Airport supply chain distribution optimization method and system facing dynamic demand

Abstract

The application discloses an airport supply chain distribution optimization method and system facing dynamic demand, and relates to the technical field of airport meal distribution optimization. The method comprises the following steps: S1, real-time collection of multi-source flight data and external information data, data preprocessing, evaluation of data source alignment, data correction operation according to the data source alignment evaluation result, and identification of a high-confidence data set; S2, construction of a multi-factor feature set through external information data, prediction of meal demand, and implementation of a prediction feedback optimization mechanism according to the meal demand prediction result; S3, prediction of inventory replenishment quantity within a meal replenishment cycle, and implementation of a replenishment strategy according to the inventory replenishment quantity prediction result; and S4, allocation of distribution resources and generation of a distribution path, with visual feedback analysis. The method solves the problem that due to data islands and barriers, it is difficult to predict meal demand and accurately perform supply chain distribution.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of airport meal optimization, in particular to an airport supply chain distribution optimization method and system for dynamic demand. BACKGROUND

[0002] With the continuous increase of passenger flow and flight density in modern airports, the efficiency and fine management of airport material supply and meal distribution are increasingly demanded. In the face of multiple types of flights, multiple time period fluctuations and diversified passenger demand, airport meal supply chains are accelerating the upgrading of digitization and intelligentization.

[0003] For example, the patent for invention with publication number CN118504950A discloses an airport meal scheduling system. By constructing a precise airport three-dimensional model, the topography, buildings, meal distribution routes and other features of the airport can be clearly presented, providing accurate spatial reference for meal scheduling. The positioning module on the distribution vehicle can synchronize the vehicle position to the airport three-dimensional model in real time, so that the dispatch center can master the distribution progress in real time. The distribution planning module can reasonably allocate distribution tasks according to the meal sequence model and food vehicle information, ensuring the load balance of each distribution vehicle and avoiding the situation that some vehicles are overburdened while others are idle. The flight meal analysis module can accurately estimate the number of passengers for each flight by deeply analyzing historical flight data and ticket purchase information, and prepare food materials according to the meal sequence model and the meal demand in the subsequent time interval, so as to ensure timely supply of required food materials.

[0004] For example, the patent for invention with publication number CN119740846B discloses an airport meal scheduling system based on big data analysis, which includes a real-time data perception layer, a decision optimization core layer and an execution monitoring layer. The real-time data perception layer collects flight dynamics, resource status and environmental variable data by means of airport operation database, Internet of Things sensors and RFID tags. The fusion prediction module of the decision optimization core layer uses an improved deep learning model to predict meal demand and resource consumption. The dynamic scheduling module generates resource allocation schemes using a multi-objective genetic algorithm. The path planning module determines the optimal transportation path using an improved ant colony algorithm. The execution monitoring layer monitors scheme execution through a digital twin visualization platform and traces abnormal events through a blockchain traceability unit. The application can effectively improve the rationality of resource allocation, the efficiency of distribution and the reliability of management of airport meal service, and adapt to the operation needs of modern airports.

[0005] However, in the process of implementing the technical scheme of the embodiments of the application, the applicant found that the above-mentioned technology at least has the following technical problems:

[0006] In the traditional airport catering technology, the current flight, passenger, inventory and other multi-source data exist serious island and information barrier, which leads to the data cannot be shared and linked in real time, resulting in inaccurate meal demand prediction, and further easily causing meal shortage and large waste, affecting the airport operation efficiency and service quality.

[0007] Therefore, in view of the above problems, there is an urgent need for an airport supply chain distribution optimization method and system facing dynamic demand. SUMMARY

[0008] Technical problems to be solved

[0009] In view of the deficiencies of the prior art, the present application provides an airport supply chain distribution optimization method and system facing dynamic demand, which solves the problem that it is difficult to predict meal demand for accurate supply chain distribution due to data islands and barriers.

[0010] Technical scheme

[0011] To achieve the above purpose, the present application is realized by the following technical scheme: an airport supply chain distribution optimization method facing dynamic demand, comprising the following steps: S1, real-time collection of multi-source flight data and external information data, data preprocessing of the multi-source flight data and external information data, evaluation of the alignment degree of the data source based on the preprocessed multi-source flight data, and data correction operation according to the data source alignment degree evaluation result, and identification of high confidence data set; S2, constructing a multi-factor feature set through external information data, and combining the multi-source flight data in the high confidence data set to predict the meal demand, and implementing a prediction feedback optimization mechanism according to the meal demand prediction result; S3, predicting the meal replenishment period based on the multi-source flight data, predicting the inventory replenishment quantity within the meal replenishment period by comprehensively considering the meal demand prediction result, and implementing a replenishment strategy according to the inventory replenishment quantity prediction result; S4, integrating the meal demand prediction result and the inventory replenishment quantity prediction result, allocating distribution resources and generating a distribution path, and simultaneously performing visual feedback analysis.

[0012] Further, the real-time collection of multi-source flight data and external information data, and the specific process of data preprocessing of the multi-source flight data and the external information data is: real-time collection of multi-source flight data, including: flight scheduling, passenger ticket booking number, passenger ticket change number, passenger ticket refund number, passenger boarding number, meal inventory number, actual meal demand number; synchronously collect external information data which has influence on distribution demand, including weather data and event information, and use event-driven mechanism to trigger real-time update of external information data; unified format conversion, time synchronization, field mapping and deduplication processing are performed on the multi-source flight data, missing data is completed by interpolation method, extreme abnormal data is removed, and the multi-source flight data is normalized; a supply chain distribution database is constructed, the preprocessed multi-source flight data is stored by multi-table association and label normalization method, and the external information data of the same period is recorded synchronously.

[0013] Further, based on the preprocessed multi-source flight data, the specific process of evaluating the alignment degree of the data source is: counting the number of data sources of the multi-source flight data collected from each business interface and about the same flight and the same object, including the passenger ticket booking number of the flight management interface, the passenger boarding number of the boarding gate management interface, and the actual meal demand number of the warehouse management interface; the actual value of each multi-source flight data is obtained; the actual values of all multi-source flight data are accumulated and averaged to obtain the mean value of the multi-source flight data; for each data source, the difference between the actual value of the multi-source flight data and the mean value of the multi-source flight data is calculated to obtain the data source difference; the data source relative deviation is obtained by dividing the data source difference by the sum of the mean value of the multi-source flight data and a minimum constant term, and taking the absolute value; wherein the minimum constant term is the product of the constant 0.01 and the mean value of the multi-source flight data; based on the number of data sources, the data source relative deviation of each data source is accumulated and averaged to obtain the average deviation of the data source; the square root of the average deviation of the data source is calculated and added to the constant one, and then the natural logarithm is calculated, the result of the natural logarithm calculation is added to the constant one, and the reciprocal is obtained to obtain the data source alignment evaluation value.

[0014] Further, according to the data source alignment evaluation result, the specific process of data correction operation and identification of high confidence data set is: for the multi-source flight data data set with data source alignment evaluation value lower than the consistency threshold, mark it as an abnormal data set that needs to be reviewed, record the actual value of the original multi-source flight data of all business interfaces, the average value of the multi-source flight data and the relative deviation of the data source, identify and label the interface with the largest relative deviation of the data source; at the same time, perform data correction operation: perform data source investigation and multi-source flight data supplement, and prompt the interface person in charge to check the data link and source end state; if the abnormality occurs continuously for three times, the abnormal interface is temporarily shielded; for the multi-source flight data with short-time loss and abnormality, trend inference and multi-source flight data interpolation are performed through sliding window linear regression and K nearest neighbor interpolation method; until the data source alignment evaluation value is greater than or equal to the consistency threshold; for the multi-source flight data data set with data source alignment evaluation value greater than or equal to the consistency threshold, mark it as a high confidence data set, and preferentially include the high confidence data set into the next process; all abnormal data sets that need to be reviewed, data correction operation and data source alignment evaluation value are written into the supply chain distribution database synchronously, the consistency threshold is continuously optimized, and the business interface is upgraded through regular review.

[0015] Further, through external information data, the specific process of constructing a multi-factor feature set is: collecting the actual meal demand number of each day in history and the weather data of the day, constructing a weather influence data set, training the weather influence data set through multiple linear regression, fitting the statistical relationship between weather data and actual meal demand number change, constructing a theoretical demand quantity prediction model, and outputting the theoretical meal demand quantity influenced by the weather of the day; obtaining the actual meal demand number under normal weather in the historical sliding time window, and calculating the mean value to obtain the average meal demand quantity; the difference between the theoretical meal demand quantity and the average meal demand quantity is divided by the average meal demand quantity to obtain the weather influence meal relative change rate and perform normalization processing as the weather influence value; synchronously obtaining the event information of each day in history and the actual meal demand number of the day, in the same time window length, counting the actual meal demand number of the event occurrence day and calculating the average value to obtain the average meal demand quantity of the event; at the same time, the actual meal demand number of the day without event occurrence is counted and the average value is calculated to obtain the average meal demand quantity without event; the difference between the average meal demand quantity with event and the average meal demand quantity without event is divided by the average meal demand quantity without event to obtain the event influence meal relative change rate and perform normalization processing as the event influence value.

[0016] Further, in combination with the multi-source flight data in the high-credibility dataset, the specific process of predicting the meal demand is as follows: obtaining the number of passenger tickets, the number of passenger changes and the number of passenger refunds of flight f at time t from the high-credibility dataset, obtaining the actual boarding number by subtracting the number of passenger changes and the number of passenger refunds from the number of passenger tickets, obtaining the actual meal demand and the passenger boarding number of the same flight f in the same period from the supply chain distribution database, obtaining the historical per capita consumption by calculating the ratio of the actual meal demand to the passenger boarding number, and obtaining the historical per capita consumption mean and the historical per capita consumption standard deviation by calculating the historical per capita consumption in the sliding time window of the same flight f, obtaining the historical per capita consumption deviation by subtracting the historical per capita consumption mean from the historical per capita consumption, obtaining the standard per capita consumption offset value by dividing the historical per capita consumption deviation by the historical per capita consumption standard deviation, obtaining the consumption offset smoothing correction value by performing the hyperbolic tangent operation on the standard per capita consumption offset value, obtaining the historical correction term by multiplying the historical correction weight factor and the consumption offset smoothing correction value, obtaining the weather influence value and the event influence value, obtaining the weather correction term by multiplying the weather influence weight factor and the weather influence value, obtaining the event correction term by multiplying the event influence weight factor and the event influence value, obtaining the comprehensive correction coefficient by adding the sum of the historical correction term, the weather correction term and the event correction term to the constant one, and obtaining the meal demand prediction value by calculating the product of the comprehensive correction coefficient and the predicted actual boarding number.

[0017] Further, the specific process of implementing the prediction feedback optimization mechanism according to the meal demand prediction result is as follows: writing the meal demand prediction value of each flight and time period into the supply chain distribution database, and synchronously pushing it to the inventory management and replenishment strategy module to drive meal replenishment and distribution resource scheduling; collecting the actual meal demand and comparing it with the meal demand prediction value whenever the actual meal distribution and consumption are completed, and recording and analyzing the prediction error in real time; periodically retraining the theoretical demand prediction model and the meal demand prediction value algorithm based on the accumulated prediction error data, and adjusting the prediction parameters; if it is found that the deviation between the continuous meal demand prediction value and the actual meal demand is higher than the deviation threshold, triggering an abnormal warning, prompting manual review and data backtracking.

[0018] Further, based on the multi-source flight data, the meal replenishment cycle is predicted, the inventory replenishment quantity in the meal replenishment cycle is predicted based on the meal demand prediction result, and the specific process of implementing the replenishment strategy according to the inventory replenishment quantity prediction result is: real-time acquisition of meal inventory quantity, recording of meal consumption cycle, calculation and statistics of meal consumption speed, identification of risk of expiration and shortage, prediction of meal replenishment cycle; obtaining the meal demand prediction value in the replenishment cycle, accumulating the meal demand prediction value in each time period to obtain the total demand prediction quantity in the replenishment cycle; at the same time, the standard deviation of the meal demand prediction value in the replenishment cycle is calculated to obtain the demand prediction standard deviation; obtaining the meal inventory quantity at the current time, subtracting the meal inventory quantity at the current time from the total demand prediction quantity in the replenishment cycle to obtain the theoretical replenishment demand quantity; multiplying the demand prediction standard deviation by the fluctuation weight factor to obtain the compensation safety inventory quantity; adding the compensation safety inventory quantity to the theoretical replenishment demand quantity to obtain the inventory replenishment prediction value, performing maximum function operation on the inventory replenishment prediction value, that is, taking the maximum value between the inventory replenishment prediction value and 0, if the inventory replenishment prediction value is negative, the inventory replenishment prediction value is 0, otherwise, the inventory replenishment prediction value is the actual calculation value; according to the inventory replenishment prediction value, the replenishment strategy is implemented: generating a replenishment work order and pushing it to the procurement and warehousing departments in real time; dynamically monitoring the actual replenishment and inventory consumption process, periodically collecting the actual meal demand quantity and the meal inventory quantity, and comparing them with the meal demand prediction value and the inventory replenishment prediction value, and statistically analyzing the shortage and overstock situations, adjusting the distribution resources and meal storage; periodically backtracking to evaluate the replenishment performance brought by different fluctuation weight factors, adjusting and optimizing the fluctuation weight factor and the replenishment parameters, realizing the self-learning and continuous evolution of the replenishment strategy; at the same time, all replenishment results, meal inventory quantity fluctuations and parameter optimization records are real-time archived.

[0019] Further, the specific process of fusing the meal demand prediction result and the inventory replenishment prediction result, allocating distribution resources and generating a distribution path, while performing visual feedback analysis, is as follows: according to the meal demand prediction value and the inventory replenishment prediction value, the distribution demand of each type of meal is refined into a specific distribution task order, and the corresponding flight, time node and meal category are determined; based on the currently available vehicles, distribution personnel, distribution time window and resource conditions of cold chain support, the optimal distribution resources are allocated, and the flight priority, distribution distance and meal preservation business constraints are fully considered to achieve scientific scheduling; in combination with the real-time road, traffic conditions and climate information of the airport, the shortest path algorithm is used to determine the optimal route for each batch of distribution tasks, and the task merging of multiple flights in the same region is supported; the distribution task is issued, and the distribution task is monitored and positioned throughout the execution process, and if an emergency occurs, real-time rescheduling is triggered to ensure that the distribution task is completed as needed, and all distribution process exceptions are pushed in synchronization; at the same time, the state of each distribution task, the distribution resource allocation, the real-time distribution vehicle and personnel position distribution, the optimal path trajectory, and the key node warning and historical distribution abnormal heat map are dynamically presented in real time.

[0020] The second aspect of the present application provides an airport supply chain distribution optimization system for dynamic demand, comprising: a multi-source data acquisition and fusion module for acquiring multi-source flight data and external information data in real time, pre-processing the multi-source flight data and external information data, evaluating the alignment degree of the data sources based on the pre-processed multi-source flight data, and performing data correction operation according to the data source alignment degree evaluation result to identify a high-confidence data set; a demand dynamic prediction module for constructing a multi-factor feature set through external information data, predicting meal demand in combination with multi-source flight data in the high-confidence data set, and implementing a prediction feedback optimization mechanism according to the meal demand prediction result; an inventory management and replenishment strategy module for predicting meal replenishment period based on multi-source flight data, predicting inventory replenishment quantity within the meal replenishment period in combination with the meal demand prediction result, and implementing a replenishment strategy according to the inventory replenishment quantity prediction result; a distribution scheduling and path optimization module for fusing the meal demand prediction result and the inventory replenishment quantity prediction result, allocating distribution resources and generating a distribution path, while performing visual feedback analysis.

[0021] Advantages

[0022] The present application has the following advantages:

[0023] (1) The present application, through real-time acquisition and unified preprocessing of multi-source flight data, uses a quantitative consistency evaluation method to effectively improve the data credibility; and can automatically identify and correct abnormal data, realize high-quality data-driven business flow, and ensure the accuracy and traceability of supply chain decision-making.

[0024] (2) The application is based on historical business and real-time external information data, constructs a multi-factor feature set, dynamically predicts meal demand, and realizes optimization of prediction parameters and evolution of algorithms through continuous error backflow and model self-learning, thereby improving the accuracy and adaptability of demand prediction.

[0025] (3) The application realizes adaptive decision of replenishment quantity through calculation of demand prediction standard deviation and compensation safety stock number, dynamic adjustment of standard deviation compensation mechanism and historical risk, effectively balances material shortage risk prevention and control and inventory cost control, and improves the safety and economy of replenishment strategy.

[0026] (4) The application realizes dynamic allocation of resources and optimal path through intelligent linkage of demand, inventory and distribution, combination of real-time traffic and weather multi-dimensional factors, supports multi-task combined distribution and emergency rescheduling, monitors distribution status, on-time rate and abnormal events throughout the process, realizes timely early warning and closed-loop tracking.

[0027] Of course, any product implementing the present application does not necessarily need to achieve all the advantages described above. BRIEF DESCRIPTION OF DRAWINGS

[0028] Figure 1 Flow chart of airport supply chain distribution optimization method for dynamic demand;

[0029] Figure 2 Structural diagram of airport supply chain distribution optimization system for dynamic demand;

[0030] Figure 3 Decomposition diagram of inventory replenishment demand driving force;

[0031] Figure 4 Airport meal distribution task panoramic visualization rendering diagram. DETAILED DESCRIPTION

[0032] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application, as understood by those skilled in the art, the described embodiments are only some of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0033] Please refer to Figures 1-4The embodiment of the present application provides a technical scheme: an airport supply chain distribution optimization method and system facing dynamic demand, comprising the following steps: S1, real-time collection of multi-source flight data and external information data, data preprocessing of the multi-source flight data and the external information data, alignment degree evaluation of the data source based on the preprocessed multi-source flight data, data correction operation according to the data source alignment degree evaluation result, and identification of a high-confidence data set; S2, construction of a multi-factor feature set through the external information data, prediction of meal demand in combination with the multi-source flight data in the high-confidence data set, implementation of a prediction feedback optimization mechanism according to the meal demand prediction result; S3, prediction of a meal replenishment period based on the multi-source flight data, prediction of the inventory replenishment quantity within the meal replenishment period in combination with the meal demand prediction result, and implementation of a replenishment strategy according to the inventory replenishment quantity prediction result; S4, distribution resource allocation and generation of a distribution path while performing visual feedback analysis by fusing the meal demand prediction result and the inventory replenishment quantity prediction result.

[0034] Specifically, real-time collection of multi-source flight data and external information data, the specific process of data preprocessing of multi-source flight data and external information data is: connecting with airport flight management, passenger information, inventory, warehousing system, realizing API bus direct connection, real-time collection of multi-source flight data, multi-source flight data includes: flight scheduling, passenger ticket booking number, passenger ticket change number, passenger ticket refund number, passenger boarding number, food inventory number, actual food demand number; synchronously collect external information data which has influence on distribution demand, including weather data and event information, wherein the event information is, for example, holidays, large activities, and an event-driven mechanism is used to trigger real-time update of external information data; unified format conversion, time synchronization, field mapping and deduplication processing are performed on the multi-source flight data, missing data is completed by interpolation method, and extreme abnormal data is removed, and the multi-source flight data is normalized; wherein the unified format conversion includes but is not limited to unified standardization of data types, coding methods and field naming of different source data, so as to eliminate the heterogeneity between systems. Time synchronization refers to aligning the time fields of different data sources to the same time zone and time granularity, ensuring the consistency of subsequent data association. Field mapping is to merge the original data fields to a unified label system according to the preset mapping rules during data processing, for example, ticket booking number and ticket purchase number are uniformly mapped to passenger ticket booking number. The deduplication processing only retains the latest data of the same object. A supply chain distribution database is constructed, and the preprocessed multi-source flight data is stored by multi-table association and label normalization method, and the external information data of the same period is recorded synchronously. Among them, multi-table association refers to joint query and associated storage of different source data tables through setting primary keys such as flight number, time, and material number, forming a data calling relationship that can support cross-business; label normalization is to merge and store different source data items through a unified label system on the basis of multi-table association, for example, ticket purchase number and total ticket number from different interfaces are uniformly marked as passenger ticket booking number, and a standardized data dictionary is formed through the label system.

[0035] In the embodiment, by unified collection, standardized preprocessing and intelligent completion of multi-source flight data and external information data, the data heterogeneity and information island problem are effectively eliminated, the data is structured, labeled and time-space consistent archived. Through multi-table association and label normalization, heterogeneous data from different business interfaces can be fused and managed according to unified standards, improving the uniqueness, accuracy and traceability of the data, and providing a high-quality and time-efficient data basis for the core links of subsequent supply chain demand prediction, replenishment optimization and intelligent scheduling.

[0036] Specifically, based on the pre-processed multi-source flight data, the specific process of evaluating the alignment degree of the data source is: the number of data sources of the multi-source flight data collected from each business interface and about the same object of the same flight is counted, including the number of passenger tickets of the flight management interface, the number of passengers boarding of the boarding gate management interface, and the actual number of meal demand of the warehouse management interface; the data source can come from multiple business sub-interfaces, and a source label is automatically attached to each data during collection, realizing subsequent multi-source tracing and difference comparison; at the same time, the actual value of each multi-source flight data is obtained, in order to ensure the consistency of the evaluation, the unit standardization processing is performed on all collected values, so as to ensure the comparability of the statistical values; the actual values of all multi-source flight data are accumulated, and the average value of the multi-source flight data is obtained; for each data source, the difference between the actual value of the multi-source flight data and the average value of the multi-source flight data is calculated to obtain the data source difference; the data source relative deviation is obtained by dividing the data source difference by the sum of the average value of the multi-source flight data and the minimum constant term, and taking the absolute value; wherein the minimum constant term is the product of the constant 0.01 and the average value of the multi-source flight data, and the introduction of the minimum constant term can effectively avoid the error amplification caused by the too small denominator when the average value is close to zero; based on the number of data sources, the data source relative deviation of each data source is accumulated, and the average value is calculated to obtain the data source average deviation; the square root of the data source average deviation is calculated, and the constant one is added, and then the natural logarithm is calculated, the natural logarithm calculation result is added to the constant one, and the reciprocal is obtained to obtain the data source alignment degree evaluation value.

[0037] The specific formula of the data source alignment degree evaluation value is:

[0038] ;

[0039] In the formula, The data source alignment degree evaluation value is used to dynamically evaluate the consistency of multi-source data at the same time, that is, whether the data collected by different business interfaces at the same time on the same object is close enough; the closer the data source alignment degree evaluation value is to 1, the better the consistency, and the closer to 0, the greater the divergence and the existence of data island; wherein the natural logarithm calculation is performed, in order to smooth the small deviation and stretch the large deviation; The actual value of the i-th multi-source flight data reflects the observation of different business interfaces on the same object, and is used to measure the actual collected data of each business interface on the same object, which is the basis for the data source alignment degree evaluation value; The average value of the multi-source flight data is used as a reference benchmark for data alignment, and is used to measure the deviation of the actual value of the multi-source flight data of each interface; The number of data sources of the multi-source flight data is used to standardize and average the data source alignment degree evaluation values of all data sources, so as to ensure the consistency of the data source alignment degree evaluation values at different quantities; represents a minimum constant term, is a minimum positive number that is adaptive, is set as the product of the constant 0.01 and the mean of the multi-source flight data, prevents the denominator from being zero when the mean is zero, and makes the formula still work stably in the case of small data magnitude; represents the relative deviation of the data source, represents the normalized deviation of each data source, and ensures that different types and different magnitudes of data can participate in the data source alignment evaluation fairly.

[0040] In the embodiment, by automatically attaching source labels to multi-source flight data, uniform unit standardization processing, and using a quantitative calculation data source alignment evaluation formula, efficient comparison and quality monitoring of different business interface data are realized. Through deviation analysis and alignment scoring of multi-source data, the scientificity and transparency of data fusion are improved, and abnormal and low-credibility data sources can be accurately identified, thereby enhancing the reliability of subsequent supply chain prediction and decision-making and data traceability.

[0041] Specifically, according to the data source alignment evaluation result, the specific process of identifying a high-credibility data set is as follows: for a multi-source flight data set with a data source alignment evaluation value lower than the consistency threshold, mark it as an abnormal data set that needs to be reviewed, record the actual values of the original multi-source flight data of all business interfaces, the mean of the multi-source flight data, and the relative deviation of the data source, identify and label the interface with the largest relative deviation of the data source, which helps to systematically analyze the main source of data deviation and assist in subsequent responsibility attribution and link investigation; at the same time, perform data correction operations: perform data source investigation and multi-source flight data resampling, and prompt the interface person in charge to check the data link and source status; trigger the interface API re-pulling mechanism during resampling, and support manual review and comparison with the third-party authoritative source; if abnormality occurs continuously for three times, temporarily shield the abnormal interface to prevent low-quality data from repeatedly interfering with the business process and improve the overall data credibility; for short-time missing and abnormal multi-source flight data, perform trend inference and multi-source flight data interpolation through sliding window linear regression and K-nearest neighbor interpolation method; not only improves the accuracy of missing repair, but also selects the optimal completion scheme according to the historical trend and multi-dimensional features. Until the data source alignment evaluation value is greater than or equal to the consistency threshold; for a multi-source flight data set with a data source alignment evaluation value greater than or equal to the consistency threshold, mark it as a high-credibility data set, and preferentially include the high-credibility data set into the next process; write all abnormal data sets that need to be reviewed, data correction operations, and data source alignment evaluation values into the supply chain distribution database, continuously optimize the consistency threshold, and upgrade the business interface through regular review. Archive the relevant historical data correction and optimization logs to provide traceable basis for self-learning, abnormal source tracing, and interface capability evolution.

[0042] In this embodiment, through quantitative evaluation and dynamic correction, abnormal and low alignment data sources can be accurately identified and isolated, systematic tracking and responsibility attribution of data deviation can be realized. Through intelligent resampling and interpolation algorithm, short-term missing and abnormal data can be effectively repaired, and the accuracy and self-adaptability of data completion can be improved. Abnormal interface shielding and high credibility data set priority mechanism ensure that subsequent business processes are always based on high-quality data, enhancing the reliability of data and the scientificity of supply chain decision-making. At the same time, it promotes the intelligentization and closed-loop continuous optimization of airport supply chain data governance.

[0043] Specifically, the specific process of constructing the multi-factor feature set through external information data is as follows: the actual meal demand number of each day in history and the weather data of the day are collected to construct a weather influence data set. Through multiple linear regression training on the weather influence data set, the statistical relationship between weather data and actual meal demand number change is fitted to construct a theoretical demand quantity prediction model, which outputs the theoretical meal demand quantity affected by weather on the day. Multiple linear regression training introduces temperature, rainfall, wind speed, humidity and other weather variables as independent variables, and actual meal demand number as dependent variable. After fitting the theoretical demand quantity prediction model, the weight coefficient of each weather element is obtained, realizing the quantification of the influence of weather on demand quantity. Batch training and regular review are supported to adapt to different seasons and abnormal climate changes. The actual meal demand number under normal weather in the historical sliding time window is obtained, and the average meal demand quantity is calculated by taking the mean value. The difference between the theoretical meal demand quantity and the average meal demand quantity is divided by the average meal demand quantity to obtain the weather-affected meal relative change rate and normalize it as the weather influence value. The maximum and minimum normalization method is adopted to ensure that the characteristic quantity has the same dimension and consistent numerical range. The event information of each day in history and the actual meal demand number of the day are obtained synchronously. In the same time window length, the actual meal demand number of the event-occurred day is counted and the average value is calculated to obtain the average meal demand quantity of the event-occurred day. At the same time, the actual meal demand number of the event-occurred day is counted and the average value is calculated to obtain the average meal demand quantity of the event-occurred day. Event information includes, for example, holidays, exhibitions, large-scale activities, traffic control and other business scenarios, which are archived through structured label method to distinguish the influence degree of different events on meal demand. The difference between the average meal demand quantity of the event-occurred day and the average meal demand quantity of the event-occurred day is divided by the average meal demand quantity of the event-occurred day to obtain the event-affected meal relative change rate and normalize it as the event influence value. The constructed weather influence value and event influence value are important components of the multi-factor feature set, which are input into the subsequent meal demand prediction algorithm together with the passenger and flight main business features to realize multi-dimensional, dynamic and traceable demand quantity modeling. The feature set supports dynamic expansion and update, and the model structure and parameters can be adjusted according to new event types and meteorological elements.

[0044] In this embodiment, by introducing external information of historical weather, events, using multiple linear regression upper data modeling means, the meteorological variables and structured event labels are quantified as multi-factor features that can be used for intelligent prediction, realizing the precise analysis and dynamic perception of meal demand fluctuation. Through standardization processing such as maximum minimum normalization, the dimensions of the features are unified to facilitate algorithm fusion and multi-scene migration. Regular batch model training and feature expansion mechanism enable the prediction model to flexibly adapt to seasonal changes, unexpected events and other complex business situations, providing high-quality and traceable data support for intelligent replenishment and scheduling decisions.

[0045] Specifically, in combination with multi-source flight data in the high-confidence dataset, the specific process of predicting meal demand is as follows: obtaining the number of passenger bookings, the number of passenger changes, and the number of passenger refunds of flight f at time t from the high-confidence dataset, subtracting the number of passenger changes and the number of passenger refunds from the number of passenger bookings to obtain the expected actual number of passengers, this step is automatically aggregated and calculated through a data interface to ensure the real-time and accuracy of the prediction base; obtaining the actual meal demand and the number of passengers on board of the same flight f in the same period from the historical high-confidence dataset in the supply chain distribution database, calculating the ratio of the actual meal demand and the number of passengers on board to obtain the historical per capita consumption, and calculating the mean value to obtain the historical per capita consumption mean value, and calculating the standard deviation to obtain the historical per capita consumption standard deviation; the length and step of the sliding time window are dynamically adjusted according to business needs to achieve a sensitive response to changes in demand patterns. Subtracting the historical per capita consumption mean value from the historical per capita consumption to obtain the historical per capita consumption deviation, dividing the historical per capita consumption deviation by the historical per capita consumption standard deviation to obtain the standard per capita consumption offset value, performing a hyperbolic tangent operation on the standard per capita consumption offset value to obtain a consumption offset smoothing correction value; wherein the hyperbolic tangent operation, i.e., the tanh function, is a commonly used nonlinear numerical transformation method that can smooth and compress any real number to between -1 and 1, effectively suppressing the influence of outliers and preserving the positive and negative trend information of the data, which can effectively suppress the amplification effect of extreme offset on demand prediction in this implementation method. Multiply the historical correction weight factor by the consumption offset smoothing correction value to obtain the historical correction term; obtain the weather influence value and the event influence value, multiply the weather influence weight factor by the weather influence value to obtain the weather correction term; multiply the event influence weight factor by the event influence value to obtain the event correction term; add a constant one to the sum of the historical correction term, the weather correction term, and the event correction term to obtain a comprehensive correction coefficient, calculate the product of the comprehensive correction coefficient and the expected actual number of passengers to obtain the meal demand prediction value. The comprehensive correction coefficient integrates the dynamic influence of multiple factors on demand, achieving multi-dimensional precise modeling of meal demand in complex changing scenarios. The meal demand prediction value is used to drive the subsequent supply chain business links of replenishment, production, and scheduling.

[0046] Wherein, the specific formula of the meal demand prediction value is:

[0047] ;

[0048] In the formula, represents the meal demand prediction value of flight f at time t, considering the actual boarding number, historical consumption fluctuation, weather and special event multi-dimensional influence factors, taking the latest boarding number as the demand baseline, and correcting the baseline through multiple dynamic items to reflect the meal demand elasticity change under the complex scene of actual business; represents the predicted actual boarding number of flight f at time t, which is the baseline of demand prediction, reflecting the dynamic change of actual service object; represents the historical per capita consumption of flight f in the same period, reflecting the historical law and consumption habit, and assisting in correcting the current demand baseline; represents the historical per capita consumption mean value of flight f, which is the baseline reference of historical consumption, compared with the current historical consumption; represents the historical per capita consumption standard deviation of flight f, which measures the volatility of historical consumption change; represents the weather influence value, reflecting the influence of weather change on actual meal demand, such as increasing meal and material backup in bad weather; represents the event influence value, supplementing the sudden factors outside the regular business, such as the adjustment effect of peak passenger flow and sudden capacity change on demand; represents the consumption offset smoothing correction value, which is a hyperbolic tangent function, normalizing the standard per capita consumption offset value to the interval of-1 to 1, which can reflect the positive and negative changes of historical per capita consumption, and avoid the dramatic amplification of extreme abnormality on prediction; make the prediction have sensitivity to the abnormal change of historical consumption, but not be severely disturbed by extreme historical fluctuation; represents the historical correction weight factor, based on the historical consumption offset smoothing correction value data set, using the least square method to fit the meal demand prediction value, taking the historical correction weight factor as the dependent variable and the historical consumption offset smoothing correction value as the independent variable, obtaining the optimal historical correction weight factor, the value range is between-2 and 2; represents the weather influence weight factor, based on the historical weather influence value data set, using the least square method to fit the meal demand prediction value, taking the weather influence weight factor as the dependent variable and the weather influence value as the independent variable, obtaining the optimal weather influence weight factor, the value range is between-2 and 2; represents the event influence weight factor, based on the historical event influence value data set, using the least square method to fit the meal demand prediction value, taking the event influence weight factor as the dependent variable and the event influence value as the independent variable, obtaining the optimal event influence weight factor, the value range is between-2 and 2.

[0049] In this embodiment, through deep aggregation and dynamic analysis of high-credibility multi-source flight data, combined with sliding window statistics of historical per capita consumption, hyperbolic tangent smoothing correction, and dynamic adjustment of weather and event multi-factors, high-precision, real-time, and multi-dimensional prediction of meal demand is realized. This method not only improves the robustness of demand modeling to sudden changes and extreme abnormalities, but also adjusts parameters adaptively for different business scenarios, ensuring that the prediction results always match the actual operation, providing a solid data foundation and scientific decision support for the supply chain links of airport meal replenishment, production planning, and intelligent scheduling, and enhancing the intelligent level and service guarantee capability of airport operation.

[0050] Specifically, the specific process of implementing the prediction feedback optimization mechanism according to the meal demand prediction result is as follows: write the meal demand prediction value of each flight and time period into the supply chain distribution database, and synchronously push it to the inventory management and replenishment strategy module to drive meal replenishment and distribution resource scheduling; whenever actual meal distribution and consumption is completed, collect the actual meal demand and compare it with the meal demand prediction value, and record and analyze the prediction error in real time; based on the accumulated prediction error data, periodically retrain the theoretical demand prediction model and meal demand prediction value algorithm, and adjust the prediction parameters; the model retraining period can be flexibly configured, historical data sets are called through automatic scripts, and multiple algorithms are supported for parallel evaluation and optimal strategy selection. If it is found that the deviation between the continuous meal demand prediction value and the actual meal demand is higher than the deviation threshold, an abnormality warning is triggered, prompting manual review and data backtracking. Multiple alarm modes such as SMS and pop-up window are supported, and the responsible person and traceability interface are linked to ensure that potential abnormalities are discovered and closed in a timely manner.

[0051] In this embodiment, by realizing full-process data closed loop of meal demand prediction and actual consumption, combined with real-time monitoring of prediction error and dynamic model retraining, the accuracy and adaptability of the prediction result are effectively improved. Multi-algorithm parallel optimization and automatic strategy selection ensure that the model remains in the optimal state. The abnormality warning and multi-channel linkage alarm mechanism realize the timely discovery, responsibility tracing, and efficient closed-loop disposal of data abnormalities.

[0052] Specifically, based on the multi-source flight data, the meal replenishment cycle is predicted, the meal demand prediction results are integrated, the inventory replenishment quantity in the meal replenishment cycle is predicted, and according to the inventory replenishment quantity prediction result, the specific process of implementing the replenishment strategy is: the meal inventory quantity is obtained in real time, and the meal consumption cycle is recorded, the meal consumption speed is calculated and counted, the risk of expiration and shortage is identified, and the meal replenishment cycle is predicted; the inventory collection interface supports minute-level data synchronization, and archives the historical consumption curve, providing a data basis for trend analysis and risk warning. Obtain the meal demand prediction value in the replenishment cycle, accumulate the meal demand prediction value of each time period to obtain the total demand prediction amount in the replenishment cycle; at the same time, the standard deviation of the meal demand prediction value in the replenishment cycle is calculated to obtain the demand prediction standard deviation; obtain the meal inventory quantity at the current time, subtract the meal inventory quantity at the current time from the total demand prediction amount in the replenishment cycle to obtain the theoretical replenishment demand quantity; multiply the demand prediction standard deviation by the fluctuation weight factor to obtain the compensation safety inventory quantity; add the compensation safety inventory quantity to the theoretical replenishment demand quantity to obtain the inventory replenishment prediction value, and perform maximum function operation on the inventory replenishment prediction value, that is, take the maximum value between the inventory replenishment prediction value and 0, if the inventory replenishment prediction value is negative, the inventory replenishment prediction value is 0, otherwise, the inventory replenishment prediction value is the actual calculation value; effectively avoid inventory surplus and replenishment redundancy, and balance supply safety and resource optimization. According to the inventory replenishment prediction value, the replenishment strategy is implemented: a replenishment work order is generated and pushed to the procurement and warehousing departments in real time; the actual replenishment and inventory consumption process are dynamically monitored, the actual meal demand quantity and the meal inventory quantity are collected regularly, and are compared with the meal demand prediction value and the inventory replenishment prediction value, the shortage and overstock situations are counted, and the distribution resources and meal storage are adjusted; a replenishment suggestion sheet and an adjustment task work order are generated to trigger the collaborative response of resource distribution and warehouse stocking. Periodically backtrack the replenishment performance brought by different fluctuation weight factors, adjust and optimize the fluctuation weight factor and the replenishment parameter through machine learning method, realize the self-learning and continuous evolution of the replenishment strategy, adapt to demand fluctuation and supply chain change; at the same time, all replenishment results, meal inventory quantity fluctuations and parameter optimization records are archived in real time, providing a full traceable basis for operation and decision-making and risk tracing, ensuring closed-loop optimization of supply chain management.

[0053] wherein the specific formula of the inventory replenishment prediction value is:

[0054]

[0055] In the formula, represents the inventory replenishment prediction value, which is used to calculate the required replenishment quantity for meals, combines the demand prediction in the future period, the current actual inventory, and the volatility of demand prediction, and compensates the risk through the compensation safety inventory quantity, to ensure that the replenishment is neither short nor wasted; represents the replenishment cycle of the meal; represents the total demand prediction in the replenishment cycle, reflecting how much material is expected to be consumed in the future, and is the core basis for replenishment; represents the current meal inventory, used to offset future demand and prevent repeated replenishment and inventory accumulation; represents the demand prediction standard deviation, used to compensate for possible errors in demand prediction and improve the safety and robustness of replenishment; represents the volatility weight factor, which adjusts the strength of standard deviation compensation and reflects the level of risk tolerance and defense; the total demand prediction, demand prediction standard deviation, stockout days, and meal inventory in the historical replenishment cycle are collected, a set of candidate volatility weight factors is set, historical replenishment scenarios are replayed, the total loss caused by shortage and accumulation under different volatility weight factors is calculated, the volatility weight factor that minimizes the historical total loss is selected as the optimal volatility weight factor for the current business, and the value range is between 0.2 and 2.

[0056] The volatility weight factor is set to 1, and in the replenishment cycle, the inventory replenishment prediction value of each period is calculated according to the total demand prediction of different periods, the current meal inventory, and the demand prediction standard deviation. As shown in Table 1, the inventory replenishment prediction value data table.

[0057] Table 1 Inventory Replenishment Prediction Value Data Table

[0058]

[0059] As Figure 3 shown, the inventory replenishment demand driving force decomposition diagram provided by the embodiment of the present application shows the comparison of each driving factor and the change trend of the actual replenishment amount in the airport meal replenishment decision-making process. The columnar part distinguishes the total demand prediction, the current meal inventory, and the safety inventory in the replenishment cycle by different colors, and the black line represents the actual inventory replenishment prediction value calculated in each cycle; according to Table 1 and Figure 3 It can be seen that the replenishment in each cycle is mainly determined by the total demand prediction; the combination of the black line and the columnar distribution intuitively reflects that the replenishment amount is not only affected by the demand and safety compensation, but also is offset by the inventory, such as time 5, the inventory is almost equal to the demand, and only a small amount of replenishment is needed, effectively avoiding the waste caused by excessive replenishment.

[0060] In this embodiment, dynamic monitoring and trend analysis of the whole process of meal inventory consumption are achieved, and the fine management and risk warning capability of inventory management are improved. The replenishment strategy is driven by demand prediction, inventory status and uncertainty compensation, effectively balancing replenishment safety and resource optimization, preventing inventory accumulation and shortage. Using machine learning intelligent algorithm to optimize replenishment parameters and fluctuation weight factors can continuously evolve and flexibly respond to demand fluctuations and complex business scenarios. All replenishment operations, inventory changes and parameter adjustment processes are archived, further promoting the intelligent, closed-loop and efficient development of airport supply chain replenishment management.

[0061] Specifically, the specific process of fusing meal demand prediction results and inventory replenishment prediction results, allocating distribution resources and generating distribution paths, and simultaneously performing visual feedback analysis is as follows: according to the meal demand prediction value and the inventory replenishment prediction value, the distribution demand of various meals is refined into specific distribution task sheets, and the corresponding flights, time nodes and meal categories are determined; a task sheet template is generated, which supports multiple task attribute configurations to ensure flexible and controllable task assignment. According to the available resources of vehicles, distribution personnel, distribution time window and cold chain guarantee, the optimal distribution resources are allocated, and the flight priority, distribution distance and meal preservation business constraints are fully considered to achieve scientific scheduling; a multi-objective optimization algorithm is used to balance the distribution workload and timeliness, and to improve the distribution efficiency and resource utilization. Combined with real-time road, traffic conditions and climate information in the airport, the shortest path algorithm is used to determine the optimal route for each batch of distribution tasks, and to support task consolidation for multiple flights in the same area; the path planning supports dynamic correction to realize flexible adjustment of the distribution route. The distribution task is issued, and the distribution task execution process is monitored and positioned throughout the process. If an emergency occurs, real-time rescheduling is triggered to ensure that the distribution task is completed as needed, and all distribution process exceptions are pushed in synchronization, and each scheduling adjustment and exception handling process is recorded; at the same time, the state of each distribution task, the allocation of distribution resources, the real-time distribution vehicle and personnel position distribution, the optimal path trajectory, and the key node warning and historical distribution abnormal heat map are dynamically presented in real time.

[0062] As Figure 4As shown, the airport meal delivery task panoramic visualization rendering diagram provided by the embodiment of the application intuitively identifies the states of each delivery task with different colors and shapes: a green dot represents a completed task, an orange dot represents a delivery task being executed, and a red square is used to highlight an abnormal task. A purple triangle represents the real-time location of the current delivery resource, and each delivery resource is responsible for a group of tasks with the same state. The actual scheduling and delivery path is reflected by the clear connection through a dashed line. A yellow five-pointed star is a warning node, which is superimposed on the abnormal task point to highlight the urgent processing needs. The red heat zone in the background shows the high-incidence area of historical abnormal tasks; the deeper the color, the higher the frequency of historical delivery abnormalities in the area, which is a high-risk area of abnormal problems; the lighter and colorless area means that the number of past abnormal occurrences is small, and the delivery process is relatively more stable, which intuitively reflects the prone location and frequency of abnormal tasks in the airport area.

[0063] In the embodiment, the meal demand prediction is dynamically fused with the inventory replenishment demand to realize accurate decomposition and intelligent dispatch of the delivery task. The introduction of multi-objective optimization and shortest path algorithm enables the delivery resource allocation and route planning to simultaneously consider efficiency, timeliness and safety, greatly improving the scientificity and flexibility of the delivery work. Task whole-process visualization, abnormal automatic pushing and scheduling closed-loop management provide real-time and transparent full-link data support for operation and decision-making, effectively improving the response capability, service quality and risk prevention and control level of the airport supply chain, and promoting the continuous evolution of delivery management towards intelligence, refinement and high reliability.

[0064] Reference Figure 2 As shown, the second aspect of the application provides an airport supply chain delivery optimization system for dynamic demand, which is applied to the above-mentioned airport supply chain delivery optimization method for dynamic demand and includes: a multi-source data acquisition and fusion module for real-time acquisition of multi-source flight data and external information data, data preprocessing of the multi-source flight data and the external information data, evaluation of the alignment degree of the data sources based on the preprocessed multi-source flight data, data correction operation according to the data source alignment degree evaluation result, and identification of a high-confidence data set; a demand dynamic prediction module for constructing a multi-factor feature set through the external information data, predicting a meal demand amount in combination with the multi-source flight data in the high-confidence data set, and implementing a prediction feedback optimization mechanism according to the meal demand amount prediction result; an inventory management and replenishment strategy module for predicting a meal replenishment period based on the multi-source flight data, predicting an inventory replenishment amount within the meal replenishment period in combination with the meal demand amount prediction result, and implementing a replenishment strategy according to the inventory replenishment amount prediction result; and a delivery scheduling and path optimization module for fusing the meal demand amount prediction result and the inventory replenishment amount prediction result, allocating delivery resources and generating a delivery path, and simultaneously performing visual feedback analysis.

[0065] In the embodiment, through modular integration of multi-source data acquisition and fusion, demand dynamic prediction, inventory management and replenishment strategy, distribution scheduling and path optimization core functions, the full-process intelligent optimization of the airport supply chain is realized. The modules are coordinated and linked, not only ensuring high consistency and high reliability of the data, but also being able to complete meal prediction, inventory replenishment and resource scheduling in real time and efficiently in response to dynamic demand changes. Multi-factor feature modeling and feedback optimization are supported, with self-adaptive learning and continuous evolution capabilities, greatly improving the scientificity, intelligence and safety of the supply chain operation, effectively preventing shortages and waste, and providing strong technical support and decision basis for airport operation management and service guarantee.

[0066] It should be noted that, in this document, the terms such as first and second are used merely to distinguish one entity or action from another, and do not necessarily require or imply that these entities or actions exist in any actual relationship or order. Moreover, the terms "include", "contain" or any other variants thereof are intended to cover non-exclusive inclusion, so that the process, method, article or equipment including a series of elements not only includes those elements, but also includes other elements not explicitly listed or inherent to such process, method, article or equipment.

[0067] The preferred embodiments of the application disclosed above are only used to help explain the application. The preferred embodiments do not describe all the details and limit the application to the specific embodiments described. As understood by those skilled in the art, many modifications and variations can be made according to the content of the specification. The specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the application, so that those skilled in the art can well understand and utilize the application. The application is limited only by the claims and their full scope and equivalents.

Claims

1. An airport supply chain distribution optimization method oriented towards dynamic demand, characterized in that, Includes the following steps: S1 collects multi-source flight data and external information data in real time, performs data preprocessing on the multi-source flight data and external information data, evaluates the alignment of the data sources based on the preprocessed multi-source flight data, and performs data correction operations based on the data source alignment evaluation results to identify high-confidence datasets. S2 constructs a multi-factor feature set through external information data and combines it with multi-source flight data in a high-confidence dataset to predict the demand for meals. Based on the prediction results of the demand for meals, a prediction feedback optimization mechanism is implemented. The specific process of constructing a multi-factor feature set using external information data is as follows: Collect the actual food demand for each day in history and the weather data of that day to construct a weather impact dataset. By training the weather impact dataset with multiple linear regression, fit the statistical relationship between weather data and changes in actual food demand, construct a theoretical demand prediction model, and output the theoretical food demand due to weather on that day. Obtain the actual food demand under normal weather conditions within the historical sliding time window, calculate the mean to obtain the average food demand, divide the difference between the theoretical food demand and the average food demand by the average food demand to obtain the relative change rate of food demand affected by weather, and normalize it as the weather impact value. The system synchronously acquires event information and actual food demand for each day in history. Within the same time window, it counts the actual food demand for days with events and calculates the average food demand for days with events. Simultaneously, it counts the actual food demand for days without events and calculates the average food demand for days without events. The difference between the average food demand for days with events and the average food demand for days without events is divided by the average food demand for days without events to obtain the relative rate of change in food demand affected by the event. This rate is then normalized and used as the event impact value. The specific process of predicting meal demand by combining multi-source flight data from a high-reliability dataset is as follows: The number of passenger bookings, passenger rebookings, and passenger refunds for flight f at time t are obtained from the high-reliability dataset. The number of passenger rebookings and passenger refunds is subtracted from the number of passenger bookings to obtain the estimated actual number of passengers boarding. The actual number of meals required and the number of passengers boarding for the same flight f during the same period are obtained from the historical high-reliability dataset from the supply chain distribution database. The historical average consumption per passenger is obtained by calculating the ratio of the actual number of meals required to the number of passengers boarding. The historical average consumption per passenger for the same flight f within the sliding time window is statistically analyzed, and the mean value is calculated to obtain the historical average consumption per passenger. At the same time, the standard deviation is calculated to obtain the historical average consumption per passenger standard deviation. The historical average per capita consumption is obtained by subtracting the historical average per capita consumption from the historical average per capita consumption. The standard per capita consumption offset is obtained by dividing the historical average per capita consumption offset by the historical average per capita consumption standard deviation. The consumption offset smoothing correction value is obtained by performing a hyperbolic tangent operation on the standard per capita consumption offset value. The historical correction weighting factor is multiplied by the consumption offset smoothing correction value to obtain the historical correction term. The weather impact value and the event impact value are obtained. The weather impact weighting factor is multiplied by the weather impact value to obtain the weather correction term. The event impact weighting factor is multiplied by the event impact value to obtain the event correction term. The sum of the historical correction term, the weather correction term, and the event correction term is added to a constant one to obtain the comprehensive correction coefficient. The comprehensive correction coefficient is calculated and multiplied by the expected actual number of passengers to obtain the predicted food demand value. S3, based on multi-source flight data, predicts the meal replenishment cycle, combines the meal demand forecast results, predicts the inventory replenishment quantity during the meal replenishment cycle, and implements replenishment strategies based on the inventory replenishment quantity forecast results; S4 integrates the forecast results of food demand and inventory replenishment, allocates delivery resources and generates delivery routes, and performs visual feedback analysis.

2. The airport supply chain distribution optimization method for dynamic demand as described in claim 1, characterized in that, The specific process of real-time acquisition of multi-source flight data and external information data, and data preprocessing of multi-source flight data and external information data is as follows: Real-time collection of multi-source flight data, including flight schedules, number of passenger bookings, number of passenger rebookings, number of passenger refunds, number of passenger boardings, food inventory, and actual food demand; synchronous real-time collection of external information data that affects delivery demand, including weather data and event information, and the use of an event-driven mechanism to trigger real-time updates of external information data; The system performs unified format conversion, time synchronization, field mapping, and deduplication on multi-source flight data. It also completes missing data through interpolation and removes extreme abnormal data. In addition, it normalizes the multi-source flight data. A supply chain distribution database is built to store the pre-processed multi-source flight data through multi-table association and label normalization, and external information data of the same period is recorded synchronously.

3. The airport supply chain distribution optimization method for dynamic demand as described in claim 1, characterized in that, The specific process for evaluating the alignment of data sources based on preprocessed multi-source flight data is as follows: The system counts the number of data sources for multi-source flight data related to the same flight and the same object collected from various business interfaces, including the number of passenger bookings from the flight management interface, the number of passengers boarding from the gate management interface, and the actual number of meals required from the warehouse management interface; it also obtains the actual value of each multi-source flight data; it sums up the actual values ​​of all multi-source flight data and calculates the average value of the multi-source flight data. For each data source, the difference between the actual value of the multi-source flight data and the mean of the multi-source flight data is calculated to obtain the data source difference. The relative deviation of the data source is obtained by dividing the difference of the data source by the sum of the mean of the multi-source flight data and the minimum constant term, and taking the absolute value. The minimum constant term is the product of the constant 0.01 and the mean of the multi-source flight data. Based on the number of data sources, the relative deviations of the data sources of each data source are accumulated and the average value is calculated to obtain the average deviation of the data sources. The square root of the average deviation of the data sources is calculated and added to the constant 1. Then, the natural logarithm is calculated. The result of the natural logarithm calculation is added to the constant 1 and the reciprocal is taken to obtain the data source alignment evaluation value.

4. The airport supply chain distribution optimization method for dynamic demand as described in claim 1, characterized in that, The specific process of performing data correction operations based on the data source alignment evaluation results to identify high-confidence datasets is as follows: For multi-source flight data datasets with data source alignment assessment values ​​below the consistency threshold, mark them as abnormal datasets requiring review. Record the actual values ​​of the original multi-source flight data, the average value of the multi-source flight data, and the relative deviation of the data source for all business interfaces. Identify and label the interfaces with the largest relative deviation of the data source. At the same time, perform data correction operations: conduct data source investigation and supplement multi-source flight data collection, and prompt the interface responsible person to check the data link and source end status. If three consecutive data collection attempts fail, the abnormal interface will be temporarily blocked. For short-term missing and abnormal multi-source flight data, trend inference and multi-source flight data imputation are performed by sliding window linear regression and K-nearest neighbor imputation. Until the data source alignment assessment value is greater than or equal to the consistency threshold; For multi-source flight data datasets with a data source alignment evaluation value greater than or equal to the consistency threshold, they are marked as high-confidence datasets and are given priority for inclusion in the next process. All abnormal datasets requiring review, data correction operations, and data source alignment assessment values ​​are synchronously written into the supply chain distribution database. Consistency thresholds are continuously optimized, and business interfaces are upgraded through regular reviews.

5. The airport supply chain distribution optimization method for dynamic demand as described in claim 1, characterized in that, The specific process of implementing the prediction feedback optimization mechanism based on the predicted food demand is as follows: The forecast values ​​of meal demand for each flight and time period are written into the supply chain distribution database and simultaneously pushed to the inventory management and replenishment strategy module to drive meal replenishment and distribution resource scheduling. Whenever actual meal delivery and consumption are completed, the actual meal demand is collected and compared with the predicted meal demand, and the prediction error is recorded and analyzed in real time. Based on the accumulated prediction error data, the theoretical demand prediction model and the meal demand prediction algorithm are retrained periodically, and the prediction parameters are adjusted. If the deviation between the consecutive meal demand predictions and the actual meal demand is found to be higher than the deviation threshold, an anomaly warning is triggered, prompting manual review and data backtracking.

6. The airport supply chain distribution optimization method for dynamic demand as described in claim 1, characterized in that, The specific process of estimating the meal replenishment cycle based on multi-source flight data, comprehensively considering the meal demand forecast, predicting the inventory replenishment quantity within the meal replenishment cycle, and implementing the replenishment strategy based on the inventory replenishment quantity forecast is as follows: Real-time acquisition of food inventory, recording of food consumption cycles, calculation and statistics of food consumption rates, identification of near-expiration and shortage risks, and prediction of food replenishment cycles; Obtain the food demand forecast value within the replenishment cycle, sum up the food demand forecast values ​​for each time period to obtain the total demand forecast value within the replenishment cycle; at the same time, calculate the standard deviation of the food demand forecast value within the replenishment cycle to obtain the demand forecast standard deviation. Obtain the current food inventory level. Subtract the current food inventory level from the total demand forecast within the replenishment cycle to obtain the theoretical replenishment demand. Multiply the standard deviation of the demand forecast by the volatility weighting factor to obtain the compensation safety stock. Add the compensation safety stock to the theoretical replenishment demand to obtain the inventory replenishment forecast. Perform a maximum function operation on the inventory replenishment forecast, i.e., take the maximum value between the inventory replenishment forecast and 0. If the inventory replenishment forecast is negative, the inventory replenishment forecast is 0; otherwise, the inventory replenishment forecast is the actual calculated value. Implement replenishment strategies based on inventory replenishment forecasts: generate replenishment work orders and push them to the purchasing and warehousing departments in real time; The actual replenishment and inventory consumption process is dynamically monitored. The actual food demand and food inventory are collected regularly and compared with the food demand forecast and inventory replenishment forecast. The shortage and backlog situation is statistically analyzed, and adjustments are made to delivery resources and food warehousing. Periodically backtest the replenishment performance brought about by different volatility weighting factors, adjust and optimize volatility weighting factors and replenishment parameters to achieve self-learning and continuous evolution of replenishment strategies; at the same time, record and archive all replenishment results, fluctuations in food inventory and parameter optimizations in real time.

7. The airport supply chain distribution optimization method for dynamic demand as described in claim 1, characterized in that, The specific process of integrating the predicted food demand and inventory replenishment results, allocating delivery resources and generating delivery routes, while simultaneously performing visual feedback analysis, is as follows: Based on the forecasts of food demand and inventory replenishment, various food delivery needs are broken down into specific delivery task orders, specifying the corresponding flights, time nodes, and food categories. Based on the available resources such as vehicles, delivery personnel, delivery time windows, and cold chain support, the optimal delivery resources are allocated, and the business constraints of flight priority, delivery distance, and food preservation are fully considered to achieve scientific scheduling. By combining real-time airport road, traffic and weather information, the shortest path algorithm is used to determine the optimal route for each batch of delivery tasks and supports the merging of tasks from multiple flights in the same area. Delivery tasks are issued and monitored and located throughout the delivery process. In case of emergencies, rescheduling is triggered in real time to ensure that delivery tasks are completed on demand and all anomalies in the delivery process are pushed out simultaneously. At the same time, it dynamically displays the status of each delivery task, the allocation of delivery resources, the real-time location distribution of delivery vehicles and personnel, the optimal route trajectory, as well as key node warnings and historical delivery anomaly heatmaps.

8. An airport supply chain distribution optimization system oriented towards dynamic demand, employing the airport supply chain distribution optimization method oriented towards dynamic demand as described in any one of claims 1-7, characterized in that, include: The multi-source data acquisition and fusion module is used to acquire multi-source flight data and external information data in real time, perform data preprocessing on the multi-source flight data and external information data, evaluate the alignment of the data sources based on the preprocessed multi-source flight data, and perform data correction operations based on the data source alignment evaluation results to identify high-confidence datasets. The dynamic demand forecasting module is used to construct a multi-factor feature set through external information data and combine it with multi-source flight data in a high-confidence dataset to predict the demand for meals. Based on the forecast results, a prediction feedback optimization mechanism is implemented. The inventory management and replenishment strategy module is used to predict the meal replenishment cycle based on multi-source flight data, combine the meal demand forecast results, predict the inventory replenishment quantity within the meal replenishment cycle, and implement replenishment strategies based on the inventory replenishment quantity forecast results. The delivery scheduling and route optimization module is used to integrate the forecast results of food demand and inventory replenishment, allocate delivery resources and generate delivery routes, and perform visual feedback analysis.

Images