A method for predicting the service time of airport water trucks

Abstract

The present invention discloses a method for predicting the service time of a clean water truck at an airport, comprising four steps: preliminary statistics of the clean water truck service time, grouping of the clean water truck service time, probability distribution fitting and testing of the clean water truck service time, and prediction of the clean water truck service time. The present invention proposes a grouping method for clean water truck operation support data, which can group various types of clean water truck operation support data indiscriminately, and ensures that subsequent calculations and predictions can be performed within a feasible time. Based on the grouping method, a clean water truck service time prediction method of "classification, initial prediction and final prediction" is further proposed. Compared with the traditional empirical probability prediction method, the present invention greatly improves the prediction accuracy of the clean water truck service time. A higher prediction accuracy is conducive to the evaluation of airport support capabilities and airport support vehicle scheduling research based on the clean water truck service time, so that the research content has stronger practical significance and meets the needs of actual engineering applications.

Description

Technical Field

[0001] The invention relates to a method for predicting the service time of an airport water cleaning vehicle, and belongs to the technical field of water cleaning vehicle service time prediction. Background Art

[0002] The airport water truck is a type of airport support vehicle that provides drinking water refueling services for aircraft and plays a vital role in the flight support process. At present, the research on the service time of support vehicles is mainly applied to two major fields: airport support capability evaluation and airport support vehicle scheduling research. In previous studies, researchers basically all used a fixed value or the average value of the support vehicle service time as the service time of the support vehicle. Very few studies used triangular distribution or gamma distribution to represent the service time of support vehicles, but the scientificity and significance of the distribution were not verified.

[0003] Only a few airports in China have the data collection capabilities for the service time of clean water trucks, such as Beijing Daxing International Airport, Beijing Capital International Airport, Shanghai Pudong International Airport, and Guangzhou Baiyun International Airport. At the same time, there is currently no research on the service time of clean water trucks. The current research on the service time of clean water trucks at airports has the following main shortcomings: (1) Directly using a fixed value or average value as the service time of clean water trucks, which is inconsistent with objective facts; (2) Directly using a certain distribution as the distribution of the service time of clean water trucks, but only through manual direct observation and simple probability distribution fitting, which lacks rationality and scientificity; (3) Due to the shortcomings of previous research, it is impossible to make a reasonable prediction of the service time of clean water trucks. If, as in previous studies, a certain probability distribution is used to represent all the service times of clean water trucks, and then a prediction is made at a certain confidence level, this will result in a low prediction accuracy rate and cannot meet the needs of actual engineering applications. Summary of the invention

[0004] The technical problem to be solved by the present invention is to provide a method for predicting the service time of a water truck at an airport, which can be applied to different types of water truck support data forms, and improves the prediction accuracy of the water truck service time compared with the original related research. A higher prediction accuracy is conducive to the evaluation of airport support capabilities based on the water truck service time and the study of airport support vehicle scheduling, making it have stronger practical significance and meet the needs of actual engineering applications.

[0005] The present invention adopts the following technical solutions to solve the above technical problems:

[0006] A method for predicting the service time of an airport water cleaning vehicle comprises the following steps:

[0007] Step 1: For the airport to be predicted, obtain the water truck operation support data of the airport, including the departure flight number, aircraft model, aircraft model category, departure airport, destination airport, planned departure time, actual departure time, water truck start service time and water truck end service time, a total of 9 attributes, and clean and calculate the support data;

[0008] Step 2: For the 9 attributes of the security data, the actual departure time and the determined time of the end of the clean water truck service time are both later than the start of the clean water truck service time. The remaining 7 attributes after removing these two attributes are used as the initial grouping attributes;

[0009] Step 3: Combine the remaining 7 attributes, calculate the number of grouping schemes that guarantee data division under each combination, that is, the complexity, and set a complexity threshold. If the number of grouping schemes calculated for a certain combination is less than the complexity threshold, then retain the grouping scheme corresponding to the combination, otherwise screen out the grouping scheme corresponding to the combination;

[0010] Step 4: For the grouping schemes corresponding to the combinations retained in step 3, the attribute label of the combination is the attribute label of the corresponding grouping scheme. If the number of guaranteed data in a grouping scheme is less than 40, the grouping scheme is deleted. For the remaining grouping schemes, the chi-square test method is used to perform two cross-over chi-square tests. If there are two grouping schemes whose two cross-over chi-square tests are significantly not in the same empirical probability distribution at the 99.9% confidence level, both grouping schemes are retained; otherwise, the two grouping schemes are merged into one grouping scheme.

[0011] Step 5: For all the grouping schemes obtained in step 4, perform probability distribution fitting and probability distribution test on each grouping scheme to obtain the final probability distribution of the grouping scheme;

[0012] Step 6: For the guarantee data of the water truck service time to be predicted, if the guarantee data satisfies all attribute labels of a grouping scheme obtained in step 4, the grouping scheme is used as an alternative grouping of the guarantee data, thereby obtaining all the alternative groups;

[0013] Step 7: For each candidate group, an initial prediction is made on the service time of the water cleaning truck to be predicted according to a preset prediction accuracy, and the lower limit and upper limit of the initial predicted service time of the water cleaning truck corresponding to the candidate group are obtained;

[0014] Step 8, according to the lower and upper limits of the initial predicted service time of the clean water truck corresponding to all the candidate groups, the expected final prediction result is obtained.

[0015] As a preferred embodiment of the present invention, the departure airport described in step 1 is the airport to be predicted; the planned departure time is the planned time for the departure flight to withdraw the wheel block from the departure airport as declared in advance, with a time accuracy of 5 minutes; the actual departure time is the actual time for the departure flight to withdraw the wheel block from the departure airport in actual operation, with a time accuracy of 1 minute; the start service time of the fresh water truck and the end service time of the fresh water truck are the time points when the airport fresh water truck actually starts to fill fresh water for the departing flight and ends the fresh water filling operation in actual operation, respectively, and the time accuracy is 1 minute.

[0016] As a preferred solution of the present invention, the cleaning and calculation of the security data in step 1 is specifically as follows:

[0017] For each piece of security data, if at least one of the clean water truck service start time and the clean water truck service end time of the data is in an incorrect form, the data will be removed; for each piece of security data after cleaning, the clean water truck service start time will be subtracted from the clean water truck service end time of the data to obtain the clean water truck service time of the data, and the time accuracy of the clean water truck service time is 1 minute.

[0018] As a preferred solution of the present invention, the number of grouping schemes for ensuring data division under each combination described in step 3, i.e., the complexity, is calculated by the following formula:

[0019]

[0020] Where D represents the complexity of securing data partitioning when m attributes are selected from the remaining 7 attributes for combination, m represents the total number of all attributes used to calculate the complexity, n = 1, ..., m, a n Represents the total number of data after deduplication of the nth attribute, i = 1, ..., a n ;

[0021] The attributes that need to be deduplicated are the planned departure time and the clean water truck start service time. The planned departure time with a time accuracy of 5 minutes is converted to a planned departure time with a time accuracy of 1, 2, 3, 4, 6, 8, 12, or 24 hours. The clean water truck start service time with a time accuracy of 1 minute is converted to a clean water truck start service time with a time accuracy of 1, 2, 3, 4, 6, 8, 12, or 24 hours.

[0022] As a preferred solution of the present invention, in step 5, each grouping scheme is subjected to probability distribution fitting and probability distribution test to obtain the final probability distribution of the grouping scheme, specifically:

[0023] For a grouping scheme, 20 probability distributions are used to fit the service time of the water truck respectively, and the maximum likelihood method is used to solve the parameters of each probability distribution. The 20 probability distributions include four categories: single-parameter probability distribution, two-parameter probability distribution, three-parameter probability distribution and four-parameter probability distribution; the single-parameter probability distribution includes: exponential distribution, Poisson distribution and Rayleigh distribution; the two-parameter distribution includes: Birnbaum-Saunders distribution, extreme value distribution, gamma distribution, half-normal distribution, inverse Gaussian distribution, logical distribution, log-logistic distribution, log-normal distribution, upper-middle distribution, normal distribution, Rice distribution and Weibull distribution; the three-parameter distribution includes: Burle distribution, generalized extreme value distribution, generalized Pareto distribution and t-location scale distribution; the four-parameter distribution includes: stable distribution; the chi-square test is performed on the fitting results of the 20 probability distributions of the grouping scheme, and the probability distribution with the highest goodness of fit is selected as the final probability distribution of the grouping scheme.

[0024] As a preferred solution of the present invention, the calculation formula of the initial prediction in step 7 is:

[0025]

[0026]

[0027] Wherein, d represents the prediction correct probability value under the preset prediction accuracy, b represents the predicted lower limit of the clean water truck service time, Δc represents the preset accuracy of the predicted clean water truck service time, b+Δc represents the predicted upper limit of the clean water truck service time, and p(x) represents the probability distribution function of the alternative grouping;

[0028] When b When ′ is always greater than 0, b takes the value of its maximum value in the domain minus Δc; when d b ′ is always less than 0, b takes the minimum value of its domain; when d b When ' is not always greater than and not always less than 0, b takes the value of d b The maximum value of d when ′ is equal to 0.

[0029] As a preferred solution of the present invention, the calculation formula of step 8 is:

[0030]

[0031] Among them, F 0 and F 1 Respectively represent the lower and upper limits of the final prediction results, and They represent the lower and upper limits of the initial prediction of the jth candidate group, G j They respectively represent the goodness of fit of the probability distribution corresponding to the j-th alternative grouping, and k represents the number of all alternative groups.

[0032] Compared with the prior art, the present invention adopts the above technical solution and has the following technical effects:

[0033] 1. The present invention proposes a method for grouping water truck operation support data. Various types of water truck operation support data can be grouped indiscriminately, and subsequent calculation and prediction can be performed within a feasible time.

[0034] 2. The present invention is based on the grouping method and further proposes a "classification, initial prediction and final prediction" method for predicting the service time of a clean water truck. Compared with the traditional empirical probability prediction method, it greatly improves the prediction accuracy of the clean water truck service time. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 The present invention is a flow chart of a method for predicting the service time of an airport water cleaning vehicle.

[0036] Figure 2 It is a partial probability distribution fitting of a certain group of the service time of the water truck of the present invention.

[0037] Figure 3 It is the goodness of fit of the probability distribution of a certain grouping of the water truck service time of the present invention.

[0038] Figure 4 The present invention is used to predict the service time of a water truck, and the accuracy of the prediction method is compared with the traditional prediction method. DETAILED DESCRIPTION

[0039] The embodiments of the present invention are described in detail below, and examples of the embodiments are shown in the accompanying drawings. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and cannot be interpreted as limiting the present invention.

[0040] like Figure 1 The figure is a flow chart of a method for predicting the service time of an airport water truck according to the present invention. The specific steps are as follows:

[0041] Step 1 Preliminary statistics of water truck service time

[0042] The preliminary statistics of the service time of the water truck are derived from the operation and support data of the airport water truck. Common airport water truck operation and support data generally include: departure flight number, aircraft model, aircraft model category, departure airport, destination airport, planned departure time, actual departure time, water truck start service time and water truck end service time.

[0043] The "departure flight number" of a piece of airport water truck operation support data is the call sign of the departure flight corresponding to the data. The "aircraft model" of the support data is the specific model code of the departure flight, including but not limited to: A3XX (Airbus series), B7XX (Boeing series), etc. The "aircraft model category" of the support data is the wake vortex type of the departure flight model, including six categories: A / B / C / D / E / F. The "origin airport" of the support data is all the same, indicating that all departure flights take off from the airport under study, expressed as the ICAO (International Civil Aviation Organization) four-character code of the airport under study. The "destination airport" of the support data is the ICAO four-character code of other airports served by the airport under study. The "planned departure time" of the support data is the planned time for the departure flight to withdraw the wheel block from the departure airport reported in advance, expressed as an integer multiple of 5 minutes (that is, the time accuracy is 5 minutes). The "actual departure time" of the guarantee data is the actual time when the outbound flight is unblocked from the departure airport during actual operation, expressed as an integer multiple of 1 minute (i.e., the time accuracy is 1 minute). The "fresh water truck start service time" and "fresh water truck end service time" of the guarantee data are the time points when the airport fresh water truck actually starts to add fresh water to the outbound flight and ends the fresh water addition operation during actual operation of the outbound flight, expressed as an integer multiple of 1 minute (i.e., the time accuracy is 1 minute).

[0044] The preliminary statistical process of the clean water truck service time is divided into two sub-steps: data cleaning and data calculation. Data cleaning means that for each piece of security data, if the "clean water truck service start time" or "clean water truck service end time" of the data is in an incorrect format (including blank and non-time data format), the data will be removed. Data calculation means that for each piece of security data after data cleaning, the "clean water truck service end time" of the data is subtracted from the "clean water truck service start time" to obtain the "clean water truck service time" of the data. Since the time accuracy of "clean water truck service start time" and "clean water truck service end time" is 1 minute, the time accuracy of "clean water truck service time" is also 1 minute.

[0045] Step 2 Grouping of water truck service time

[0046] Based on step 1, the water truck service time grouping process includes three sub-steps: determining the initial grouping attributes, complexity screening, and determining the grouping.

[0047] 2.1 The purpose of determining the initial grouping attributes is to preliminarily determine the single or multiple attributes of the support data used for grouping. As described in step 1, the support data usually has 9 attributes. Except for the "actual departure time" and the "clean water truck end service time", the other 7 attributes can be determined as the initial grouping attributes. The reason for excluding the "actual departure time" and the "clean water truck end service time" is: the purpose of the present invention is to predict the clean water truck service time, and the time nodes of the "actual departure time" and the "clean water truck end service time" are both after the "clean water truck start service time", so these attributes cannot be used for predictive analysis. In an extended way, if the determination time of a certain attribute of the support data (regardless of whether it belongs to the time type) is later than the "clean water truck start service time", then the attribute cannot be used to determine the initial grouping attribute.

[0048] 2.2 The purpose of complexity screening is to calculate the number of schemes that guarantee data partitioning based on the single or multiple combinations of the initial grouping attributes determined above, and to screen the grouping method with a reasonable number of schemes. If the number of schemes is too large (sometimes the number of digits is as high as dozens, hundreds or even higher), the computer cannot complete the grouping operation within a feasible time, let alone execute the subsequent steps within a feasible time. The calculation of complexity is shown in formula (1): D represents complexity, m represents the total number of all attributes used to calculate complexity, n represents the sequence of a certain attribute, and a n Represents the total number of deduplicated data of the nth attribute, and i represents a data sequence after deduplicating the nth attribute. If the complexity is greater than the preset complexity threshold, this solution is screened out; otherwise, it is retained. For attributes in time or numerical form, its precision can be adjusted (for example, the "planned departure time" with a precision of 5 minutes is converted to the number of hours of the "planned departure time" with a precision of 1 / 2 / 3 / 4 / 6 / 8 / 12 / 24 hours, so that 288 deduplicated data in a day become 24 / 12 / 8 / 6 / 4 / 3 / 2 / 1 deduplicated data); for attributes in other forms, remain unchanged.

[0049]

[0050] 2.3 Determining the grouping is based on determining the initial grouping attributes and the complexity screening operation, the purpose is to ensure that the distribution of the service time of the water truck between independent groups is significantly different at a certain confidence level. The present invention uses the traditional chi-square test method to perform two crossover chi-square tests on the empirical probability distribution of the grouping between each two independent groups (the two groups are theoretical values ​​or observed values ​​of each other) to verify the necessity of the grouping. If the number of guaranteed data of one of the two independent groups is less than 40, the group is directly removed, because the necessity of the grouping cannot be judged due to the small number of data, and the service time of the water truck cannot be predicted on this basis. If the two crossover chi-square tests of the two independent groups both show that they significantly do not belong to the same empirical probability distribution at a confidence level of 99.9%, both independent groups are retained; otherwise, the two independent groups are merged into one group.

[0051] Step 3: Probability distribution fitting and testing of water truck service time

[0052] Based on step 1 and step 2, probability distribution fitting and testing are performed on the grouping results of step 2. The probability distribution fitting and testing of the service time of the water truck includes probability distribution fitting and probability distribution testing, which are two sub-steps in total.

[0053] 3.1 Probability distribution fitting

[0054] Generally speaking, the accuracy of the clean water truck service time in step 1 is 1 minute, so when the amount of data is sufficient, 1 minute should be used as the interval to count the frequency of the clean water truck service time. If the amount of data is small, resulting in the clean water truck service time frequency being 0 at most intervals, the interval should be appropriately expanded.

[0055] Using 20 common probability distributions in engineering, the service time of water trucks is fitted respectively, and the maximum likelihood method is used to solve the parameters of each probability distribution. The 20 common distributions in engineering include 4 categories: single-parameter probability distribution, two-parameter probability distribution, three-parameter probability distribution and four-parameter probability distribution. Single-parameter probability distribution includes: exponential distribution, Poisson distribution and Rayleigh distribution, a total of 3 types; two-parameter distribution includes: Birnbaum-Saunders distribution, extreme value distribution, gamma distribution, half-normal distribution, inverse Gaussian distribution, logical distribution, log-logistic distribution, log-normal distribution, upper-middle distribution, normal distribution, Rice distribution and Weibull distribution, a total of 12 types; three-parameter distribution includes: Burle distribution, generalized extreme value distribution, generalized Pareto distribution and t-location scale distribution, a total of 4 types; four-parameter distribution includes: stable distribution, a total of 1 type.

[0056] 3.2 Probability distribution test

[0057] Then, a chi-square test was performed on the fitting results of various probability distributions of each group, and the probability distribution with the highest goodness of fit was selected as the final probability distribution of the group.

[0058] Step 4 Prediction of water truck service time

[0059] Based on steps 1 to 3, the service time of a water truck for a certain guarantee data is predicted. The prediction of the service time of a water truck includes classification, initial prediction and final prediction, which has three sub-steps in total.

[0060] 4.1 Classification is based on the departure flight number, aircraft type, aircraft type category, departure airport, destination airport, planned departure time and start time of the fresh water truck of the support data to be predicted. If the attributes of a group in step 3 meet all the attributes of the support data, the support data is classified into the group and the group is used as one of the candidate groups of the support data.

[0061] 4.2 The initial prediction is to perform an initial prediction calculation on the clean water truck service time of each candidate group of the guarantee data according to the preset prediction accuracy. The initial prediction calculation of a candidate group is shown in formula (2): d represents the correct prediction probability value under the preset prediction accuracy, b represents the lower limit of the predicted clean water truck service time, Δc represents the preset accuracy of the predicted clean water truck service time, b+Δc represents the upper limit of the predicted clean water truck service time, and p(x) represents the probability distribution function of the candidate group. In order to improve the accuracy of the prediction, it is necessary to find the maximum value of d. The derivation process is shown in formula (3): When d b When ′ is always greater than 0, b takes the value of its maximum value in the domain minus Δc; when d b ′ is always less than 0, b takes the minimum value of its domain; when d b When ' is not always greater than and not always less than 0, b takes the value of d b The maximum value of d when ′ is equal to 0.

[0062]

[0063]

[0064] 4.3 The final prediction is based on the initial prediction of all candidate groups of the guarantee data, and the expectation is calculated to obtain the final prediction result. According to the preset accuracy Δc, the upper and lower limits of the final prediction are solved as shown in formula (4): F 0 and F 1 Represent the lower and upper limits of the final prediction, respectively. and They represent the lower and upper limits of the initial prediction of the jth candidate group of the guarantee data, respectively. jThey represent the chi-square test significant correlation probability of the initial predicted probability distribution of the jth candidate grouping of the guarantee data (regarded as the goodness of fit of the probability distribution), and k represents the number of candidate groups of the guarantee data. and Both So F 1 -F 0 =Δc, that is, the accuracy of the final prediction and the accuracy of the initial prediction are both Δc.

[0065]

[0066] Combination Figures 2 to 4 Taking some water truck support data of Beijing Daxing International Airport as an example, the following embodiment is described in detail. The method for predicting the service time of the water truck in this embodiment mainly includes the following steps:

[0067] Step 1: Preliminary statistics of water truck service time

[0068] Taking Beijing Daxing International Airport from September 2019 to May 2020 as an example, the common airport fresh water truck operation and support data are shown in Table 1, including: departure flight number, aircraft model, aircraft model category, departure airport, destination airport, planned departure time, actual departure time, fresh water truck start service time and fresh water truck end service time.

[0069] Table 1 Data format example

[0070]

[0071] The departure flight number of this security data is CA757, the aircraft model is B737, and the aircraft category is Class C; the departure airport is Beijing Daxing International Airport (ZBAD), and the destination airport is Thailand Don Mueang International Airport (VTBD). The "planned departure time" of this security data is the planned off-block time of the departure flight CA757 at the departure airport, and the value is 7:50 (time accuracy is 5 minutes). The "actual departure time" of this security data is the actual off-block time of the departure flight CA757 at the departure airport in actual operation, and the value is 7:51 (time accuracy is 1 minute). The "clean water truck start service time" and "clean water truck end service time" of this security data are the time points when the airport clean water truck actually starts to add clean water to the departure flight and ends the clean water addition operation in actual operation of the departure flight CA757, and the values ​​are 6:22 (time accuracy is 1 minute) and 6:27 (time accuracy is 1 minute) respectively.

[0072] 1) Data cleaning

[0073] Before data cleaning, the instance has 31,076 guaranteed data items. For each guaranteed data item, if the "clean water truck service start time" or "clean water truck service end time" of the data is in an incorrect format (including blank and non-time data format), the data item will be removed. After data cleaning, the instance has 15,933 valid guaranteed data items.

[0074] 2) Data calculation

[0075] Data calculation means that for each piece of guarantee data after data cleaning, the "clean water truck service end time" of the data is subtracted from the "clean water truck service start time" to obtain the "clean water truck service time" of the data. Taking Table 1 as an example, the clean water truck service time of this guarantee data is 5 minutes. Since the time accuracy of "clean water truck service start time" and "clean water truck service end time" is 1 minute, the time accuracy of "clean water truck service time" is also 1 minute.

[0076] Step 2: Grouping of water truck service time

[0077] 1) Determine the initial grouping

[0078] The purpose of determining the initial grouping attributes is to preliminarily determine one or more attributes of the support data used for grouping. According to step 1, the support data usually has 9 attributes. Except for "actual departure time" and "clean water truck end service time", the other 7 attributes can be determined as initial grouping attributes.

[0079] 2) Complexity screening

[0080] The purpose of complexity screening is to calculate the number of schemes to ensure data partitioning based on a single or multiple combinations of the initial grouping attributes determined above, and to screen the grouping method with a reasonable number of schemes.

[0081] Taking the single attribute of "machine model" as an example, the number of valid guaranteed data items in the output instance of step 1 is 15933, and there are 14 different machine models. According to formula (1), it can be concluded that there are 65535 grouping schemes, that is, the complexity is 65535. If the preset complexity threshold is 1000, the complexity of this attribute exceeds the threshold, so it is screened out.

[0082] Taking the single attribute of "machine model category" as an example, the number of valid guaranteed data items of the output instance in step 1 is 15933, and there are 5 different machine model categories. According to formula (1), it can be concluded that there are 31 grouping schemes, that is, the complexity is 31. If the preset complexity threshold is 1000, the complexity of this attribute does not exceed the threshold, so it is retained.

[0083] Taking the multiple attributes of "aircraft type" and "planned departure time (accuracy 1 hour, i.e., divided into 24 time slices such as 0 to 1 o'clock and 1 to 2 o'clock)" as an example, there are 5 different aircraft types and 24 different planned departure times. According to formula (1), there are 31×16777215=520093665 grouping schemes, i.e., the complexity is 520093665. If the preset complexity threshold is 1000, the complexity of this attribute exceeds the threshold, so it is screened out.

[0084] Taking the multi-attributes of "aircraft type" and "planned departure time" (with a precision of 6 hours, i.e., 0 to 6 o'clock, 6 to 12 o'clock, 12 to 18 o'clock, and 18 to 24 o'clock) as an example, there are 5 different aircraft types and 4 different planned departure times. According to formula (1), there are 31×15=465 grouping schemes, i.e., the complexity is 465. If the preset complexity threshold is 1000, the complexity of this attribute does not exceed the threshold, so it is retained.

[0085] 3) Determine the grouping

[0086] Determining the grouping is based on determining the initial grouping attributes and the complexity screening operation, the purpose of which is to ensure that the service time distribution of the water truck between independent groups is significantly different at a certain confidence level.

[0087] Taking the multi-attribute grouping of support data according to "type category" and "planned departure time" as an example, there are a total of 465 grouping schemes. First, among the 465 groupings, 30 groups have less than 40 support data, so they are directly removed, leaving 435 groupings. Using the chi-square test method, if the chi-square test of two crossovers of two independent groups (the two groups are theoretical values ​​or observed values ​​of each other) shows that they significantly do not belong to the same empirical probability distribution at a confidence level of 99.9%, then both independent groups will be retained; otherwise, the two independent groups will be merged into one group. After screening by the chi-square test method, 435 groupings were combined into 36 groups.

[0088] Step 3: Probability distribution fitting and testing of water truck service time

[0089] Based on steps 1 and 2, the probability distribution fitting and testing are performed on the grouping results of step 2. For the grouping results of step 2, 20 common probability distributions in engineering are used to fit the water truck service time data. The water truck service time of each group is fitted, and the maximum likelihood method is used to solve the parameters of each probability distribution. Figure 2 As shown, the significance of the chi-square test is ranked from high to low as follows Figure 3 shown. Figure 3The higher the significance of a probability distribution, the better the goodness of fit of the probability distribution. Figure 2 and Figure 3 In the group analysis, the 16th probability distribution (Burr distribution) with the highest goodness of fit was selected as the final probability distribution of the group.

[0090] Step 4: Prediction of water truck service time

[0091] 1) Classification

[0092] Taking the support data in Table 1 as an example, if there is a certain grouping method: for example, multi-attribute grouping based on "aircraft model category" and "planned departure time (accuracy 6 hours, i.e. 0:00 to 6:00, 6:00 to 12:00, 12:00 to 18:00 and 18:00 to 24:00)", then the support data should be classified into the group with "aircraft model category" as C and "planned departure time" as 6:00 to 12:00 and other groups (if the attributes of a group meet all the attributes of the support data).

[0093] 2) Initial prediction

[0094] According to the above classification results, the guarantee data in Table 1 has multiple candidate groups. According to formula (3), multiple initial prediction lower limits and upper limits can be output according to different preset precisions Δc.

[0095] 3) Final prediction

[0096] Based on the multiple lower limits and upper limits of the above initial predictions, the final prediction results under different preset accuracies Δc can be calculated according to formula (4).

[0097] According to the final predicted F 0 and F 1 (i.e., lower limit and upper limit), if the service time of the water truck of the guarantee data is greater than or equal to F 0 and less than or equal to F 1 , then the clean water truck service time prediction of the guarantee data with respect to the preset accuracy Δc is correct; if the clean water truck service time of the guarantee data is less than F 0 or greater than F 1 , then the forecast of the water truck service time of the guarantee data with respect to the preset accuracy Δc is wrong. According to the number of correct and incorrect predictions, the accuracy of the prediction can be output.

[0098] The water truck support data of Beijing Daxing International Airport in June 2020 was used as the prediction test data. The traditional water truck service time prediction method based on empirical probability distribution was first used for prediction, and then the prediction method of the present invention was used for prediction. The comparison of prediction accuracy is shown in Table 2 and Table 3. Figure 4 shown.

[0099] Table 2 Comparison of prediction accuracy between the present invention and traditional prediction methods

[0100]

[0101] The above embodiments are only for illustrating the technical idea of ​​the present invention, and cannot be used to limit the protection scope of the present invention. Any changes made on the basis of the technical solution in accordance with the technical idea proposed by the present invention shall fall within the protection scope of the present invention.

Claims

1. A method for predicting the service time of an airport water truck, characterized in that: The steps include: Step 1: For the airport to be predicted, obtain the water truck operation support data of the airport, including the departure flight number, aircraft model, aircraft model category, departure airport, destination airport, planned departure time, actual departure time, water truck start service time and water truck end service time, a total of 9 attributes, and clean and calculate the support data; Step 2: For the 9 attributes of the security data, the actual departure time and the determined time of the end of the clean water truck service time are both later than the start of the clean water truck service time. The remaining 7 attributes after removing these two attributes are used as the initial grouping attributes; Step 3: Combine the remaining 7 attributes, calculate the number of grouping schemes that guarantee data division under each combination, that is, the complexity, and set a complexity threshold. If the number of grouping schemes calculated for a certain combination is less than the complexity threshold, then retain the grouping scheme corresponding to the combination, otherwise screen out the grouping scheme corresponding to the combination; The number of grouping schemes to ensure data partition under each combination, i.e., the complexity, is calculated as follows: Where D represents the complexity of securing data partitioning when m attributes are selected from the remaining 7 attributes for combination, m represents the total number of all attributes used to calculate the complexity, n = 1, ..., m, a n Represents the total number of data after deduplication of the nth attribute, i = 1, ..., a n ; The attributes that need to be deduplicated are the planned departure time and the start time of the clean water truck. The planned departure time with a time accuracy of 5 minutes is converted to the planned departure time with a time accuracy of 1, 2, 3, 4, 6, 8, 12 or 24 hours. The clean water truck start time with a time accuracy of 1 minute is converted to the clean water truck start time with a time accuracy of 1, 2, 3, 4, 6, 8, 12 or 24 hours. Step 4: For the grouping schemes corresponding to the combinations retained in step 3, the attribute label of the combination is the attribute label of the corresponding grouping scheme. If the number of guaranteed data in a grouping scheme is less than 40, the grouping scheme is deleted. For the remaining grouping schemes, the chi-square test method is used to perform two cross-over chi-square tests. If there are two grouping schemes whose two cross-over chi-square tests are significantly not in the same empirical probability distribution at the 99.9% confidence level, both grouping schemes are retained; otherwise, the two grouping schemes are merged into one grouping scheme. Step 5: For all the grouping schemes obtained in step 4, perform probability distribution fitting and probability distribution test on each grouping scheme to obtain the final probability distribution of the grouping scheme; Step 6: For the guarantee data of the water truck service time to be predicted, if the guarantee data satisfies all attribute labels of a grouping scheme obtained in step 4, the grouping scheme is used as an alternative grouping of the guarantee data, thereby obtaining all the alternative groups; Step 7: For each candidate group, an initial prediction is made on the service time of the water cleaning truck to be predicted according to a preset prediction accuracy, and the lower limit and upper limit of the initial predicted service time of the water cleaning truck corresponding to the candidate group are obtained; The calculation formula for the initial prediction is: Wherein, d represents the prediction correct probability value under the preset prediction accuracy, b represents the predicted lower limit of the clean water truck service time, Δc represents the preset accuracy of the predicted clean water truck service time, b+Δc represents the predicted upper limit of the clean water truck service time, and p(x) represents the probability distribution function of the alternative grouping; When b When ′ is always greater than 0, b takes the value of its maximum value in the domain minus Δc; when d b ′ is always less than 0, b takes the minimum value of its domain; when d b When ' is not always greater than and not always less than 0, b takes the value of d b The maximum value of d when ′ is equal to 0; Step 8, according to the lower and upper limits of the initial predicted service time of the clean water truck corresponding to all the candidate groups, the expected final prediction result is obtained.

2. The method for predicting the service time of an airport water truck according to claim 1, characterized in that: The departure airport mentioned in step 1 is the airport to be predicted; the planned departure time is the planned time for the departure flight to withdraw the wheel block from the departure airport as reported in advance, with a time accuracy of 5 minutes; the actual departure time is the actual time for the departure flight to withdraw the wheel block from the departure airport in actual operation, with a time accuracy of 1 minute; the start service time of the fresh water truck and the end service time of the fresh water truck are the time points when the airport fresh water truck actually starts to fill fresh water and ends the fresh water filling operation for the departing flight in actual operation, and the time accuracy is 1 minute.

3. The method for predicting the service time of an airport water truck according to claim 1, characterized in that: Step 1 is to clean and calculate the security data, specifically: For each piece of security data, if at least one of the clean water truck service start time and the clean water truck service end time of the data is in an incorrect form, the data will be removed; for each piece of security data after cleaning, the clean water truck service start time will be subtracted from the clean water truck service end time of the data to obtain the clean water truck service time of the data, and the time accuracy of the clean water truck service time is 1 minute.

4. The method for predicting the service time of an airport water truck according to claim 1, characterized in that: In step 5, probability distribution fitting and probability distribution test are performed on each grouping scheme to obtain the final probability distribution of the grouping scheme, which is specifically: For a grouping scheme, 20 probability distributions are used to fit the service time of the water truck, and the maximum likelihood method is used to solve the parameters of each probability distribution. The 20 probability distributions include four categories: single-parameter probability distribution, two-parameter probability distribution, three-parameter probability distribution and four-parameter probability distribution. The single-parameter probability distributions include: exponential distribution, Poisson distribution and Rayleigh distribution; the two-parameter distributions include: Birnbaum-Saunders distribution, extreme value distribution, gamma distribution, half-normal distribution, inverse Gaussian distribution, logistic distribution, log-logistic distribution, log-normal distribution, upper-middle distribution, normal distribution, Rice distribution and Weibull distribution; the three-parameter distributions include: Burle distribution, generalized extreme value distribution, generalized Pareto distribution and t-location scale distribution; the four-parameter distributions include: stable distribution; a chi-square test was performed on the fitting results of the 20 probability distributions of the grouping scheme, and the probability distribution with the highest goodness of fit was selected as the final probability distribution of the grouping scheme.

5. The method for predicting the service time of an airport water truck according to claim 1, characterized in that: The calculation formula of step 8 is: Among them, F 0 and F 1 Respectively represent the lower and upper limits of the final prediction results, and They represent the lower and upper limits of the initial prediction of the jth candidate group, G j They respectively represent the goodness of fit of the probability distribution corresponding to the j-th alternative grouping, and k represents the number of all alternative groups.

Images