an electronic device to perform desired functions.

[0037] The memory may include one or more computer program products, which may include various forms of computer-readable storage media, such as volatile memory and / or non-volatile memory. Volatile memory may include, for example, random access memory (RAM) and / or cache memory. Non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc.

[0038] In one example, the electronic device may also include input devices and output devices, which are interconnected via a bus system and / or other forms of connection mechanisms (not shown). In addition, depending on the specific application, the electronic device may include any other suitable components.

[0039] Example 4: Embodiments of this application may also be computer-readable storage media storing computer program instructions thereon, which, when executed by a processor, cause the processor to perform the steps described in the "Exemplary Methods" section above according to the various embodiments of this application.

[0040] Computer-readable storage media may take the form of any combination of one or more readable media. A readable medium may be a readable signal medium or a readable storage medium. A readable storage medium may, for example, include, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatuses, or devices, or any combination thereof. More specific examples of readable storage media (a non-exhaustive list) include: electrical connections having one or more wires, portable disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0041] The basic principles of this application have been described above with reference to specific embodiments. However, it should be noted that the advantages, benefits, and effects mentioned in this application are merely examples and not limitations, and should not be considered as essential features of each embodiment of this application. Furthermore, the specific details disclosed above are for illustrative and facilitative purposes only, and are not limitations. These details do not restrict the application from being implemented using the specific details described above.

[0042] The block diagrams of devices, apparatuses, devices, and systems involved in this application are merely illustrative examples and are not intended to require or imply that they must be connected, arranged, or configured in the manner shown in the block diagrams. As those skilled in the art will recognize, these devices, apparatuses, devices, and systems can be connected, arranged, and configured in any manner. Words such as “comprising,” “including,” “having,” etc., are open-ended terms meaning “including but not limited to,” and are used interchangeably with them. The terms “or” and “and” as used herein refer to the terms “and / or,” and are used interchangeably with them unless the context clearly indicates otherwise. The term “such as” as used herein refers to the phrase “such as but not limited to,” and is used interchangeably with it.

[0043] It should also be noted that in the apparatus, equipment, and methods of this application, the components or steps can be disassembled and / or recombined. These disassemblies and / or recombinations should be considered as equivalent solutions of this application.

[0044] The above description of the disclosed aspects is provided to enable any person skilled in the art to make or use this application. Various modifications to these aspects will be readily apparent to those skilled in the art, and the general principles defined herein can be applied to other aspects without departing from the scope of this application. Therefore, this application is not intended to be limited to the aspects shown herein, but rather to be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0045] The above description has been given for purposes of illustration and description. Furthermore, this description is not intended to limit the embodiments of this application to the forms disclosed herein. Although numerous exemplary aspects and embodiments have been discussed above, those skilled in the art will recognize certain variations, modifications, alterations, additions, and sub-combinations thereof.

Claims

1. A method for short-term prediction of supply and demand imbalance in urban areas for ride-hailing platforms, characterized in that, include: Obtain the current period's intersection turning restrictions, terrain slope, remaining driving range of the transport vehicles, cumulative online time of the transport vehicles, spatial distribution data of the transport vehicles, real-time order demand, and regional transaction deviation; Based on the intersection turning restrictions and the terrain slope, path resistance variables are generated, and an efficiency decay benchmark table is constructed using the remaining driving range of the transport vehicle and the cumulative online time of the transport vehicle. The path resistance variable is injected into the efficiency decay benchmark table, the physical power consumption increment of the transport vehicle is calculated, and the spatial distribution data of the transport vehicle is converted into an effective supply energy value with physical energy loss attributes. The difference between the effective supply capacity and the real-time order demand is calculated to generate an imbalance prediction intensity index that indicates the intensity of the supply and demand contradiction in the current period. Based on the fluctuation frequency of the imbalance prediction intensity index, the data collection granularity of the remaining driving range and the cumulative online time of the transport vehicle in the next cycle is adjusted in reverse. as well as The influence operator weights in the effective supply energy value conversion process are corrected by using the regional transaction deviation. The corrected influence operator weights are then used as input parameters for the next cycle to perform cyclical compensation for the effective supply energy value.

2. The method for short-term prediction of urban regional supply and demand imbalance for ride-hailing platforms according to claim 1, characterized in that, Before generating the imbalance prediction intensity index, the following steps are also included: Obtain historical supply and demand distribution data for the current cycle; Based on the historical supply and demand distribution data and the real-time order demand, a dynamic weighting factor is generated to correct the effective supply energy value. The weighted effective supply energy value is obtained based on dynamic weighting factors; A revised imbalance prediction intensity index is generated based on the weighted effective supply energy value and the real-time order demand.

3. The method for short-term prediction of urban area supply and demand imbalance for ride-hailing platforms according to claim 2, characterized in that, After generating the revised imbalance prediction intensity index, the following is also included: Based on the corrected imbalance prediction intensity index and the regional transaction deviation, a linear deviation value is obtained; The dynamic weighting factor for the next cycle is corrected in real time using the linear deviation value. The contribution weight of the effective supply energy value in generating the corrected imbalance prediction intensity index is adjusted based on the linear deviation value.

4. The method for short-term prediction of urban area supply and demand imbalance for ride-hailing platforms according to claim 1, characterized in that, The conversion of the effective energy supply includes: Based on the number of turning restrictions at intersections and the frequency of terrain slope fluctuations, the expected power consumption variable of the transport vehicle per unit distance is calculated. The remaining service range of the vehicle is obtained by applying physical loss correction to the remaining range of the vehicle using the expected power consumption variable. The order-taking efficiency weight of the transport vehicle in the current period is determined based on the cumulative online time of the transport vehicle. The remaining service range located at different coordinate positions is spatiotemporally aggregated using the order acceptance efficiency weight to generate the effective supply energy value.

5. The method for short-term prediction of urban area supply and demand imbalance for ride-hailing platforms according to claim 4, characterized in that, The remaining service range of the transport vehicle is obtained, including: Based on the expected power consumption variable, match the energy consumption sensitivity coefficient corresponding to the power source type of the transport vehicle; The energy consumption sensitivity coefficient is coupled with the expected power consumption variable in a time-dependent manner to determine the real-time energy consumption redundancy of the transport vehicle in the current road segment. The remaining service range is obtained by dynamically thresholding the remaining driving range using the real-time energy consumption redundancy.

6. The method for short-term prediction of urban area supply and demand imbalance for ride-hailing platforms according to claim 1, characterized in that, The reverse adjustment of the data acquisition granularity includes: The transient disturbance characteristic value is determined by utilizing the time-domain distribution of the imbalance prediction intensity index within the current period; The transient disturbance feature values ​​are mapped and matched with the preset sampling confidence interval to determine the sensing resource scheduling weight for the next cycle; The data collection granularity of the remaining driving range and the cumulative online time of the transportation vehicle is adjusted in reverse by using the perceived resource scheduling weight.

7. The method for short-term prediction of urban area supply and demand imbalance for ride-hailing platforms according to claim 6, characterized in that, The determination of transient disturbance characteristic values ​​includes: The trend term of the imbalance prediction intensity index is stripped in the time domain to obtain the high-frequency residual sequence of the imbalance prediction intensity index; Calculate the energy distribution entropy of the high-frequency residual sequence within the current period; The intensity of fluctuations in the imbalance prediction intensity index is quantified using the energy distribution entropy to generate the transient disturbance characteristic value.

8. A short-term prediction system for supply and demand imbalance in urban areas for ride-hailing platforms, characterized in that: include: Multi-dimensional data acquisition module: used to acquire the intersection turning restrictions, terrain slope, remaining driving range of the transport vehicle, cumulative online time of the transport vehicle, spatial distribution data of the transport vehicle, real-time order demand, and regional transaction deviation for the current period; Energy efficiency benchmark construction module: used to generate path resistance variables based on the intersection turning restrictions and the terrain slope, and to construct an efficiency decay benchmark table using the remaining driving range of the transport vehicle and the cumulative online time of the transport vehicle; Effective supply conversion module: used to inject the path resistance variable into the efficiency decay benchmark table, calculate the physical power consumption increment of the transport vehicle, and convert the spatial distribution data of the transport vehicle into an effective supply energy value with physical energy loss attributes; Imbalance Intensity Prediction Module: Used to calculate the difference between the effective supply energy value and the real-time order demand, and generate an imbalance prediction intensity index to indicate the intensity of the supply and demand contradiction in the current period. Feedback correction module: used to adjust the data collection granularity of the remaining driving range and the cumulative online time of the transport vehicle in the next cycle in reverse according to the fluctuation frequency of the imbalance prediction intensity index; as well as The influence operator weights in the effective supply energy value conversion process are corrected by using the regional transaction deviation. The corrected influence operator weights are then used as input parameters for the next cycle to perform cyclical compensation for the effective supply energy value.

9. An electronic device comprising a memory and a processor, characterized in that: The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions, which, when executed by the processor, implement the steps of the method as described in any one of claims 1 to 7.

10. A computer storage medium storing computer-executable instructions thereon, characterized in that: When the computer-executable instructions are executed by a processor, they implement the steps of the method as described in any one of claims 1 to 7.

Images