Flight delay dynamic prediction method and system based on multi-source data fusion
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- BEIJING DEXUN AVIATION SERVICE CO LTD
- Filing Date
- 2026-05-27
- Publication Date
- 2026-06-23
AI Technical Summary
In traditional flight delay dynamic prediction methods based on multi-source data fusion, the fixed-parameter EMA model cannot capture the sudden changes in the relationships between multiple flight data sources, resulting in reduced prediction accuracy.
By acquiring real-time data such as flight position, speed, estimated arrival time, landing airspace traffic, and runway occupancy, the attenuation factor in the EMA model is adjusted. Combined with indicators such as position deviation coefficient, flight status variability, airspace traffic anomaly, and runway overload, the parameters of the EMA model are dynamically adjusted to adapt to the actual flight movement.
It improves the accuracy of flight delay prediction, can more accurately capture short-term dynamic changes in flights, enhances the EMA model's ability to respond to sudden factors, and avoids damage to prediction accuracy.
Smart Images
Figure CN122264233A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of multi-source data fusion prediction technology, specifically to a method and system for dynamic prediction of flight delays based on multi-source data fusion. Background Technology
[0002] With the continued growth of global air transport and the rapid increase in the number of flights, flight delays have become a significant challenge for the aviation industry. Flight delays not only significantly increase airlines' operating costs in various aspects but also exacerbate air traffic congestion risks, further straining airport resources such as parking spaces, baggage handling, and gate arrangements. Dynamic flight delay prediction methods based on multi-source data fusion can integrate information from different data sources and achieve dynamic prediction of flight delays through correlation analysis, helping to improve the data governance and intelligent operation capabilities of airlines and airports.
[0003] The Exponential Moving Average (EMA) model can effectively cope with the impact of sudden factors in the flight delay prediction process, balancing noise and trends in the data, thus ensuring the robustness of the prediction results in the process of predicting flight delays from multiple sources of data that are complex and ever-changing. However, in the process of dynamic prediction of flight delays for multi-source data fusion, the fixed-parameter EMA model cannot capture the sudden changes in the multi-source flight data and is easily affected by the complex changes in the multi-source data, misjudging sudden situations as short-term noise, thereby reducing the accuracy of the EMA model in the dynamic prediction of flight delays. Summary of the Invention
[0004] In light of the above, it is necessary to provide a method and system for dynamic prediction of flight delays based on multi-source data fusion. Compared with traditional methods for dynamic prediction of flight delays based on multi-source data fusion, this method improves the prediction accuracy of the EMA model for flight delays. In a first aspect, embodiments of this application provide a method for dynamic prediction of flight delays based on multi-source data fusion, the method comprising the following steps: The system can obtain the location, altitude, speed, and estimated arrival time of the flight to be predicted in real time, as well as the landing airspace traffic and runway occupancy rate of the airport where the flight is to be predicted. An adjustment range for the attenuation factor in the EMA model used to predict the estimated arrival time is preset. The position deviation coefficient of the flight to be predicted in each interval is obtained by observing the changing trend and dispersion of the position deviation of the flight to be predicted from the preset route in each interval. By comparing the altitude distribution range of the flight to be predicted between each interval and its adjacent intervals, and combining the energy distribution of the speed of the flight to be predicted in the frequency domain in each interval, the flight state change degree of the flight to be predicted in each interval is obtained. This is then fused with the position deviation coefficient to obtain the flight deviation degree of the flight to be predicted in each interval. By analyzing the differences between the landing airspace traffic and the standard landing airspace traffic in each interval, as well as the changing trends of the landing airspace traffic, the airspace traffic anomaly degree of the flights to be predicted in each interval is obtained. Combined with the average level and attenuation of runway occupancy rate in each interval, the runway high load degree of the flights to be predicted in each interval is obtained, and the load coefficient of the flights to be predicted in each interval is obtained. Then, combined with the flight deviation degree, the attenuation factor adjustment coefficient of the flights to be predicted in each interval is obtained, so as to adjust the attenuation factor in each interval, thereby predicting the delay of the flights to be predicted.
[0005] In one embodiment, the process of obtaining the position deviation coefficient is as follows: Obtain the trend term of the deviation of the position of the flight to be predicted from the preset route in each interval in the time series, and calculate the Hurst exponent of the trend term; The dispersion of the position of the flight to be predicted within each interval relative to the preset route is denoted as the deviation dispersion. The position deviation coefficient is positively correlated with the Hurst exponent and the deviation dispersion, respectively.
[0006] In one embodiment, the process of obtaining the degree of change in flight state is as follows: Calculate the interquartile range of the altitude of the flight to be predicted in each interval, and record the difference between the interquartile range in each interval and the interquartile range in the adjacent preceding interval as the range difference. Obtain the frequency components of the speed of the flight to be predicted in the frequency domain within each interval, count the frequency with the highest energy in each frequency component, obtain the segmentation threshold of all frequencies counted in each interval, and record the frequency components that are greater than the segmentation threshold in each interval as high-frequency components, and record the frequency components other than high-frequency components in each interval as low-frequency components; calculate the ratio of the energy accumulation result of all high-frequency components in each interval to the energy accumulation result of all low-frequency components. The degree of change in flight state is positively correlated with the range difference and the ratio, respectively.
[0007] In one embodiment, the flight deviation is the normalized value of the product of the position deviation coefficient and the flight state change.
[0008] In one embodiment, the process of obtaining the airspace traffic anomaly degree is as follows: The difference between the average landing airspace flow in each interval and the standard landing airspace flow is denoted as the flow difference. Obtain trend test statistics for landing airspace traffic within each interval; The airspace traffic anomaly is positively correlated with the absolute value of the traffic difference and the absolute value of the trend test statistic, respectively.
[0009] In one embodiment, the process of obtaining the high load index of the runway is as follows: Calculate the average runway occupancy rate within each interval; obtain the abrupt change points of runway occupancy rate in time series within each interval; obtain the fitted straight line of all abrupt change points in time series within each interval. When the slope of the fitted line is non-negative, the high load of the runway is positively correlated with the average value and the slope of the fitted line, respectively; when the slope of the fitted line is negative, the high load of the runway is inversely proportional to the absolute value of the slope of the fitted line.
[0010] In one embodiment, the load factor is a normalized value of the product of the airspace traffic anomaly and the runway high load.
[0011] In one embodiment, the attenuation factor adjustment coefficient is the product of the flight deviation and the load factor.
[0012] In one embodiment, adjusting the attenuation factor within each interval to predict flight delays includes: Based on the estimated arrival time, location, altitude, speed, landing airspace flow, and runway occupancy rate of each interval and all previous times, the EMA model is used to predict the estimated arrival time of the next time moment in each interval. Specifically, a threshold is set for the attenuation factor adjustment coefficient of a preset number of delayed flights during their flight process. If the attenuation factor adjustment coefficient of the flight to be predicted in each interval is greater than the threshold, the attenuation factor in the EMA model is adjusted; otherwise, the attenuation factor in the EMA model is not adjusted. The expression for adjusting the attenuation factor in the EMA model is as follows: In the formula, This represents the adjusted attenuation factor in the EMA model within the i-th interval; This represents the preset initial value of the decay factor in the EMA model; represents the attenuation factor adjustment coefficient for the flight to be predicted in the i-th interval; norm() is the normalization function.
[0013] Secondly, embodiments of this application also provide a flight delay dynamic prediction system based on multi-source data fusion, including a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements the steps of any of the above-described flight delay dynamic prediction methods based on multi-source data fusion.
[0014] This application has at least the following beneficial effects: This application quantifies the degree and trend of a flight's deviation from a preset route during flight by calculating the position deviation coefficient. This reflects whether the flight is affected by unforeseen factors, transforming complex flight trajectory information into concise numerical features for easier subsequent analysis. By calculating the flight state change rate and comprehensively considering the frequency domain characteristics of the flight's altitude and speed, it can more accurately reflect the changes in the flight state. By fusing the position deviation coefficient and the flight state change rate, the flight deviation rate is obtained, which comprehensively reflects the degree of impact of unforeseen factors on the flight in the short term. This provides more accurate short-term deviation information for the EMA model, helping it to more accurately capture the short-term dynamic changes of the flight and thus improve the prediction accuracy of flight delays. Furthermore, by calculating the airspace traffic anomaly and runway high load, and combining them to obtain the load coefficient, the utilization of airport airspace and runway resources is comprehensively considered, which can more accurately assess the comprehensive impact of airport airspace and ground factors on flight delays. Furthermore, by combining flight deviation with load factor, an attenuation factor adjustment coefficient is obtained, which can comprehensively reflect the degree of influence of various factors on the flight to be predicted. By adjusting the attenuation factor through the attenuation factor adjustment coefficient, the parameters of the EMA model can be adapted to the actual motion of the flight to be predicted, thereby improving the prediction accuracy of the EMA model for flight delays. Attached Figure Description
[0015] To more clearly illustrate the technical solutions and advantages in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0016] Figure 1 A flowchart illustrating the steps of a flight delay dynamic prediction method based on multi-source data fusion, as provided in one embodiment of this application; Figure 2 This is a schematic diagram illustrating the process of obtaining the attenuation factor adjustment coefficient. Detailed Implementation
[0017] In the description of the embodiments in this application, the words "exemplary," "or," and "for example" are used to indicate examples, illustrations, or descriptions. Any embodiment or design scheme described as "exemplary" or "for example" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design schemes. Specifically, the use of the words "exemplary," "or," and "for example" is intended to present the relevant concepts in a specific manner.
[0018] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. It should be understood that, unless otherwise stated, " / " in this application means "or".
[0019] It should also be noted that the terms "first" and "second" in this application are used to distinguish similar objects, rather than to describe a specific order or sequence.
[0020] The following description, in conjunction with the accompanying drawings, details the specific scheme of the flight delay dynamic prediction method and system based on multi-source data fusion provided in this application.
[0021] Please see Figure 1 The diagram illustrates a flowchart of a flight delay dynamic prediction method based on multi-source data fusion provided in an embodiment of this application. The method includes the following steps: Step 1: Obtain the real-time location, altitude, speed, estimated arrival time of the flight to be predicted, as well as the landing airspace traffic and runway occupancy rate of the airport where the flight will land.
[0022] The system utilizes the Aircraft Communications Addressing and Reporting System (ACARS) to access the fielded flight plan message (FPL) in real time to obtain the preset route and initial estimated arrival time of the flight to be predicted, and to acquire the estimated arrival time of the flight at each time point in real time. The estimated arrival time refers to the time when the flight is expected to arrive at its destination. This is based on data from the Automatic Dependent Surveillance-Broadcasting System (ADSB). (ADS-B surveillance) acquires the real-time position, altitude, and speed of the flight to be predicted, where position refers to the latitude and longitude coordinates of the flight. Based on the airport information integration system, it acquires the landing airspace traffic and runway occupancy rate of the landing airport for the flight to be predicted in real time. Landing airspace traffic refers to the number of aircraft in the landing airspace of the landing airport at each time point, and runway occupancy rate refers to the percentage of unusable runways to the total number of runways at the landing airport at each time point. When any runway in the landing airport becomes unusable due to weather conditions or other unforeseen circumstances, or due to flight takeoffs and landings, that runway is marked as unusable, and the remaining runways are marked as usable. GPS clocks are used to unify the timestamps of multi-source flight data.
[0023] In this embodiment, the acquisition frequency for the estimated arrival time, location, altitude, speed, landing airspace traffic and runway occupancy rate is set to 1Hz. The acquisition frequency value is preset by the user and can be set by the implementer according to the actual situation. This application does not impose any special restrictions.
[0024] Furthermore, to prevent data loss due to network fluctuations during the acquisition of multi-source flight data, a data filling method was used to fill in the missing data in each type of data. At the same time, to avoid the impact of different units of measurement on subsequent analysis, the data after filling were normalized.
[0025] In this embodiment, cubic spline interpolation is used to fill in missing data, and Min-Max normalization is used to normalize various data. Both cubic spline interpolation and Min-Max normalization are well-known techniques and will not be elaborated upon here. As other implementation methods, implementers may use other existing feasible techniques to achieve missing data filling and data normalization; this application does not impose any special restrictions. In this application, unless otherwise specified, Min-Max normalization is used for normalization.
[0026] Step 2: Preset the adjustment range of the attenuation factor in the EMA model used to predict the expected arrival time. By observing the changing trend and dispersion of the position deviation of the flight to be predicted from the preset route in each range, obtain the position deviation coefficient of the flight to be predicted in each range. By comparing the altitude distribution range of the flight to be predicted between each range and its adjacent ranges, and combining the energy distribution of the speed of the flight to be predicted in the frequency domain in each range, obtain the flight state change degree of the flight to be predicted in each range. Then, fuse it with the position deviation coefficient to obtain the flight deviation degree of the flight to be predicted in each range.
[0027] In the process of using the EMA model to achieve dynamic prediction of flight delays through multi-source data fusion, the preset route is the optimal flight plan formulated by the airline based on factors such as the shortest path, fuel efficiency, and air traffic control restrictions. Therefore, the deviation between the actual position of the flight and the preset route is a key real-time indicator reflecting the air traffic situation. The deviation of the actual position of the flight from the preset route may be caused by a variety of factors such as sudden severe weather and emergency adjustments to air traffic. In order to consider both the stability of historical data and the rapid response to recent route deviations in the EMA model, it is necessary to dynamically adjust the attenuation factor in the EMA model so that it can change flexibly according to the actual situation.
[0028] Specifically, in the process of using the EMA model to achieve dynamic prediction of flight delays through multi-source data fusion, the more severe the impact of unforeseen circumstances on the flight to be predicted, the more severe the short-term deviation between the actual route and the preset route. That is, the deviation distance between the actual position of the flight and the preset route during flight shows a more obvious trend, and the dispersion of the position deviation is more pronounced. Simultaneously, if the position deviation of the flight to be predicted is caused by sudden weather interference, high-altitude jet streams, or abnormal engine thrust output during flight, the altitude of the flight to be predicted will change abruptly, resulting in a larger range of altitude variations. Furthermore, the speed will be affected by high-frequency maneuvers, making the proportion of high-frequency energy more significant. High-frequency maneuvers refer to frequent and rapid adjustments to flight attitude or speed made by the aircraft within a short period. These adjustments are usually made to cope with unforeseen circumstances, such as avoiding severe weather, avoiding air conflicts, or adjusting flight speed. In this case, the attenuation factor in the EMA model should be increased to avoid a lag in the response to sudden disturbances during flight delay prediction, which could compromise prediction accuracy.
[0029] Based on the above analysis, an adjustment range for the attenuation factor in the EMA model used to predict the estimated arrival time is preset, and this range is sliding. The position deviation coefficient of the flight to be predicted within each range is obtained by analyzing the changing trend and dispersion of the flight's position deviation from the preset route. The specific process is as follows: Obtain the trend term of the deviation of the position of the flight to be predicted from the preset route in each interval over time, and calculate the Hurst exponent of the trend term; denote the dispersion of the deviation of the position of the flight to be predicted from the preset route in each interval as the deviation dispersion. The position deviation coefficient of the flight to be predicted within each interval is positively correlated with the Hurst exponent and the deviation dispersion, respectively. The calculation of the Hurst exponent is a well-known technique and will not be elaborated upon in this application.
[0030] It should be noted that dispersion refers to the degree of unevenness in the distribution of data, which can be achieved by calculating variance, standard deviation, coefficient of variation, etc. This application does not impose any special restrictions on it.
[0031] It should be noted that positive correlation means that the variables change in the same direction; when one variable increases, the other variable also increases, and when one variable decreases, the other variable also decreases.
[0032] In this embodiment, the length of the interval is 20 minutes and the sliding step size of the interval is 1 second. The length of the interval and the sliding step size are preset by the user. The implementer can set them according to the actual situation. This application does not impose any special restrictions.
[0033] In this embodiment, the deviation of the position of the flight to be predicted from the preset route is specifically: the Euclidean distance between the preset position of the flight to be predicted on the preset route and the actual position of the flight to be predicted at each time.
[0034] In this embodiment, the STL (Seasonal and Trend decomposition using Loess) algorithm is used to obtain the trend item. The STL algorithm is a well-known technology and will not be described in detail in this application. As other implementation methods, based on the ability to obtain the trend item, the implementer may use other existing technologies, such as X-12-ARIMA, etc. This application does not impose any special restrictions.
[0035] In this embodiment, the dispersion of the deviation of the position of the flight to be predicted within each interval from the preset route is the coefficient of variation.
[0036] In this embodiment, the Hurst exponent and the deviation dispersion are mapped to a first positive number and a second positive number, respectively. The product of the first positive number and the second positive number is used as the position deviation coefficient of the flight to be predicted in each interval. The purpose of mapping the Hurst exponent and the deviation dispersion to positive numbers is to avoid the situation where the calculated result of the position deviation coefficient is forced to be 0 when the Hurst exponent or the deviation dispersion is 0. In this application, unless otherwise specified, the purpose of mapping the data to positive numbers is achieved by calculating the sum of the data and a preset value greater than 0. The preset value greater than 0 is 0.01, and the value of the preset value greater than 0 is preset by the user. The implementer can set it according to the actual situation. There are many methods to map the data to positive numbers. The implementer can choose other existing feasible methods. This application does not impose any special restrictions on this.
[0037] In another embodiment, the mean of the Hurst exponent and the deviation dispersion is used as the position deviation coefficient of the flight to be predicted in each interval.
[0038] It should be noted that the position deviation coefficient reflects the long-term trend and distribution of the deviation of the flight to be predicted from the preset route in each interval. When using the EMA model to dynamically predict flight delays, if the flight to be predicted is affected by unforeseen circumstances that increase the risk of delay, the deviation between its actual position and the preset route will be more obvious and more discrete, and the calculated position deviation coefficient will increase.
[0039] Furthermore, by comparing the altitude distribution range of the flights to be predicted between each interval and its adjacent intervals, and combining this with the energy distribution of the speed of the flights to be predicted in the frequency domain within each interval, the degree of change in the flight state of the flights to be predicted within each interval is obtained. The specific process is as follows: Calculate the interquartile range of the altitude of the flight to be predicted in each interval, and record the difference between the interquartile range in each interval and the interquartile range in the adjacent preceding interval as the range difference. Obtain the frequency components of the speed of the flight to be predicted in the frequency domain within each interval, count the frequency with the highest energy in each frequency component, obtain the segmentation threshold of all frequencies counted in each interval, and record the frequency components that are greater than the segmentation threshold in each interval as high-frequency components, and record the frequency components other than high-frequency components in each interval as low-frequency components; calculate the ratio of the energy accumulation result of all high-frequency components in each interval to the energy accumulation result of all low-frequency components. The degree of change in the flight status of the flight to be predicted within each interval is positively correlated with the range difference and the ratio, respectively. It should be noted that for the first interval, the range difference is set to 0.
[0040] In this embodiment, Fourier transform is used to obtain the frequency spectrum of the speed of the flight to be predicted in each interval, and based on the frequency spectrum, Hilbert transform is used to obtain the frequency components of the speed of the flight to be predicted in each interval in the frequency domain. Both Fourier transform and Hilbert transform are well-known technologies and will not be described in detail in this application. As other implementation methods, based on the ability to obtain the frequency components of the speed of the flight to be predicted in each interval in the frequency domain, the implementer may use other existing feasible technologies, and this application does not impose any special restrictions.
[0041] In this embodiment, the Otsu threshold segmentation algorithm is used to obtain the frequency segmentation threshold. The Otsu threshold segmentation algorithm is a well-known technology and will not be described in detail in this application. As other implementation methods, based on the ability to obtain the frequency segmentation threshold, implementers may use other existing technologies, such as iterative threshold segmentation, global threshold segmentation, etc. This application does not impose any special restrictions.
[0042] In this embodiment, the range difference is used as the exponent of an exponential function with a base greater than 1. The product of the calculated result of the exponential function and the ratio is used as the degree of change in the flight status of the flight to be predicted within each interval. The exponential function is used to map the range difference to a positive number. The value of the value greater than 1 is preset by the user. The implementer can set it according to the actual situation. In this embodiment, the value of the value greater than 1 is preset to be a natural constant.
[0043] It should be noted that: the flight state change degree characterizes the extent to which the altitude change range expands and the proportion of high-frequency energy in the speed when the flight to be predicted makes high-frequency maneuvering adjustments due to unforeseen factors in each interval. When using the EMA model to dynamically predict flight delays, if the flight to be predicted is affected by unforeseen circumstances, leading to an increased risk of delay, the altitude change range of the flight to be predicted will increase, and the proportion of high-frequency maneuvering frequency components in the speed will also increase. At this time, the calculated flight state change degree will increase.
[0044] Furthermore, by combining the position deviation coefficient and flight status change degree of the flight to be predicted within each interval, the flight deviation degree of the flight to be predicted within each interval is obtained, expressed as: In the formula, This represents the flight deviation degree of the flight to be predicted within the i-th interval; This represents the position deviation coefficient of the flight to be predicted within the i-th interval; This represents the degree of change in the flight status of the flight to be predicted within the i-th interval; norm() is the normalization function.
[0045] It should be noted that the flight deviation rate comprehensively reflects the dispersion of the position deviation trend and the degree of flight status amplification of the flight to be predicted due to unforeseen factors within each interval. When using the EMA model for dynamic prediction of flight delays, if the flight to be predicted is affected by unforeseen circumstances leading to an increased risk of delay, both the calculated position deviation coefficient and the degree of flight status change will increase, and the calculated flight deviation rate will also increase. In this case, it is necessary to increase the attenuation factor in the EMA model to enhance the EMA model's responsiveness to recent unforeseen changes, thereby ensuring prediction accuracy.
[0046] Step 3: By analyzing the difference between the landing airspace traffic and the standard landing airspace traffic in each interval, as well as the changing trend of the landing airspace traffic, the airspace traffic anomaly degree of the flights to be predicted in each interval is obtained. Combined with the average level and attenuation of the runway occupancy rate in each interval, the runway high load degree of the flights to be predicted in each interval is obtained, and the load factor of the flights to be predicted in each interval is obtained.
[0047] In the process of using the EMA model to achieve dynamic prediction of flight delays through multi-source data fusion, flight delays not only depend on real-time dynamic factors during aircraft flight, but also on landing airspace traffic and runway occupancy rates at the landing airport, which are fundamental causes and prerequisite constraints for flight delays. Simply adjusting the attenuation factor in the EMA model based solely on the flight deviation of the flight to be predicted in each interval still has certain drawbacks. Specifically, if the EMA model only focuses on the in-flight status and ignores the impact of ground systems on flight delays, it falls into the trap of a "limited aerial perspective." Sudden and cumulative events in the ground system can significantly impact flight delays, but because the EMA model does not consider these factors, it cannot comprehensively predict flight delays, thus reducing prediction accuracy.
[0048] Specifically, in the process of using the EMA model to achieve dynamic prediction of flight delays through multi-source data fusion, the more severely airport ground resources are constrained by sudden and cumulative events, the greater the difference between the average level of real-time landing airspace traffic and the standard landing airspace traffic. Furthermore, sudden severe weather or air traffic congestion causes landing airspace traffic to exhibit a strong trend. Simultaneously, runway occupancy rates remain high over time, and the rate of decay is slow. In this situation, the decay factor in the EMA model should be increased to improve the model's response to flight delay risks caused by recent sudden and cumulative ground system events, avoiding a "limited air view" in flight delay prediction that ignores the impact of ground systems and reduces prediction accuracy. The standard landing airspace traffic is updated in real-time. The process of obtaining the standard landing airspace traffic within any interval is as follows: select a preset number of adjacent intervals, and use the average of all landing airspace traffic within all selected intervals as the standard landing airspace traffic within that interval.
[0049] In this embodiment, the preset quantity is 100. The preset quantity is preset by the user and can be set by the implementer according to the actual situation. This application does not impose any special restrictions. If there is insufficient data in the process of calculating the standard landing airspace traffic, the mean filling method is used to fill the missing data. The mean filling method is a well-known technology and will not be described in detail in this application. The implementer can choose other existing feasible methods to fill the missing data. This application does not impose any special restrictions.
[0050] Based on the above analysis, the airspace traffic anomaly of the flights to be predicted in each interval is obtained by analyzing the difference between the landing airspace traffic and the standard landing airspace traffic in each interval, as well as the changing trend of the landing airspace traffic. The specific process is as follows: The difference between the mean landing airspace flow in each interval and the standard landing airspace flow is recorded as the flow difference; the trend test statistic of the landing airspace flow in each interval is obtained. The airspace traffic anomaly degree of the flights to be predicted in each interval is positively correlated with the absolute value of the traffic difference and the absolute value of the trend test statistic.
[0051] In this embodiment, the trend test statistic is the statistic Z in the Mann-Kendall trend test algorithm. The Mann-Kendall trend test algorithm is a well-known technology and will not be described in detail here. As other implementation methods, based on the ability to measure the changing trend of landing airspace traffic in each interval, implementers may use other existing technologies, such as the slope method, Cox-Stuart test, etc. This application does not impose any special restrictions.
[0052] In this embodiment, the absolute values of the flow difference and the trend test statistic are mapped to a third positive number and a fourth positive number, respectively. The product of the third positive number and the fourth positive number is used as the airspace flow anomaly degree for the flight to be predicted in each interval. The purpose of mapping the absolute values of the flow difference and the trend test statistic to positive numbers is to avoid the situation where the calculation result of the airspace flow anomaly degree is forced to be 0 when the absolute values of the flow difference or the trend test statistic are 0.
[0053] In another embodiment, the mean of the absolute value of the traffic flow difference and the absolute value of the trend test statistic is used as the airspace traffic anomaly degree of the flight to be predicted in each interval.
[0054] It should be noted that: the larger the absolute value of the trend test statistic, the more obvious the increasing or decreasing trend of landing airspace traffic, indicating that the landing airport is more severely affected by sudden severe weather or air traffic congestion within each interval, leading to a continuous increase or decrease in landing airspace traffic. The airspace traffic anomaly is used to measure the degree of difference between the landing airspace traffic of the airport to be predicted and the normal steady state in each interval, as well as the trend of landing airspace traffic change. When using the EMA model to dynamically predict flight delays, if the ground resources of the airport to be predicted are severely affected by sudden or cumulative events, the difference between the landing airspace traffic and the normal steady state is greater, and the trend of landing airspace traffic change caused by sudden severe weather or air traffic congestion is more obvious, then the calculated airspace traffic anomaly is greater.
[0055] Furthermore, by analyzing the average runway occupancy rate and its decay within each section, the high runway load factor for predicted flights within each section is obtained. The specific process is as follows: Calculate the average runway occupancy rate within each interval; obtain the abrupt change points of runway occupancy rate in time series within each interval; obtain the fitted straight line of all abrupt change points in time series within each interval. When the slope of the fitted line is non-negative, the runway high occupancy rate of the flights to be predicted in each interval is positively correlated with the average value and the slope of the fitted line, respectively. When the slope of the fitted line is negative, the runway high occupancy rate of the flights to be predicted in each interval is inversely proportional to the absolute value of the slope of the fitted line. It should be noted that if there are no abrupt change points or too few abrupt change points to make line fitting impossible, it indicates that the runway occupancy rate is relatively stable in each interval, and the slope of the fitted line is assigned a value of 0.
[0056] In this embodiment, the Bayesian mutation point detection algorithm is used to obtain mutation points in the runway occupancy rate. The Bayesian mutation point detection algorithm is a well-known technology and will not be described in detail in this application. As other implementation methods, based on the ability to obtain mutation points in the runway occupancy rate, implementers may use other existing technologies, such as the Pettt mutation point detection algorithm, the Bernaola Galvan segmentation algorithm, etc. This application does not impose any special restrictions.
[0057] In this embodiment, the least squares method is used to obtain the fitted line of the mutation point. The least squares method is a well-known technique and will not be described in detail in this application. As other implementation methods, based on the ability to obtain the fitted line of the mutation point, the implementer may use other existing techniques, such as linear regression analysis, weighted least squares method, etc. This application does not impose any special restrictions.
[0058] In this embodiment, the expression for the runway high load factor of the flights to be predicted in each interval is: In the formula, This indicates the runway high load factor for the flights to be predicted within the i-th interval; This represents the average runway occupancy rate within the i-th interval; This represents the slope of the fitted straight line over time for all abrupt changes within the i-th interval; , All of these represent preset values greater than 0, used to avoid... or When the value is 0, it indicates that the calculation result of the runway high load degree is forced to be 0. , The values are all preset by humans, and the implementer can set them according to the actual situation. In this embodiment, , The value of is 0.01; This indicates the absolute value operation.
[0059] In another embodiment, the expression for the runway high load factor of the flights to be predicted in each interval is: In the formula, This indicates the runway high load factor for the flights to be predicted within the i-th interval; This represents the average runway occupancy rate within the i-th interval; represents the slope of the fitted line on the time series for all abrupt change points in the i-th interval; exp() represents the exponential function with the natural constant as the base, used to map the data to positive numbers.
[0060] It's important to note that identifying abrupt change points is beneficial because it avoids the problem of a large proportion of steady-state runway occupancy data in real-world scenarios, which could lead to inaccurate data reflecting the runway occupancy response to sudden or cumulative events in the ground system. A steeper slope in the fitted line at the abrupt change point indicates a more significant increase in runway operational load or a slower rate of runway occupancy decay. Runway high load is used to characterize the high load and low decay status of runway occupancy at the airport where the predicted flight is landing within each interval. When using the EMA model for dynamic flight delay prediction, if the ground resources of the airport where the predicted flight is landing are severely affected by sudden or cumulative events, the airport runway occupancy will exhibit a high load or slow decay state, resulting in a higher calculated runway high load.
[0061] Furthermore, by combining the airspace traffic anomaly and runway high load within each interval, the load factor for the predicted flights within each interval is obtained, expressed as: In the formula, This represents the load factor of the flights to be predicted in the i-th interval; This represents the airspace traffic anomaly degree of the flight to be predicted within the i-th interval; This represents the runway high load factor for the flights to be predicted within the i-th interval; norm() is the normalization function.
[0062] It should be noted that the load factor is used to measure the airspace traffic anomalies and runway overload levels caused by sudden and cumulative ground system events affecting the airport where the flight to be predicted is to land in each interval. When using the EMA model for dynamic prediction of flight delays, if the ground resources of the airport where the flight to be predicted is to land are severely affected by sudden or cumulative events, the calculated airspace traffic anomalies and runway overload levels will both increase, and the calculated load factor will also increase. In this case, it is necessary to increase the attenuation factor in the EMA model to enhance the EMA model's ability to respond to recent sudden changes, thereby ensuring prediction accuracy.
[0063] Step 4: Obtain the attenuation factor adjustment coefficient of the flight to be predicted in each interval by using the load coefficient and flight deviation of the flight to be predicted in each interval.
[0064] In the process of using the EMA model to achieve dynamic prediction of flight delays through multi-source data fusion, the more severe the sudden changes in the flight status of the flight to be predicted, and the more severe the constraints of sudden and cumulative events on the ground resources of the landing airport, the more the attenuation factor in the EMA model should be increased to improve the robustness of the EMA model to sudden flight situations and "limited aerial perspective" situations, and to ensure prediction accuracy.
[0065] Based on the above analysis, by combining the flight deviation and load coefficient of the flights to be predicted within each interval, the attenuation factor adjustment coefficient for the flights to be predicted within each interval is obtained. This coefficient characterizes the degree of impact of sudden changes in flight status and ground system constraints on the flights to be predicted within each interval. Specifically, the product of the flight deviation and load coefficient of the flights to be predicted within each interval is used as the attenuation factor adjustment coefficient. A larger attenuation factor adjustment coefficient indicates that the response capability of the flights to be predicted within each interval should be improved to sudden changes in multi-source flight data, avoiding EMA model lag and thus reducing the accuracy of flight delay prediction. A schematic diagram of the process for obtaining the attenuation factor adjustment coefficient is shown below. Figure 2 As shown.
[0066] Step 5: Adjust the attenuation factor in each interval to predict the delay of the flight to be predicted.
[0067] The estimated arrival times of all times within and before each interval are arranged chronologically to form the estimated arrival time series for each interval. The position, altitude, velocity, landing airspace flow, and runway occupancy rate of all times within and before each interval are arranged chronologically to form the position, altitude, velocity, landing airspace flow, and runway occupancy rate sequences for each interval. Using these sequences as input, an EMA model is employed to output the predicted estimated arrival time for the next time interval. The weights of the position, altitude, velocity, landing airspace flow, and runway occupancy rate sequences are the normalized values of the cosine similarity between these sequences and the estimated arrival time series. The calculation of the cosine similarity is a well-known technique and will not be elaborated upon in this application.
[0068] The system obtains the position, altitude, speed, landing airspace traffic, and runway occupancy rate of M delayed flights during their flight process. Based on the attenuation factor adjustment coefficient of the flight to be predicted, it obtains the attenuation factor adjustment coefficient for each delayed flight in each interval. The system then performs distribution fitting on the attenuation factor adjustment coefficients of all delayed flights across all intervals, setting the fitting interval to [0,1] to fit the attenuation factor adjustment coefficients of all delayed flights to the same range. The minimum value from the fitting interval is selected as the attenuation factor adjustment coefficient threshold. If the attenuation factor adjustment coefficient of the flight to be predicted in any interval is greater than the attenuation factor adjustment coefficient threshold, the attenuation factor in the EMA model is adjusted; otherwise, the attenuation factor in the EMA model is not adjusted. The expression for adjusting the attenuation factor in the EMA model is as follows: In the formula, This represents the adjusted attenuation factor in the EMA model within the i-th interval; This represents the preset initial value of the decay factor in the EMA model; represents the attenuation factor adjustment coefficient for the flight to be predicted in the i-th interval; norm() is the normalization function.
[0069] In this embodiment, the Softrmax function is used to obtain the normalized value of cosine similarity so that the sum of the weights of the position sequence, altitude sequence, speed sequence, landing airspace traffic sequence, and runway occupancy sequence is 1. The Softrmax function is a well-known technology and will not be described in detail in this application.
[0070] In this embodiment, the value of M is 100. The value of M is preset by the user and can be set by the implementer according to the actual situation. This application does not impose any special restrictions.
[0071] In this embodiment, the process of obtaining the attenuation factor adjustment coefficient threshold is as follows: for all delayed flights in all intervals, obtain the probability distribution function of the attenuation factor adjustment coefficient. The horizontal axis of the probability distribution function is the attenuation factor adjustment coefficient, and the vertical axis is the probability of the attenuation factor adjustment coefficient. The attenuation factor adjustment coefficient when the probability distribution function is 0 for the first time in the interval [0,1] is used as the attenuation factor adjustment coefficient threshold.
[0072] In this embodiment, the preset initial value of the attenuation factor is 0.5. The preset initial value of the attenuation factor is preset by humans, and the implementer can set it according to the actual situation. This application does not impose any special restrictions.
[0073] The later the predicted arrival time of the next moment in each predicted interval is compared to the initial predicted arrival time, the higher the risk of delay for the predicted flight.
[0074] Based on the same inventive concept as the above methods, this application also provides a flight delay dynamic prediction system based on multi-source data fusion, including a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements the steps of any one of the above-described flight delay dynamic prediction methods based on multi-source data fusion.
[0075] In summary, this application quantifies the degree and trend of a flight's deviation from a preset route during flight by calculating the position deviation coefficient. This reflects whether the flight is affected by unforeseen factors, transforming complex flight trajectory information into concise numerical features for easier subsequent analysis. By calculating the flight state change rate and comprehensively considering the frequency domain characteristics of the flight's altitude and speed, it can more accurately reflect the changes in the flight state. By fusing the position deviation coefficient with the flight state change rate, the flight deviation rate is obtained, which comprehensively reflects the degree of impact of unforeseen factors on the flight in the short term. This provides more accurate short-term deviation information for the EMA model, helping it to more accurately capture the short-term dynamic changes of the flight and thus improve the prediction accuracy of flight delays. Furthermore, by calculating the airspace traffic anomaly and runway high load, and combining them to obtain the load coefficient, the utilization of airport airspace and runway resources is comprehensively considered, which can more accurately assess the comprehensive impact of airport airspace and ground factors on flight delays. Furthermore, by combining flight deviation with load factor, an attenuation factor adjustment coefficient is obtained, which can comprehensively reflect the degree of influence of various factors on the flight to be predicted. By adjusting the attenuation factor through the attenuation factor adjustment coefficient, the parameters of the EMA model can be adapted to the actual motion of the flight to be predicted, thereby improving the prediction accuracy of the EMA model for flight delays.
[0076] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to embodiments of this disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. In some alternative implementations, the functions marked in the blocks may occur in a different order than that shown in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. In the descriptions corresponding to the flowcharts and block diagrams in the accompanying drawings, the operations or steps corresponding to different blocks may also occur in a different order than disclosed in the description, and sometimes there is no specific order between different operations or steps. For example, two consecutive operations or steps may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. Each block in a block diagram and / or flowchart, and combinations of blocks in a block diagram and / or flowchart, can be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.
[0077] It will be apparent to those skilled in the art that this application is not limited to the details of the exemplary embodiments described above, and that this application can be implemented in other specific forms without departing from its essential characteristics. Therefore, the embodiments described above should be considered exemplary and non-limiting in all respects.
Claims
1. A method for dynamic prediction of flight delays based on multi-source data fusion, characterized in that, The method includes the following steps: The system can obtain the location, altitude, speed, and estimated arrival time of the flight to be predicted in real time, as well as the landing airspace traffic and runway occupancy rate of the airport where the flight is to land. An adjustment range for the attenuation factor in the EMA model used to predict the estimated arrival time is preset. The position deviation coefficient of the flight to be predicted in each interval is obtained by observing the changing trend and dispersion of the position deviation of the flight to be predicted from the preset route in each interval. By comparing the altitude distribution range of the flight to be predicted between each interval and its adjacent intervals, and combining the energy distribution of the speed of the flight to be predicted in the frequency domain in each interval, the flight state change degree of the flight to be predicted in each interval is obtained. This is then fused with the position deviation coefficient to obtain the flight deviation degree of the flight to be predicted in each interval. By analyzing the differences between the landing airspace traffic and the standard landing airspace traffic in each interval, as well as the changing trends of the landing airspace traffic, the airspace traffic anomaly degree of the flights to be predicted in each interval is obtained. Combined with the average level and attenuation of runway occupancy rate in each interval, the runway high load degree of the flights to be predicted in each interval is obtained, and the load coefficient of the flights to be predicted in each interval is obtained. Then, combined with the flight deviation degree, the attenuation factor adjustment coefficient of the flights to be predicted in each interval is obtained, so as to adjust the attenuation factor in each interval, thereby predicting the delay of the flights to be predicted.
2. The method for dynamic prediction of flight delays based on multi-source data fusion as described in claim 1, characterized in that, The process for obtaining the position deviation coefficient is as follows: Obtain the trend term of the deviation of the position of the flight to be predicted from the preset route in each interval in the time series, and calculate the Hurst exponent of the trend term; The dispersion of the position of the flight to be predicted within each interval relative to the preset route is denoted as the deviation dispersion. The position deviation coefficient is positively correlated with the Hurst exponent and the deviation dispersion, respectively.
3. The method for dynamic prediction of flight delays based on multi-source data fusion as described in claim 1, characterized in that, The process of obtaining the flight state change degree is as follows: Calculate the interquartile range of the altitude of the flight to be predicted in each interval, and record the difference between the interquartile range in each interval and the interquartile range in the adjacent preceding interval as the range difference. Obtain the frequency components of the speed of the flight to be predicted in the frequency domain within each interval, count the frequency with the highest energy in each frequency component, obtain the segmentation threshold of all frequencies counted in each interval, and record the frequency components that are greater than the segmentation threshold in each interval as high-frequency components, and record the frequency components other than high-frequency components in each interval as low-frequency components; calculate the ratio of the energy accumulation result of all high-frequency components in each interval to the energy accumulation result of all low-frequency components. The degree of change in flight state is positively correlated with the range difference and the ratio, respectively.
4. The method for dynamic prediction of flight delays based on multi-source data fusion as described in claim 1, characterized in that, The flight deviation is the normalized value of the product of the position deviation coefficient and the flight state change.
5. The method for dynamic prediction of flight delays based on multi-source data fusion as described in claim 1, characterized in that, The process of obtaining the airspace traffic anomaly degree is as follows: The difference between the average landing airspace flow in each interval and the standard landing airspace flow is denoted as the flow difference. Obtain trend test statistics for landing airspace traffic within each interval; The airspace traffic anomaly is positively correlated with the absolute value of the traffic difference and the absolute value of the trend test statistic, respectively.
6. The method for dynamic prediction of flight delays based on multi-source data fusion as described in claim 1, characterized in that, The process of obtaining the high load index of the runway is as follows: Calculate the average runway occupancy rate within each interval; obtain the abrupt change points of runway occupancy rate in time series within each interval; obtain the fitted straight line of all abrupt change points in time series within each interval. When the slope of the fitted line is non-negative, the high load of the runway is positively correlated with the average value and the slope of the fitted line, respectively; when the slope of the fitted line is negative, the high load of the runway is inversely proportional to the absolute value of the slope of the fitted line.
7. The method for dynamic prediction of flight delays based on multi-source data fusion as described in claim 1, characterized in that, The load factor is the normalized value of the product of the airspace flow anomaly and the runway high load factor.
8. The method for dynamic prediction of flight delays based on multi-source data fusion as described in claim 1, characterized in that, The attenuation factor adjustment coefficient is the product of the flight deviation and the load factor.
9. The method for dynamic prediction of flight delays based on multi-source data fusion as described in claim 1, characterized in that, The adjustment of the attenuation factor within each interval to predict flight delays includes: Based on the estimated arrival time, location, altitude, speed, landing airspace flow, and runway occupancy rate of each interval and all previous times, the EMA model is used to predict the estimated arrival time of the next time moment in each interval. Specifically, a threshold is set for the attenuation factor adjustment coefficient of a preset number of delayed flights during their flight process. If the attenuation factor adjustment coefficient of the flight to be predicted in each interval is greater than the threshold, the attenuation factor in the EMA model is adjusted; otherwise, the attenuation factor in the EMA model is not adjusted. The expression for adjusting the attenuation factor in the EMA model is as follows: In the formula, This represents the adjusted attenuation factor in the EMA model within the i-th interval; This represents the preset initial value of the decay factor in the EMA model; represents the attenuation factor adjustment coefficient for the flight to be predicted in the i-th interval; norm() is the normalization function.
10. A flight delay dynamic prediction system based on multi-source data fusion, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the flight delay dynamic prediction method based on multi-source data fusion as described in any one of claims 1-9.