A method for evaluating and predicting airport flight operation resilience under severe weather conditions

Abstract

The present invention provides an airport flight operation resilience assessment and prediction method under severe weather conditions, comprising the following steps: step 1: constructing an airport flight operation resilience assessment model, and evaluating the airport flight operation performance through different metrics; step 2: constructing an airport flight operation resilience prediction model based on LSTM, and predicting the airport performance under severe weather conditions. The present invention can predict the airport flight operation resilience based on meteorological factors, thereby reducing disasters and improving emergency response capabilities.

Description

Technical Field

[0001] The invention belongs to the field of aviation technology, and in particular relates to a method for evaluating and predicting the resilience of airport flight operations under severe weather conditions. Background Art

[0002] In recent years, the aviation transportation industry has maintained a high-speed growth trend, and the airport transportation volume has increased rapidly year by year. Due to the continuous updating and iteration of software and hardware, the continuous improvement of aircraft performance, control technology, ground support and operating procedures, and the increasingly scientific scheduling of airlines, the impact of equipment and human factors on the safety and efficiency of airport operations has been continuously reduced, and bad weather has become the main factor restricting the normal flight of flights. In order to resist the serious impact of bad weather, the research on the resilience of airport flight operations has emerged.

[0003] Unlike the extensive literature on the impact of extreme weather on travel activities, transportation system capacity, reliability, robustness, and vulnerability, research on the resilience of transportation systems in the post-disaster stage is still rare, and most existing studies on transportation resilience use topological methods to quantify the resilience of transportation systems. The commonly used method is to simulate emergency scenarios by continuously removing nodes / links in the transportation network, and to simulate the gradual dissipation of emergencies and system recovery scenarios by restoring nodes / links. Each time a node / link is removed or restored, the system's resilience is quantified. Although these studies provide preliminary explorations for the measurement and evaluation of transportation resilience, they are mainly focused on the network level, and few studies quantify the resilience of transportation network nodes or infrastructure, especially the lack of methods to evaluate the resilience of major transportation hubs. In addition, most of them use hypothetical emergency scenarios and simulation data, lacking the use of real case data to evaluate the extent to which airport performance is affected by different factors, which is limited in reflecting the resilience of airport flights.

[0004] In studying the impact of bad weather on airport flight delay prediction, deep learning has shown strong potential in general prediction research due to its flexible model structure and strong learning ability. For data with strong temporal relationships, the recurrent neural network (RNN) method has shown competitive capabilities. However, traffic in the traffic system has complex non-Euclidean correlations and directions, showing strong topological properties rather than general Euclidean spatial correlations. For these data, the original RNN is not applicable, and the long short-term memory (LSTM) network may perform better.

[0005] Overall, research on the impact of severe weather on airport flight operations has achieved certain results, but the models are mostly based on certain assumptions and are out of touch with actual flight operations. Most studies do not delve deeply into the coupling relationship between severe weather and flight operations, and there are few studies on the changes in airport flight operation performance under the overall process of severe weather. In addition, existing studies mostly use traditional statistical prediction methods and lack the combination of deep learning. Summary of the invention

[0006] In view of this, the present invention aims to propose a method for evaluating and predicting the resilience of airport flight operations under severe weather conditions.

[0007] To achieve the above object, the technical solution of the present invention is achieved as follows:

[0008] A method for evaluating and predicting the resilience of airport flight operations under severe weather conditions comprises the following steps:

[0009] Step 1: Construct an airport flight operation resilience assessment model and evaluate the airport flight operation resilience through different metrics;

[0010] Step 2: Build an airport flight operation resilience prediction model based on LSTM to predict airport performance under severe weather conditions.

[0011] Furthermore, in step 1, the departure rate is used as an evaluation indicator of the airport flight operation resilience model under severe weather conditions, and the departure rate is defined as follows:

[0012]

[0013] in: is the total number of scheduled departure flights in the time period [T1, T2]; represents the number of flights whose scheduled departure time is within the time period [T1, T2] and whose departure delay is less than 30 minutes; Represents the number of flights that were delayed before T1 and departed during the time period [T1, T2].

[0014] Furthermore, step 1 also includes taking different resilience measurement indicators at different stages to measure the performance of the evaluation indicators at different stages. The measurement indicators include response time, destruction time, destruction rate, robustness, recovery time, recovery rate, recovery capability, performance loss, and average performance loss.

[0015] Furthermore, the specific metrics are as follows:

[0016] Response Time (RST): The time from the start of the interference event to the time when the system performance begins to degrade. d'<t<t d , which indicates the system's ability to resist external interference;

[0017] RST=t d -t d '

[0018] Among them, t d ' is the time when the bad weather starts, t d This is the moment when the system is disturbed by external factors and its performance begins to decline;

[0019] Destruction Time (DSS): The time period from the initial decline in system performance to the lowest performance. d <t<t r , indicating the length of time the system performance has been degraded;

[0020] DSS=t r -t d

[0021] Among them, t r This is the moment when the system is disturbed by external factors and its performance drops to the lowest value;

[0022] Damage rate (RAPI DP ): In the destruction stage t d <t<t r , which indicates the speed at which the system degrades from initial performance to minimum performance;

[0023]

[0024] Among them, t r The moment when the system is disturbed by external factors and its performance drops to the lowest value; t d The moment when the system is disturbed by external factors and its performance begins to decline; MORP(t d ) is the system performance at the moment when the performance starts to decline, MORP(t r ) represents the minimum value of system performance under external interference;

[0025] Robustness (R): refers to the ability of a system to maintain its stability when it is disturbed by external events;

[0026] R=min{MORP(t)}(t d <t<t ns )

[0027] Among them, MORP*t (represents a discrete function that changes with time, t d Indicates the time when the system performance begins to decline after being disturbed, t nsIndicates the moment when the system returns to a new stable stage. R represents the maximum value of the system performance during this period, which can measure the maximum impact of external event disturbances on the system.

[0028] Recovery Time (RCT): The time period t from the moment when the system performance is the lowest to the moment when the system recovers to a new stable state r <t<t ns , indicating the length of time it takes for system performance to recover;

[0029] RCT=t ns -t r

[0030] Recovery rate (RAPI RP ): In the recovery phase r <t<t ns , which indicates the speed at which the system recovers from the lowest performance to the new stable stage performance;

[0031]

[0032] Where MORP(t) represents the discrete function of system performance that changes with time under a disturbance event, t r is the moment when the system is disturbed by external factors and its performance drops to the minimum value, t ns Indicates the time when the system returns to a new stable stage, MORP(t ns ) is the value of performance recovery to the stable stage, MORP(t r ) represents the minimum value of system performance under external interference;

[0033] Recovery Ability (RA): represents the recovery time at t≥t ns When the system reaches a new stable stage;

[0034]

[0035] Among them, MORP(t) represents the discrete function of system performance that changes with time under a disturbance event. ns ) is the value of performance recovery to the stable stage, MORP(t r ) represents the minimum value of system performance under external interference, MORP(t d ) is the system performance at the moment when the performance starts to degrade;

[0036] Loss of Performance (LOP): indicates the total amount of performance degradation during the entire process of the interference event;

[0037]

[0038] Time Averaged Performance Loss (TAPL): indicates the overall performance loss of the system during the destruction phase and the recovery phase;

[0039]

[0040] Where MORP(t) represents the discrete function of system performance that changes with time under a disturbance event; t d Represents the moment when system performance begins to decline; t ns The moment when the system returns to a new stable stage; MORP week (t) represents the function of system performance change under normal conditions.

[0041] Furthermore, the step 1 also includes establishing a non-negative general resilience index to measure the resilience of different systems under the same destructive event and the same system under different destructive events, and evaluating the resilience of airport flight operations by obtaining the change of airport flight performance in the time series, wherein the non-negative general resilience index (NGR) is:

[0042]

[0043] Furthermore, in the step 2, the architecture of the airport flight operation resilience prediction model is constructed based on LSTM, including using PCA technology to reduce the dimension of the data, using the LSTM model, and using the message data to perform deep learning on the airport performance MORP value. The result is predicted by the FC layer, and the prediction result is processed by the Sigmoid activation function to obtain the final prediction value MORP′.

[0044] Furthermore, the loss function of the airport flight operation resilience prediction model based on LSTM adopts the mean square error (MSE) between MORP′ and the true value MORP.

[0045] loss = MSE(MORP,MORP′).

[0046] Furthermore, the mean square error (MSE) is used as a metric to evaluate the difference between the predicted value and the true value, and to judge the degree of fit between the predicted value and the true value of the prediction model.

[0047]

[0048] Among them, y is the true value of MORP, y iIt represents the true value within a certain period of time, y′ is the predicted value obtained by the prediction model for the input data, and the mean square error (MSE) ranges from [0, +∞]. The closer it is to 0, the closer the predicted value is to the true value, and the better the model prediction performance.

[0049] Compared with the prior art, the method for evaluating and predicting the resilience of airport flight operations under severe weather conditions described in the present invention has the following advantages:

[0050] (1) Extending the existing resilience research methods to a wider range of real disaster situations, involving multiple types of severe weather rather than just one type of severe weather, will help aviation decision makers take countermeasures in advance to reduce the impact of severe weather based on the changing patterns and prediction results of airport flight operation performance under different severe events. This is of great significance for reducing flight delays and ensuring the safety of airport flight operation systems.

[0051] (2) A new airport operation performance resilience index, the departure rate, is proposed, and a non-negative comprehensive resilience index is proposed based on the actual situation of airport flight operations. Studying the changing law of airport flight operation performance under different severe weather conditions from occurrence to dissipation, and exploring the relationship between severe weather events and airport flight operation performance are of great practical significance for reducing the impact of severe weather on airport flight operations.

[0052] (3) Based on meteorological factors, the deep learning model LSTM is used to predict the resilience of airport flight operations, providing theoretical support for airport apron control and operation management center personnel to command flights and coordinate decisions, which is conducive to ensuring the safe and efficient operation of the airport. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] The accompanying drawings constituting a part of the present invention are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the accompanying drawings:

[0054] Figure 1 It is a schematic diagram of the change of system resilience;

[0055] Figure 2 This is a schematic diagram of the LSTM-based airport flight resilience prediction model under severe weather conditions;

[0056] Figure 3 is the Sigmoid function image;

[0057] Figure 4 This is a schematic diagram of the changes in airport flight operation performance under different severe weather conditions;

[0058] Figure 5 It is a schematic diagram comparing GR and NGR;

[0059] Figure 6Training loss value for the LSTM-based airport flight operation resilience prediction model under severe weather conditions. DETAILED DESCRIPTION

[0060] It should be noted that, in the absence of conflict, the embodiments of the present invention and the features in the embodiments may be combined with each other.

[0061] The present invention will be described in detail below with reference to the accompanying drawings and in conjunction with embodiments.

[0062] 1 Airport flight operation resilience assessment model

[0063] 1.1 Construction of resilience assessment model

[0064] The present invention proposes an indicator that has memory and can comprehensively measure the actual operating performance of airport flights under bad weather conditions, namely, the departure rate (Departure Rate), which is defined as follows:

[0065]

[0066] Where: is the total number of scheduled departure flights in the time period [T1, T2]; represents the number of flights whose scheduled departure time is within the time period [T1, T2] and whose departure delay is less than 30 minutes; Represents the number of flights that were delayed before T1 and departed in the time period [T1, T2]. This indicator can accurately describe the actual operational performance of airport flights, taking into account both the proportion of flights departing on time in a certain period of time to the number of flights scheduled to depart, and the performance of previously delayed flights in the subsequent time period.

[0067] The present invention uses the departure punctuality rate as the evaluation index of the airport flight operation resilience model under bad weather. The model divides the system before and after the disturbance into five stages, such as Figure 1 The vertical axis shown is the system performance resilience index (Measurement of Performance Resilience, MORP) that changes with time. The selection of the index to characterize the system performance resilience depends on the content of the study, and the value range fluctuates between [0,1], where 0 indicates that the system is in a paralyzed state and 1 indicates that the system is in an ideal performance state.

[0068] The different stages before and after the system is disturbed and the corresponding evaluation methods are summarized in Table 1:

[0069] Table 1 Summary of different stages of toughness process

[0070]

[0071] 1.2 Resilience metrics

[0072] In order to describe and analyze the airport flight operation performance, the following resilience metrics are selected:

[0073] 1. Response Time (RST): The time from when the interference event starts to when the system performance starts to degrade. d '<t<t d , which indicates the system's ability to resist external interference;

[0074] RST=t d -t d ′(1.2)

[0075] Among them, t d ' is the time when the bad weather starts, t d This is the moment when the system is disturbed by external factors and its performance begins to decline.

[0076] 2. Destruction Time (DSS): The time period from the initial decline in system performance to the lowest performance, t d <t<t r , indicating the length of time the system performance has been degraded;

[0077] DSS=t r -t d (1.3)

[0078] Among them, t r This is the moment when the system is disturbed by external factors and its performance drops to the lowest value.

[0079] 3. Damage rate (RAPI DP ): In the destruction stage t d <t<t r , which indicates the speed at which the system degrades from initial performance to minimum performance;

[0080]

[0081] Among them, t r The moment when the system is disturbed by external factors and its performance drops to the lowest value; t d The moment when the system is disturbed by external factors and its performance begins to decline; MORP(t d ) is the system performance at the moment when the performance starts to decline, MORP(t r ) represents the minimum value of system performance under external interference.

[0082] 4. Robustness (R): refers to the ability of the system to maintain its stability when disturbed by external events;

[0083] R=min{MORP(t){(t d <t<t ns ) (1.5)

[0084] Among them, MORP(t) represents the discrete function that changes with time, t d Indicates the time when the system performance begins to decline after being disturbed, t ns It indicates the moment when the system recovers to a new stable stage. R represents the maximum value of the system performance during this period of time, which can measure the maximum impact of external events on the system.

[0085] 5. Recovery Time (RCT): The time period from the lowest system performance to the system recovery to a new stable state t r <t<t ns , indicating the length of time it takes for system performance to recover;

[0086] RCT=t ns -t r (1.6)

[0087] 6. Recovery Rate (RAPI RP ): In the recovery phase r <t<t ns , which indicates the speed at which the system recovers from the lowest performance to the new stable stage performance;

[0088]

[0089] Where MORP(t) represents the discrete function of system performance that changes with time under a disturbance event, t r is the moment when the system is disturbed by external factors and its performance drops to the minimum value, t ns Indicates the time when the system returns to a new stable stage, MORP(t ns ) is the value of performance recovery to the stable stage, MORP(t r ) represents the minimum value of system performance under external interference.

[0090] 7. Recovery Ability (RA): indicates that when t ≥ t ns When the system reaches a new stable stage;

[0091]

[0092] Among them, MORP(t) represents the discrete function of system performance that changes with time under a disturbance event. ns ) is the value of performance recovery to the stable stage, MORP(t r) represents the minimum value of system performance under external interference, MORP(t d ) is the system performance at the moment when performance starts to degrade.

[0093] 8. Loss of Performance (LOP): indicates the total amount of performance degradation during the entire process of the interference event;

[0094]

[0095] 9. Time Averaged Performance Loss (TAPL): indicates the overall performance loss of the system during the destruction phase and the recovery phase;

[0096]

[0097] Where MORP(t) represents the discrete function of system performance that changes with time under a disturbance event; t d Represents the moment when system performance begins to decline; t ns The moment when the system returns to a new stable stage; MORP week (t) represents the function of system performance change under normal conditions.

[0098] In order to comprehensively compare and analyze the resilience of different systems under different interference events from a holistic perspective, a general resilience index (GR) is proposed to comprehensively measure the entire resilience process of the system before and after external interference. The definition is as follows:

[0099]

[0100] The shortcoming of this indicator is that when the system's performance level reaches the minimum value of 0 during an external interference event, that is, the system's robustness R is 0, and the calculated comprehensive resilience index value is 0. At this time, from an overall perspective, the system's comprehensive resilience capability is 0. This indicator does not take into account the situation where the system performance recovers after reaching the minimum value of 0. Therefore, if the robustness is 0 in a certain interference event, it is directly considered that the system's comprehensive resilience value in this event is 0, which is obviously not reasonable. In addition, this situation will also make it impossible to analyze and compare the comprehensive resilience values ​​between different systems.

[0101] On this basis, a new comprehensive resilience index is proposed: Nonnegative General Resilience (NGR):

[0102]

[0103] The above indicators provide important metrics for measuring the resilience of different systems under the same destructive events and the same system under different destructive events. The resilience of airport flight operations can be evaluated by obtaining changes in airport flight performance over a time series.

[0104] In the airport flight operation resilience assessment model under severe weather, a new airport operation performance resilience index, the departure rate, is proposed; secondly, based on the actual situation of airport flight operation, improvements are made and a non-negative comprehensive resilience index is proposed; the airport flight operation performance under normal weather is used as the baseline to reduce the error in calculating performance loss. The model compares and studies the changing rules of the system under different severe weather conditions and under severe weather conditions of different intensities, and provides a method for empirical research and analysis of airport flight operation performance.

[0105] 2. Airport flight operation resilience prediction

[0106] Due to the frequent occurrence of severe weather events, which have widespread destructive impacts on the functions of transportation infrastructure and systems, and even cause casualties, it will be more effective for mitigating disasters and improving emergency response capabilities if the spatiotemporal distribution pattern of transportation resilience under various severe weather events can be quantitatively observed and estimated based on transportation system performance and meteorological data, showing the specific severity of resilience loss and the overall recovery time of the system.

[0107] Based on the actual situation of predicting the resilience of airport flight operation performance through message data, an LSTM-based airport performance prediction model under severe weather conditions was adaptively optimized and built. Figure 2 shown.

[0108] The Sigmoid activation function is used to solve the problem that the airport flight operation performance resilience index MORP takes values ​​between 0 and 1. The Sigmoid activation function is shown in formula (2.1), and its function graph is shown in Figure 3 shown.

[0109]

[0110] The specific architecture details of the LSTM-based airport flight operation performance resilience prediction model under severe weather are shown in Table 2. The values ​​of each factor of the original message data are normalized to between -1 and 1, and the PCA technology is used to reduce the dimension of the data. The LSTM model is used to perform deep learning on the airport performance MORP value using the message data, and the results are predicted through the FC layer. The prediction results are processed by the Sigmoid activation function to obtain the final prediction value MORP′.

[0111] Table 2 Detailed information of the LSTM-based airport flight operation resilience prediction model

[0112]

[0113]

[0114] The loss function of the LSTM-based airport performance prediction model under severe weather conditions uses the mean square error (MSE) between MORP′ and the true value MORP, as shown in formula (2.2). In order to reduce the loss function value, the optimizer selects Adam.

[0115] The model parameters are continuously optimized through gradient descent to improve the model accuracy.

[0116] loss = MSE(MORP,MORP′)(2.2)

[0117] In order to verify the accuracy of the predicted values ​​obtained by the model, the mean square error (MSE) is used as a measurement indicator to evaluate the difference between the predicted value and the true value to determine the degree of fit between the predicted value and the true value of the model proposed in this paper, as shown in formula (2.3).

[0118]

[0119] Among them, y is the true value of MORP, y i represents the true value within a certain period of time, and y′ is the predicted value of the input data obtained by the model proposed in this paper. The mean square error (MSE) ranges from [0, +∞]. The closer it is to 0, the closer the predicted value is to the true value, and the better the model prediction performance.

[0120] In addition, the mean relative error (MRE), root mean square error (RMSE), mean absolute error (MAE), and symmetric mean absolute percentage error (SMAPE) are used to reflect the accuracy of the proposed model for airport performance prediction, as shown in formulas (2.4), (2.5), (2.6), and (2.7).

[0121]

[0122]

[0123] Among them, k is the total number of message data. The value range of MRE, RMSE, MAE, and SMAPE is [[0,+∞], so the smaller the value is, the closer it is to 0, which means that the closer the MORP prediction result in this paper is to the true value, the better the prediction performance of the model.

[0124] 3 Empirical analysis

[0125] The meteorological message data, meteorological warning information and flight takeoff data of Beijing Capital Airport from 00:00 on January 1, 2021 to 00:00 on August 1, 2021 (a total of 5088 hours) were selected. After data preprocessing and other operations, a total of 7 severe weather events were screened out, including 2 snowfall events (snowfall event on January 18 and snowfall event on January 25), 1 sandstorm event (sandstorm event on March 14), and 4 rain events (rain event on July 1, rain event on July 5, rain event on July 11, and rain event on July 26), represented by event 1, event 2, event 3, event 4, event 5, event 6 and event 7 respectively. The meteorological warning information for these severe weather events obtained from the Beijing Meteorological Observatory website are as follows: a blue warning for heavy rain was issued at 16:50 on July 1; an orange warning for heavy rain was issued at 12:59 on July 5; an orange warning for heavy rain was issued at 20:00 on July 11; a yellow warning for heavy rain was issued at 21:25 on July 26, and then a red warning for heavy rain was issued at 23:40.

[0126] 3.1 Airport flight operation resilience assessment

[0127] The indicators are calculated by the selected severe events and resilience measurement index formula, as shown in Table 3.

[0128] Table 3 Metrics of airport flight operation resilience at each stage under severe weather events

[0129]

[0130] Taking the departure rate under normal weather as the baseline, the changes in departure rate under snowfall, sandstorm and thunderstorm events are depicted. Figure 4 The dashed line in the figure represents the airport departure rate under normal weather conditions, and the green line represents the airport departure rate under severe weather events; the three dashed vertical lines represent the time t when the airport flight operation performance begins to decline. d , the performance reaches the lowest time t r , the performance returns to the stable time t ns As can be seen from the figure, different severe weather events have different degrees of impact on airport flight operations.

[0131] For different severe weather events, the comprehensive resilience index values ​​of GR and NGR when the corresponding severe weather occurs are calculated, and the differences in airport flight operation resilience shown by GR and NGR under seven severe weather events are compared and analyzed. Figure 5As shown. The bars in the figure represent the comprehensive resilience index GR, and the broken line represents the improved non-negative comprehensive resilience index NGR. As can be seen from the figure, under the severe weather of thunderstorms (events 4 to 7), the airport flight operation performance level is reduced to the lowest value of 0 at a certain moment, that is, the robustness R is 0. Therefore, the GR values ​​of events 4 to 7 calculated according to formula (1.12) are all 0, indicating that the airport flight operation system has no "rebound" process under the interference of events 4 to 7, which is contrary to the actual situation. In addition, since the GR of events 4 to 7 is 0, it is impossible to explore the law between these severe weather events and GR. From the overall airport flight operation performance changes before and after the event, under the interference of events 4 to 7, the performance of the airport flight operation system is only 0 in a certain period of time. After that, with the dissipation of severe weather and the coordination and scheduling of management personnel, the system is in a gradual recovery process and restored to the original state, that is, rebound, and GR is directly 0, which shows that the NGR index comprehensively evaluates the resilience of airport flight operations. It is more realistic than the original GR.

[0132] 3.2 Prediction of Airport Flight Operation Resilience

[0133] 3.2.1 Data processing

[0134] Data normalization is used to deal with the problem of large differences in the dimensions of values ​​among different message factors, as shown in formula (3.1), where x represents a certain message factor.

[0135]

[0136] Principal component analysis (PCA) is used to reduce the dimension of the data. Let the result of normalized message data be X * , use formulas (3.2) and (3.3) to calculate X * The covariance matrix C of is used, and formulas (3.4) and (3.5) are used to calculate the cumulative contribution rate of the principal components and the sample data of m principal components respectively.

[0137]

[0138] Where Cov(x,y) is the matrix X * The covariance between the xth column and the yth column of , where n is the number of factors in each message data.

[0139]

[0140] In the formula, k is the number of samples.

[0141] According to the PCA technology, the message data can be effectively reduced in dimension, thereby achieving the effect of compacting the data or simplifying the model, while maintaining the original data information to the greatest extent.

[0142] 3.2.1 Experimental details

[0143] The METAR message is collected every half hour by the airport automatic weather station, and the time period is from 00:00 on January 1, 2021 to 23:30 on July 31, 2021. Excluding the three factors of time, message type, and airport code, the following 39 factors are selected as the message data every half hour: wind speed (m / s), visibility, shallow, scattered, low blowing, blowing, showery, thunderstorm, supercooling, drizzle, rain, snow, sleet, ice needles, ice grains, hail, sleet, fog, smoke, volcanic ash, floating, sand, haze, sand roll, squall, tornado, sandstorm, dust storm, cloud amount 1, cloud base height 1, cumulus 1, cloud amount 2, cloud base height 2, cumulus 2, cloud amount 3, cloud base height 3, cumulus 3, temperature, dew point temperature, a total of 39 message data as the factors of each sample, and a total of 10176 samples are collected as model input. Based on the quantization of message data, three experimental settings of time series steps are adopted: 6 hours, 12 hours, and 24 hours. 70% of the data is used for training, 20% for testing, and 10% for validation. Formula (2.8) is used to optimize the model, and the model parameters that show the best fit (MSE minimum) on the validation set during model training are saved. The specific data set division is shown in Table 4. The experimental parameter settings are shown in Table 5. The experimental environment is shown in Table 6.

[0144] Table 4 Experimental data division

[0145]

[0146] Table 5 Experimental parameter settings

[0147]

[0148] Table 6 Experimental environment

[0149]

[0150] The principal component eigenvalues, contribution rates and cumulative contribution rates of the 39 factors can be obtained through formulas (3.1-3.5) as shown in Table 7:

[0151] Table 7 Principal component eigenvalues, contribution rates and cumulative contribution rates of each factor

[0152]

[0153]

[0154] 3.2.3 Experimental results and analysis

[0155] The processed data is used to train, verify and test the LSTM-based airport performance prediction model under severe weather conditions. The loss function loss value of the model training process is as follows: Figure 6 shown.

[0156] Depend on Figure 6 It can be seen that the loss value fluctuates greatly in 24 hours, the loss decreases slowly in 12 hours, and the loss decreases more gently in 6 hours, but there are still some fluctuations. In general, the overall trend of the loss value under the three time periods is that it gradually decreases with the increase in the number of training times, and finally drops to the same order of magnitude. The trend of gradually decreasing loss values ​​indicates that the airport performance prediction model under severe weather conditions based on LSTM is constantly being optimized.

[0157] In order to evaluate the prediction accuracy of the LSTM-based airport performance prediction model under severe weather conditions, the mean square error (MSE) at different time series step sizes was calculated, as shown in Table 8.

[0158] Table 8 MSE results under different step lengths

[0159]

[0160] The mean square error (MSE) is a measure used to reflect the difference between the predicted value and the true value. If the predicted value is closer to the true value, the model has a stronger prediction ability, and the MSE will be closer to 0. In general, the model trained with a time step of 6 hours is better than the model trained at other times, achieving the best results on the training set and the test set, and has good stability.

[0161] The mean relative error (MRE), root mean square error (RMSE), mean absolute error (MAE), and symmetric mean absolute percentage error (SMAPE) indicators are calculated and shown in Table 9.

[0162] Table 9 Test set metric results under different step lengths

[0163]

[0164] Table 9 evaluates the prediction performance of the proposed model from the perspective of error analysis between the predicted value and the true value. From the error results of the test set with different time series steps, all error index values ​​in the 6-hour case are less than 0.22, indicating that the predicted value is highly similar to the true value.

[0165] In order to reasonably evaluate the problems raised by the present invention and the effects of the corresponding improvements, an ablation experiment was conducted, and the results are shown in Table 10. The LSTM model (hidden size = 128, num layers = 2) was used as the base model (Baseline), and then relevant ablation experiments were conducted based on whether data normalization processing, data PCA dimensionality reduction, and Sigmoid activation function were used. Since the data PCA dimensionality reduction technology includes data normalization processing, data normalization processing must be used when using PCA.

[0166] Table 10 Ablation experiment

[0167]

[0168] As shown in Table 10, data normalization has a significant impact on the performance of the model. When the Sigmoid function (Baseline5) is added to the model using data normalization, the model performance is further improved, and the error value calculated on the test set is further reduced. The model (Our) using all three has the best performance, which proves the effectiveness of the proposed method to solve the problems of data units and dimensions, too long and redundant information, and MORP values ​​ranging from 0 to 1.

[0169] The present invention predicts the airport performance and measures it using a variety of quantitative evaluation methods for error analysis. By comparing and analyzing the error analysis indicators under different time strategies, the optimal performance of the prediction model under message data can be evaluated, and the resilience of airport flight operations can be predicted based on meteorological factors, thereby reducing disasters and improving emergency response capabilities.

[0170] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for evaluating and predicting the resilience of airport flight operations under severe weather conditions, characterized by: The following steps are included Step 1: Construct an airport flight operation resilience assessment model and evaluate the airport flight operation resilience through different metrics; The departure rate is used as an evaluation indicator of the airport flight operation resilience model under severe weather conditions. The departure rate is defined as follows: in: is the total number of scheduled departure flights in the time period [T1, T2]; represents the number of flights whose scheduled departure time is within the time period [T1, T2] and whose departure delay is less than 30 minutes; Represents the number of flights that were delayed before T1 and departed during the [T1, T2] time period; It also includes taking different resilience metrics at different stages to measure the performance of evaluation indicators at different stages. The metrics include response time, damage time, damage rate, robustness, recovery time, recovery rate, recovery capability, performance loss, and average performance loss. Step 2: Build an airport flight operation resilience prediction model based on LSTM to predict airport performance under severe weather conditions; The architecture of the airport flight operation resilience prediction model based on LSTM includes using PCA technology to reduce the dimension of data, using the LSTM model and message data to conduct deep learning on the airport performance MORP value, and the results are predicted by the FC layer. The prediction results are processed by the Sigmoid activation function to obtain the final prediction value. The loss function of the airport flight operation resilience prediction model based on LSTM adopts the mean square error (MSE) between MORP′ and the true value MORP. loss = MSE(MORP,MORP'); The mean square error (MSE) is used as a metric to evaluate the difference between the predicted value and the true value, and to judge the degree of fit between the predicted value and the true value of the prediction model. Among them, y is the true value of MORP, y i It represents the true value within a certain period of time, y′ is the predicted value obtained by the prediction model for the input data, and the mean square error (MSE) ranges from [0, +∞]. The closer it is to 0, the closer the predicted value is to the true value, and the better the model prediction performance.

2. The method for evaluating and predicting the resilience of airport flight operations under severe weather conditions according to claim 1 is characterized by: in, The specific metrics are as follows: Response Time (RST): The time from the start of the interference event to the time when the system performance begins to degrade. d '<t<t d , which indicates the system's ability to resist external interference; RST=t d -t d ′ Among them, t d ' is the time when the bad weather starts, t d This is the moment when the system is disturbed by external factors and its performance begins to decline; Destruction Time (DSS): The time period from the initial decline in system performance to the lowest performance. d <t<t r , indicating the length of time the system performance has been degraded; DSS=t r -t d Among them, t r This is the moment when the system is disturbed by external factors and its performance drops to the lowest value; Damage rate (RAPI DP ): In the destruction stage t d <t<t r , which indicates the speed at which the system degrades from initial performance to minimum performance; Among them, t r The moment when the system is disturbed by external factors and its performance drops to the lowest value; t d The moment when the system is disturbed by external factors and its performance begins to decline; MORP(t d ) is the system performance at the moment when the performance starts to decline, MORP(t r ) represents the minimum value of system performance under external interference; Robustness (R): refers to the ability of a system to maintain its stability when it is disturbed by external events; R=min{MORP(t)}(t d <t<t ns ) Among them, MORP(t) represents the discrete function that changes with time, t d Indicates the time when the system performance begins to decline after being disturbed, t ns Indicates the moment when the system returns to a new stable stage. R represents the maximum value of the system performance during this period, which can measure the maximum impact of external event disturbances on the system. Recovery Time (RCT): The time period t from the moment when the system performance is the lowest to the moment when the system recovers to a new stable state r <t<t ns , indicating the length of time it takes for system performance to recover; RCT=t ns -t r Recovery Rate (RAPI RP ): In the recovery phase r <t <t ns , which indicates the speed at which the system recovers from the lowest performance to the new stable stage performance; Where MORP(t) represents the discrete function of system performance that changes with time under a disturbance event, t r is the moment when the system is disturbed by external factors and its performance drops to the minimum value, t ns Indicates the time when the system returns to a new stable stage, MORP(t ns ) is the value of performance recovery to the stable stage, MORP(t r ) represents the minimum value of system performance under external interference; Recovery Ability (RA): represents the recovery time at t≥t ns When the system reaches a new stable stage; Among them, MORP(t) represents the discrete function of system performance that changes with time under a disturbance event. ns ) is the value of performance recovery to the stable stage, MORP(t r ) represents the minimum value of system performance under external interference, MORP(t d ) is the system performance at the moment when the performance starts to degrade; Loss of Performance (LOP): indicates the total amount of performance degradation during the entire process of the interference event; Time Averaged Performance Loss (TAPL): indicates the overall performance loss of the system during the destruction phase and the recovery phase; Where MORP(t) represents the discrete function of system performance that changes with time under a disturbance event; t d Represents the moment when system performance begins to decline; t ns The moment when the system returns to a new stable stage; MORP week (t) represents the function of system performance change under normal conditions.

3. The method for evaluating and predicting the resilience of airport flight operations under severe weather conditions according to claim 2 is characterized by: The step 1 also includes establishing a non-negative general resilience index to measure the resilience of different systems under the same destructive event and the same system under different destructive events, and evaluating the resilience of airport flight operations by obtaining the change of airport flight performance in time series, wherein the non-negative general resilience index (NGR) is:

Images