Aircraft fuel consumption prediction method, device and medium based on QAR data

Abstract

The present invention provides a method, device, and medium for predicting aircraft fuel consumption based on QAR data, including: preprocessing the acquired raw QAR data; obtaining altitude data from the preprocessed QAR data, and obtaining the corresponding atmospheric temperature, air density, flight phase, and bank angle; processing outliers in the QAR data for each flight phase; obtaining the force coefficient and flight force corresponding to the QAR data; obtaining characteristic parameters of fuel consumption; and training a set aircraft fuel consumption prediction model using data corresponding to the characteristic parameters, obtaining the trained aircraft fuel consumption prediction model as a target aircraft fuel consumption prediction model. The present invention can improve the prediction accuracy of aircraft fuel consumption.

Description

Technical Field

[0001] The present invention relates to the field of aircraft fuel consumption prediction, and in particular to a method, device and medium for aircraft fuel consumption prediction based on QAR data. Background Art

[0002] At present, the prediction of aircraft fuel consumption mostly relies on empirical formulas and fixed parameter models, which do not fully consider the variable environmental factors and operating conditions during the flight process, and the prediction accuracy is low. With the widespread application of QAR (Quick Access Recorder) data, it provides a possibility for data-driven methods. However, how to effectively utilize these data to improve the accuracy of fuel consumption prediction is still a problem that needs to be solved. Patent document (CN 114154708A) discloses a method for predicting aircraft fuel consumption, including: using correlation analysis method to extract parameters with high correlation with flight fuel consumption as input and using grid search method to determine important parameters of random forest algorithm. This method can improve the prediction accuracy of aircraft fuel consumption. However, another method for predicting aircraft fuel consumption can also be provided. Summary of the Invention

[0003] In view of the above technical problems, the technical solution adopted by the present invention is:

[0004] According to a first aspect of the present invention, a method for predicting aircraft fuel consumption based on QAR data is provided, the method comprising the following steps:

[0005] S100, preprocessing the acquired original QAR data to obtain preprocessed QAR data.

[0006] S200, obtain the altitude data in the preprocessed QAR data, obtain the corresponding atmospheric temperature and air density, and add the obtained atmospheric temperature and air density to the first original QAR data to obtain the first QAR data, and obtain the flight phase and bank angle corresponding to the first QAR data.

[0007] S300, based on the determined m flight phases, the first QAR data is divided into m QAR data segments, and the outliers in each QAR data segment are processed to obtain the first QAR data with outlier processing as the second QAR data, and the obtained flight phase and bank angle are added to the second QAR data to obtain the third QAR data.

[0008] S400, based on the aircraft aerodynamic model and the third QAR data, obtain the force coefficient corresponding to the third QAR data, and based on the thrust model and the third QAR data, obtain the flight force corresponding to the third QAR data, and add the obtained force coefficient and flight force to the third QAR data to obtain fourth QAR data.

[0009] S500, obtaining fuel consumption data from the fourth QAR data and determining candidate parameters related to the fuel consumption data from corresponding parameters, and obtaining correlations between the candidate parameters and fuel consumption, and using the candidate parameters having correlations with the fuel consumption greater than a set correlation as characteristic parameters of the fuel consumption.

[0010] S600 , training a set aircraft fuel consumption prediction model using data corresponding to the characteristic parameters, and obtaining a trained aircraft fuel consumption prediction model as a target aircraft fuel consumption prediction model.

[0011] According to a second aspect of the present invention, an electronic device is provided, comprising a processor and a memory; the processor is configured to execute the steps of the method according to the first aspect of the present invention by calling a program or instruction stored in the memory.

[0012] According to a second aspect of the present invention, there is provided a non-transitory computer-readable storage medium storing a program or instructions for enabling a computer to execute the steps of the method according to the first aspect of the present invention.

[0013] The present invention has at least the following beneficial effects:

[0014] The aircraft fuel consumption prediction method based on QAR data provided in an embodiment of the present invention processes the original QAR data to obtain multiple candidate parameters related to fuel consumption, then obtains parameters highly correlated with fuel consumption from the multiple candidate parameters as feature parameters, and then uses the data corresponding to these feature parameters to train the prediction model, which can improve the accuracy of the prediction model.

[0015] It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present invention, nor is it intended to limit the scope of the present invention. Other features of the present invention will become readily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0017] Figure 1 A flowchart of a method for predicting aircraft fuel consumption based on QAR data provided by an embodiment of the present invention;

[0018] Figure 2 Schematic diagram of fuel flow in QAR data. DETAILED DESCRIPTION

[0019] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making any creative efforts shall fall within the scope of protection of the present invention.

[0020] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which this invention pertains. The terms used herein in the specification of the present invention are for the purpose of describing specific embodiments only and are not intended to limit the present invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0021] It should be noted that some exemplary embodiments are described as processes or methods depicted as flowcharts. Although the flowcharts describe the steps as sequential processes, many of the steps can be performed in parallel, concurrently, or simultaneously. In addition, the order of the steps can be rearranged. A process can be terminated when its operation is completed, but can also have additional steps not included in the accompanying drawings. A process can correspond to a method, function, procedure, subroutine, subprogram, etc.

[0022] The embodiment of the present invention provides a method for predicting aircraft fuel consumption based on QAR data, such as Figure 1 As shown, the method may include the following steps:

[0023] S100 , preprocessing the acquired original QAR data to obtain preprocessed QAR data as preprocessed QAR data.

[0024] In an embodiment of the present invention, the aircraft may be an airplane. The original QAR data may be obtained by an airline and may include changes in multiple parameters of a complete flight cycle, such as true airspeed, wind speed, flight time, wind direction, ground speed, longitude and latitude of the aircraft, Mach number, altitude, and true fuel flow.

[0025] In an exemplary embodiment of the present invention, the fuel flow rate in the QAR data used can be as follows: Figure 2 shown.

[0026] Furthermore, preprocessing of the original QAR data may include data cleaning, format conversion, missing value processing, and unit unification.

[0027] In an embodiment of the present invention, data cleaning may include removing duplicate records and correcting obvious erroneous data. For example, removing the content of duplicate erroneous record columns about time. Format conversion refers to unifying the format of data from different sources to facilitate subsequent processing. For example, time data is unified into a 24-hour system. Missing value processing refers to filling in missing data with interpolation. For example, for QAR data provided by a third party, the default value of missing data in the WIN_DIR, FF, GW, and TAS columns is 0. Unit unification refers to converting units into international standard units to ensure that all data units conform to international standard units. For example, the units of flight speed and altitude are specified as m / s and m:

[0028] S200, obtain the altitude data in the preprocessed QAR data, obtain the corresponding atmospheric temperature and air density, and add the obtained atmospheric temperature and air density to the first original QAR data to obtain the first QAR data, and obtain the flight phase and bank angle corresponding to the first QAR data.

[0029] In an embodiment of the present invention, the atmospheric temperature and air density can be obtained based on aerodynamic formulas. Specifically, the atmospheric temperature T can meet the following conditions:

[0030] If the height h≤11000, T=288.15-0.0065×h, if 11000<h≤20000, T=216.65, if h>20000, T=216.65+0.001×(h-20000).

[0031] The air density ρ can satisfy the following conditions:

[0032] If height h≤11000, ρ=1.225×(T / 288.15) 4.25588 , if 11000<h≤20000, ρ=0.36392×e (11000-h) / 6341.62, if h>20000, ρ=0.088035×(T / 216.65) -35.1632 .

[0033] It is known to those skilled in the art that for the atmospheric temperature and air density corresponding to each recording moment, the altitude used is the altitude corresponding to the recording moment.

[0034] In the embodiment of the present invention, the temperature and density of the atmospheric environment are calculated for the flight altitude layer, which can accurately reflect the flight conditions of the aircraft at different altitudes.

[0035] In the embodiment of the present invention, the flight phase of the aircraft is determined based on the altitude change rate of the aircraft and the elevation of the take-off and landing airport, and the bank angle is specified accordingly.

[0036] Furthermore, in S200, the flight phase FP of the aircraft corresponding to the first QAR data at the i-th recording moment i Obtained through the following steps:

[0037] S201, if the height h corresponding to the i-th recording moment i Satisfy the first condition and determine FP i For the ground stage; if h i Satisfy the second condition and determine FP i For the approach phase; if h i Satisfy the third condition and determine FP i For the landing phase; if h i If any of the first, second and third conditions is not met, execute S202; wherein the first condition is: h i =TOA i or h i =LA i , the second condition is: if 200m<|h i -TOA i |≤1000m, the third condition is: |h i -LA i |≤50m; among which, TOA i The elevation of the take-off airport of the aircraft corresponding to the first QAR data, LA i is the elevation of the landing airport corresponding to the first QAR data.

[0038] In the embodiment of the present invention, the value of i is 1 to n, where n is the number of recording moments corresponding to the first QAR data, and the time interval between two adjacent recording moments is the set time interval, which can be set based on actual needs, for example, 1s.

[0039] S202, obtaining the height difference △h i =hi+tp -h i , if △h i ≥500m, confirm FP i For the climbing stage, if |△h i |<500m, confirm FP i For the level flight stage, if △h i <-500m, confirm FP i It is a descending phase, where tp is the number of set time intervals, which can be determined based on the set duration t0. Specifically, tp = (t0 - Δt) / Δt, where Δt is the duration of the set time interval. For example, if Δt = 1s and t0 = 60s, then tp = 59.

[0040] In the embodiment of the present invention, the flight phases are divided by comparing the flight altitude with the take-off / landing airport elevation and the altitude change rate per minute, so that the division of the flight phases is more refined and the fuel consumption changes during the flight can be accurately captured.

[0041] Furthermore, in S200, the bank angle BA of the aircraft corresponding to the first QAR data at the i-th recording time i The following conditions are met:

[0042] If FP i For the climbing or descending phase, BA i =angle1, if FP i During the approach or landing phase, BA i =angle2, if FP i For the ground phase or level flight phase, BA i =angle3, angle1 is the first preset angle, angle2 is the second preset angle, and angle3 is the third preset angle.

[0043] In the embodiment of the present invention, angle1, angle2 and angle3 may be empirical values. In an exemplary embodiment, angle1 = 10°, angle2 = 15°, and angle3 = 0°.

[0044] In this embodiment of the present invention, a bank angle is specified to more accurately identify factors affecting fuel consumption. S300 : Based on the m determined flight phases, the first QAR data is divided into m QAR data segments, and outliers in each QAR data segment are processed to obtain the outlier-processed first QAR data as second QAR data. The acquired flight phase and bank angle are then added to the second QAR data to obtain third QAR data.

[0045] Furthermore, in S300, the outliers in the j-th QAR data segment may be processed by the following steps:

[0046] S301: For the parameter r in the QAR data, sort the data corresponding to the parameter r in the jth QAR data segment in ascending order to obtain the sorted data D of the parameter r. jr ; where j ranges from 1 to m, r ranges from 1 to q, and q is the number of parameters in the QAR data.

[0047] S302, obtain D jr The k1th data d in jr k1 and D rj The k2th data d in jr k2 , and based on d jr k1 and d jr k2 Get the abnormal baseline value IQR jr =d jr k2 -d jr k1 ; Among them, k1=[f(j, r)×0.25], k2=[f(j, r)×0.75], [] means rounding, f(j, r) is D jr The amount of data in .

[0048] S303, traverse D jr , if D jr The sth data d in jr s <(d jr k1 -c×IQR jr ), or d jr s >(d jr k2 +c×IQR jr ), d jr s The value of s ranges from 1 to f(j, r), and the initial value of s is 1.

[0049] S304, using the replacement value h jr Replace D jr The outliers in h jr =1 / (f(j,r)-2×np(j,r))∑ f(j，r)-np(j，r) g＝np(j，r) d jr g, where np(j, r) = f(j, r) × d, d is a preset coefficient, 0<d<1.

[0050] In an exemplary embodiment of the present invention, d=0.05.

[0051] S305, set r=r+1, if r≤q, execute S301, otherwise, exit the current control program, and obtain the j-th QAR data segment after outlier processing.

[0052] In the embodiment of the present invention, the flight phase is converted into a dummy variable. Specifically, the flight phase is converted into a dummy variable in the following manner:

[0053] If the flight phase is the climbing phase, the corresponding dummy variable is 1; if the flight phase is the level flight phase, the corresponding dummy variable is 2; if the flight phase is the descending phase, the corresponding dummy variable is 3; if the flight phase is the approach phase, the corresponding dummy variable is 4; if the flight phase is the landing phase, the corresponding dummy variable is 5; if the flight phase is the ground phase, the corresponding dummy variable is 6.

[0054] S400, based on the aircraft aerodynamic model and the third QAR data, obtain the force coefficient corresponding to the third QAR data, and based on the thrust model and the third QAR data, obtain the flight force corresponding to the third QAR data, and add the obtained force coefficient and flight force to the third QAR data to obtain fourth QAR data.

[0055] In the embodiment of the present invention, the force coefficient may include the lift coefficient C L and the drag coefficient C D .

[0056] Among them, C L =2mg / ρv 2 Scos(BA), where m is the mass of the aircraft, v is the flight speed of the aircraft, g is the acceleration due to gravity, ρ is the air density, S is the wing area, and BA is the bank angle.

[0057] Those skilled in the art know that for the lift coefficient corresponding to each recording moment, the flight speed, air density, and bank angle used are the flight speed, air density, and bank angle corresponding to the recording moment.

[0058] C D The following conditions are met:

[0059] If C D0，AP 、C D2，AP 、C D0，LDG 、C D0，△LDG and C D2，LDG Both are 0, C D =CD0，CR +C D2，CR ×C L 2 , if C D0，AP 、C D2，AP 、C D0，LDG 、C D0，△LDG and C D2，LDG All are not 0, C D =C D1 .

[0060] Among them, C D0，A is the parasite drag coefficient during the approach phase, C D2，AP is the induced drag coefficient during the approach phase, C D0，LDG is the parasitic drag coefficient during landing, C D0，△LDG is the change in drag coefficient during landing, C D2，LDG is the induced drag coefficient during landing phase, C D0，CR is the parasitic drag coefficient during the cruise phase, C D2，CR is the induced drag coefficient during the cruise phase. C D1 is an intermediate quantity. If the flight phase is the approach phase, C D1 =C D0，AP +C D0，AP ×C L 2 If the flight phase is landing phase, C D1 =C D0，LDG +C D0，△LDG +C D0，LDG ×C L 2 If the flight phase is neither approach nor landing, C D =C D0，CR +C D2，CR ×C L 2 .

[0061] In the embodiment of the present invention, C D0，AP 、C D2，AP 、C D0，LDG 、C D0，△LDG 、C D2，LDG 、C D0，△LDG and C D2，LDG This can be obtained by consulting the aircraft's performance manual.

[0062] In the embodiment of the present invention, the flight force includes resistance D, thrust T, maximum climbing thrust T maxc , cruise thrust T cruise , Maximum cruise thrust T max，cruise , descending thrust T des and idle thrust T idle .

[0063] Among them, T cruise =D=C D ρv 2 S.

[0064] T satisfies the following conditions:

[0065] If it is the climbing stage, T=T maxc If it is level flight, T = T cruise If it is the ground phase and the speed is not equal to 0, T = T idle If it is the descent phase, landing phase, approach phase, or ground phase with a speed of 0, T = T des .

[0066] Among them, T maxc =C (△ISA)eff ×T maxcs , C (△ISA)eff is the climb thrust correction factor, which can be obtained by referring to the aircraft performance manual. maxcs is the maximum standard climb thrust, T maxcs The following conditions are met:

[0067] If the engine type is a jet engine, T maxcs =C Tc1 (1-h / C Tc2 )+C Tc3 h 2 ; C Tc1 is the first thrust coefficient, C Tc2 is the second thrust coefficient, C Tc3 The first to third thrust coefficients can be obtained by referring to the performance of the aircraft.

[0068] If the engine type is a turbine engine, T maxcs =(C Tc1 (1-h / C Tc2 )) / v+C Tc3 ; v is the flight speed.

[0069] If the engine type is a piston engine, T maxcs =C Tc1 (1-h / C Tc2 )+C Tc3 / 3.

[0070] T idle =C △idle ×C Tc1 (1-h / C Tc2 )+C Tc3 h 2 . C △idleis the idle thrust correction coefficient, which can be an empirical value, for example, 0.03.

[0071] T max，cruise =C T,Cr ×T cruise , C T,Cr is the cruise thrust correction factor, which can be obtained based on the aircraft performance manual.

[0072] T des =C def ×T maxc , C def is the descent thrust correction factor, which can be obtained based on the aircraft performance manual.

[0073] S500, obtaining fuel consumption data from the fourth QAR data and determining candidate parameters related to the fuel consumption data from corresponding parameters, and obtaining correlations between the candidate parameters and fuel consumption, and using the candidate parameters having correlations with the fuel consumption greater than a set correlation as characteristic parameters of the fuel consumption.

[0074] In an embodiment of the present invention, the candidate parameters may be determined based on expert experience. In an embodiment of the present invention, all parameters in the fourth QAR data except the flight time, wind direction, longitude coordinates and latitude coordinates may be used as candidate parameters.

[0075] Furthermore, S500 may specifically include:

[0076] S501 : Perform dimensionless processing on the fuel consumption data and the data corresponding to each candidate parameter in the fourth QAR data to obtain the dimensionless processed fourth QAR data.

[0077] In the embodiment of the present invention, a minimum-maximum normalization method may be used to perform dimensionless processing on the fourth QAR data.

[0078] S502, based on the fourth QAR data after dimensionless processing, construct a data matrix R of size z×n, and calculate the covariance matrix CR of the data matrix R; wherein n is the number of recorded moments in the fourth QAR data, and z is the number of candidate parameters in the fourth QAR data.

[0079] In the embodiment of the present invention, n=9707, z=11.

[0080] S503, obtaining the characteristic value of each row of data in the CRCR, and obtaining the characteristic value corresponding to each candidate parameter.

[0081] In an embodiment of the present invention, CR may be eigendecomposed to obtain the eigenvalue corresponding to each candidate parameter, and the eigenvalue λ of the candidate parameter may satisfy the following condition: det(CR-λI)=0, det() represents the determinant, and I is the unit matrix.

[0082] S504, obtaining the weight w corresponding to the e-th candidate parameter e =λ e / ∑ q h＝1 λ h , and obtain the first correlation S1 between the e-th parameter and fuel consumption e =Rc×λ e ; Among them, λ e is the eigenvalue of the e-th candidate parameter, λ h is the eigenvalue of the hth candidate parameter, the value of e ranges from 1 to z, and Rc is the matrix obtained by centering R, that is, Rc = R-mean(R), where mean() represents the average value.

[0083] S505. Sort the weights of the z candidate parameters in descending order to obtain a sorted candidate parameter weight set, and obtain the candidate parameters corresponding to the first u weights in the sorted candidate parameter weight set as reference parameters, where u satisfies the following conditions: the sum of the first u-1 weights in the sorted candidate parameter weight set is less than 1, and the sum of the first u+1 weights in the sorted candidate parameter weight set is greater than or equal to 1.

[0084] S506, obtaining a second correlation S2 between the e-th parameter and fuel consumption e =(1 / n)∑ n v＝1 C ev , C ev is the correlation coefficient between the e-th candidate parameter and the fuel consumption at the v-th recording time, C ev =(a+x×b) / (|d0(v)-d e (v)|+x×b), a and b are intermediate quantities, a is the minimum value in the matrix R, b is the maximum value in the matrix R, where the size of the matrix R is n×z, and the element corresponding to the vth row and the eth column in R is |d0(v)-d e (v)|, d0(v) is the fuel consumption value corresponding to the vth recording moment, d e (v) is the value of the e-th candidate parameter at the v-th recording time, x is a setting coefficient, 0<x<1, and in an exemplary embodiment, x=0.5.

[0085] S507, obtaining the correlation S between the e-th parameter and fuel consumption e =β×|S1e |+(1-β)×|S2 e |, β is a setting coefficient, 0<β<1, in an exemplary embodiment, β=0.3.

[0086] S508 , sorting the correlations between the z candidate parameters and the fuel consumption in descending order, and obtaining candidate parameters corresponding to correlations greater than or equal to a set correlation threshold among the sorted correlations as characteristic parameters of the fuel consumption.

[0087] In an embodiment of the present invention, the setting of the correlation threshold is determined based on the reference parameter. In an exemplary embodiment, the setting of the correlation threshold is set so that the acquired feature parameters include the reference parameter. Figure 2 In the corresponding data, the correlation threshold is 0.5.

[0088] In an embodiment of the present invention, during the feature parameter extraction stage, the weighted values ​​of the correlations obtained by the two methods are used as the criterion for extracting feature parameters, which can more effectively screen out feature parameters that are highly correlated with fuel consumption and provide higher-quality input data for subsequent model training.

[0089] S600 , training a set aircraft fuel consumption prediction model using data corresponding to the characteristic parameters, and obtaining a trained aircraft fuel consumption prediction model as a target aircraft fuel consumption prediction model.

[0090] In an embodiment of the present invention, the aircraft fuel consumption prediction model is set to be a bidirectional long short-term memory network (BiLSTM) model, which can capture time series information during the flight process. Compared with traditional machine learning models, BiLSTM has stronger expression ability and prediction accuracy in processing time series data.

[0091] In this embodiment of the present invention, the BiLSTM model primarily consists of two BiLSTM layers. The first BiLSTM layer sets the return_sequences parameter to True. Each BiLSTM layer is followed by a fully connected layer. The BiLSTM model uses the ReLU activation function for nonlinear learning of fuel flow and uses the Adam optimizer to automatically adjust the learning rate. Furthermore, the time window size of the BiLSTM model can be 20.

[0092] Furthermore, in an embodiment of the present invention, the hyperparameters for setting the aircraft fuel consumption prediction model are obtained by the following steps:

[0093] S10, setting the first counter N1=1.

[0094] S11, if N1≤N10, set the second counter N2=1, execute S12, if N1>N10, execute S17; N10 is the number of set hyper parameters.

[0095] In an embodiment of the present invention, hyperparameters of the aircraft fuel consumption prediction model may include time step, cross-validation fold, number of iterations, initial learning rate, number of layer channels, etc. N10 can be set based on actual needs and can be determined based on a set hyperparameter search space, that is, a combination of hyperparameter search spaces. In an exemplary embodiment of the present invention, the hyperparameter search space can be as shown in Table 1 below:

[0096] Table 1

[0097] Hyperparameters Search Space Time step 10，20，30 Cross validation folds 3，5，10 Number of iterations 10，20，50 Initial learning rate 0.001，0.002，0.004 Number of layer channels 10，20，40

[0098] According to the search space in Table 1, it can be seen that N10=3×3×3×3×3=243.

[0099] S12, if N2≤N20, randomly divide the data corresponding to the characteristic parameters into N30 data subsets of the same size, set the third counter N3=1, and execute S13. If N2>N20, execute S16; N20 is the number of cross-validations set, which can be set based on actual needs, for example, N20=3.

[0100] S13, if N3≤N30, use the N3th data subset among the N30 data subsets as the validation set, and the remaining N30-1 data subsets as the training set, and execute S14, otherwise, execute S15; N30 is the set cross-validation fold, which can be set based on actual needs, for example, N30=10.

[0101] S14, using the training set to train the N1th set aircraft fuel consumption prediction model, to obtain the trained N1th set aircraft fuel consumption prediction model as the current aircraft fuel consumption prediction model M N1N2 N3 , and use the validation set to test the current aircraft fuel consumption prediction model M N1N2 N3 Verify and get the corresponding verification result RS N1N2 N3 ; Execute S15, set N3=N3+1, execute S13; wherein, the N1th set hyperparameter of the aircraft fuel consumption prediction model is the N1th set hyperparameter.

[0102] Those skilled in the art know that the method of training a model using a training set and the method of verifying a model using a verification set are both prior arts.

[0103] S15, obtain the average value AV of the verification result corresponding to the N2th cross-validation under the N1th set hyperparameters N1N2 =(RS N1N2 1 +RS N1N2 2 +……+RS N1N2 N3 +……+RS N1N2 N30 ) / N30, set N2=N2+1, and execute S12.

[0104] S16, obtain the average value AV of the verification results under the N1th set hyperparameters N1 =(AV N11 +AV N12 +……+AV N1N2 +……+AV N1N20 ) / N20, set N1=N1+1, and execute S11;

[0105] S17, obtain max(AV1, AV2, ..., AV N1 ,……,AV N10 ) is used as the hyperparameter of the aircraft fuel consumption prediction model, and max() represents the maximum value.

[0106] In this embodiment of the present invention, taking the search space shown in Table 1 as an example, the hyperparameters of the aircraft fuel consumption prediction model are shown in Table 2 below:

[0107] Table 2

[0108] Hyperparameters value Time step 20 Cross validation folds 10 Number of iterations 20 Initial learning rate 0.004 Number of channels in the first layer 20 Second layer channel number 10

[0109] In the embodiment of the present invention, the search space and multiple K-fold cross-validation methods are used to optimize the hyperparameters of the BiLSTM model, which not only improves the generalization ability of the model but also ensures the stability of the model on different data sets.

[0110] Those skilled in the art will appreciate that any method of using the data corresponding to the characteristic parameters to train a set aircraft fuel consumption prediction model and obtaining the trained aircraft fuel consumption prediction model as a target aircraft fuel consumption prediction model falls within the scope of protection of the present invention.

[0111] The embodiment of the present invention is based on Figure 2 The corresponding data was used as a test set to test the aircraft fuel consumption prediction model provided by the present invention and the fuel calculation model currently used by airlines. The test results are shown in Table 3 below:

[0112] Table 3

[0113] Error rate Fuel calculation model This model average value 17.67 9.04 Maximum 48.24 28.11 Minimum 8.37 0.08

[0114] As can be seen from Table 3, the error rate of the aircraft fuel consumption prediction model provided by the embodiment of the present invention is lower than the error rate of the fuel calculation model, indicating that the aircraft fuel consumption prediction model provided by the embodiment of the present invention can achieve higher prediction accuracy.

[0115] An embodiment of the present invention also provides an electronic device, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are configured to execute the method described in the embodiment of the present invention.

[0116] An embodiment of the present invention further provides a non-transitory computer-readable storage medium storing computer-executable instructions, wherein the computer instructions are used to execute the method described in the embodiment of the present invention.

[0117] It should be understood that the various forms of the processes shown above can be used to reorder, add, or delete steps. For example, the steps described in the present invention can be performed in parallel, sequentially, or in a different order, as long as the desired results of the technical solutions disclosed in the present invention can be achieved. This is not limited herein.

[0118] The above specific embodiments do not limit the scope of protection of the present invention. Those skilled in the art will appreciate that various modifications, combinations, sub-combinations, and substitutions may be made based on design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention are intended to be included within the scope of protection of the present invention.

Claims

1. A method for predicting aircraft fuel consumption based on QAR data, characterized in that: The method comprises the following steps: S100, preprocessing the acquired original QAR data to obtain preprocessed QAR data; S200: Acquire altitude data from the preprocessed QAR data, acquire corresponding atmospheric temperature and air density, add the acquired atmospheric temperature and air density to the first original QAR data to obtain first QAR data, and acquire the flight phase and bank angle corresponding to the first QAR data; S300: Divide the first QAR data into m QAR data segments based on the determined m flight phases, process outliers in each QAR data segment, and obtain first QAR data subjected to outlier processing as second QAR data. Add the obtained flight phase and bank angle to the second QAR data to obtain third QAR data. S400: Obtaining a force coefficient corresponding to the third QAR data based on an aircraft aerodynamic model and the third QAR data, and obtaining a flight force corresponding to the third QAR data based on a thrust model and the third QAR data, and adding the obtained force coefficient and flight force to the third QAR data to obtain fourth QAR data. S500, acquiring fuel consumption data from the fourth QAR data, determining candidate parameters related to the fuel consumption data from corresponding parameters, and obtaining correlations between the candidate parameters and the fuel consumption, and selecting a candidate parameter having a correlation with the fuel consumption greater than a set correlation as a characteristic parameter of the fuel consumption; S600 , training a set aircraft fuel consumption prediction model using data corresponding to the characteristic parameters, and obtaining a trained aircraft fuel consumption prediction model as a target aircraft fuel consumption prediction model.

2. The method according to claim 1, characterized in that In S200, the flight phase FP of the aircraft corresponding to the first QAR data at the i-th recording time i Obtained through the following steps: S201, if the height h corresponding to the i-th recording moment i Satisfy the first condition and determine FP i For the ground stage; if h i Satisfy the second condition and determine FP i For the approach phase; if h i Satisfy the third condition and determine FP i For the landing phase; if h i If any of the first, second and third conditions is not met, execute S202; wherein the first condition is: h i =TOA i or h i =LA i , the second condition is: if 200m<|h i -TOA i |≤1000m, the third condition is: |h i -LA i |≤50m; among which, TOA i The elevation of the take-off airport of the aircraft corresponding to the first QAR data, LA i is the elevation of the landing airport corresponding to the first QAR data; S202, obtaining the height difference △h i =h i+tp -h i , if △h i ≥500m, confirm FP i For the climbing stage, if |△h i |<500m, confirm FP i For the level flight stage, if △h i <-500m, confirm FP i is the descending phase, where tp is the number of set time intervals; The value of i ranges from 1 to n, where n is the number of recording moments corresponding to the first QAR data, and the time interval between two adjacent recording moments is the set time interval.

3. The method according to claim 2, characterized in that In S200, the bank angle BA of the aircraft corresponding to the first QAR data at the i-th recording time i The following conditions are met: If FP i For the climbing or descending phase, BA i =angle1, if FP i During the approach or landing phase, BA i =angle2, if FP i For the ground phase or level flight phase, BA i =angle3, angle1 is the first preset angle, angle2 is the second preset angle, and angle3 is the third preset angle.

4. The method according to claim 1, wherein In S300, the outliers in the j-th QAR data segment are processed by the following steps: S301: For the parameter r in the QAR data, sort the data corresponding to the parameter r in the jth QAR data segment in ascending order to obtain the sorted data D of the parameter r. jr ; where j ranges from 1 to m, r ranges from 1 to q, and q is the number of parameters in the QAR data; S302, obtain D jr The k1th data d in jr k1 and D rj The k2th data d in jr k2 , and based on d jr k1 and d jr k2 Get the abnormal baseline value IQR jr =d jr k2 -d jr k1 ; Among them, k1=[f(j, r)×0.25], k2=[f(j, r)×0.75], [] means rounding, f(j, r) is D jr The amount of data in S303, traverse D jr , if D jr The sth data d in jr s <(d jr k1 -c×IQR jr ), or d jr s >(d jr k2 +c×IQR jr ), d jr s Marked as an outlier; c is a preset coefficient, s ranges from 1 to f(j, r), and the initial value of s is 1; S304, using the replacement value h jr Replace D jr Outliers in, where h jr =1 / (f(j,r)-2×np(j,r))∑ f(j，r)-np(j，r) g＝np(j，r) d jr g , where np(j, r) = f(j, r) × d, d is a preset coefficient, 0<d<1; S305, set r=r+1, if r≤q, execute S301, otherwise, exit the current control program, and obtain the j-th QAR data segment after outlier processing.

5. The method according to claim 1, wherein S500 specifically includes: S501, performing dimensionless processing on the fuel consumption data and the data corresponding to each candidate parameter in the fourth QAR data to obtain dimensionless processed fourth QAR data; S502: constructing a data matrix R of size z×n based on the dimensionless fourth QAR data, and calculating a covariance matrix CR of the data matrix R; wherein n is the number of recorded moments in the fourth QAR data, and z is the number of candidate parameters in the fourth QAR data; S503, obtaining the characteristic value of each row of data in the CRCR, and obtaining the characteristic value corresponding to each candidate parameter; S504, obtaining the weight w corresponding to the e-th candidate parameter e =λ e / ∑ q h＝1 λ h , and obtain the first correlation S1 between the e-th parameter and fuel consumption e =Rc×λ e ; Among them, λ e is the eigenvalue of the e-th candidate parameter, λ h is the eigenvalue of the hth candidate parameter, the value of e ranges from 1 to z, and Rc is the matrix obtained by centering R; S505: sort the weights of the z candidate parameters in descending order of weight to obtain a sorted candidate parameter weight set, and obtain the candidate parameters corresponding to the first u weights in the sorted candidate parameter weight set as reference parameters, where u satisfies the following conditions: the sum of the first u-1 weights in the sorted candidate parameter weight set is less than 1, and the sum of the first u+1 weights in the sorted candidate parameter weight set is greater than or equal to 1; S506, obtaining a second correlation S2 between the e-th parameter and fuel consumption e =(1 / n)∑ n v＝1 C ev , C ev is the correlation coefficient between the e-th candidate parameter and the fuel consumption at the v-th recording time, C ev =(a+x×b) / (|d0(v)-d e (v)|+x×b), a and b are intermediate quantities, a is the minimum value in the matrix R, b is the maximum value in the matrix R, where the size of the matrix R is n×z, and the element corresponding to the vth row and the eth column in R is |d0(v)-d e (v)|, d0(v) is the fuel consumption value corresponding to the vth recording moment, d e (v) is the value of the e-th candidate parameter at the v-th recording time, x is the setting coefficient, 0<x<1; S507, obtaining the correlation S between the e-th parameter and fuel consumption e =β×|S1 e |+(1-β)×|S2 e |, β is the setting coefficient, 0<β<1; S508, sorting the correlations between the z candidate parameters and the fuel consumption in descending order, and obtaining the candidate parameters corresponding to correlations greater than or equal to a set correlation threshold among the sorted correlations as characteristic parameters of the fuel consumption, wherein the set correlation threshold is determined based on the reference parameter.

6. The method according to claim 1, characterized in that The aircraft fuel consumption prediction model is set to be a BiLSTM model.

7. The method according to claim 1, characterized in that The flight forces include resistance D, thrust T, maximum climb thrust T maxc , cruise thrust T cruise , Maximum cruise thrust T max，cruise , descending thrust T des and idle thrust T idle ; Among them, T cruise =D=C D ρv 2 S, C D is the drag coefficient, ρ is the atmospheric density, v is the flight speed, and S is the wing area; T satisfies the following conditions: If it is the climbing stage, T=T maxc If it is level flight, T = T cruise If it is the ground phase and the speed is not equal to 0, T = T idle If it is the descent phase, landing phase, approach phase, or ground phase with a speed of 0, T = T des ; Among them, T maxc =C (△ISA)eff ×T maxcs , C (△ISA)eff is the climb thrust correction factor, T maxcs is the maximum standard climb thrust, T maxcs The following conditions are met: If the engine type is a jet engine, T maxcs =C Tc1 (1-h / C Tc2 )+C Tc3 h 2 ; C Tc1 is the first thrust coefficient, C Tc2 is the second thrust coefficient, C Tc3 is the third thrust coefficient, h is the flight altitude; If the engine type is a turbine engine, T maxcs =(C Tc1 (1-h / C Tc2 )) / v+C Tc3 ; v is the flight speed; If the engine type is a piston engine, T maxcs =C Tc1 (1-h / C Tc2 )+C Tc3 / 3; T idle =C △idle ×C Tc1 (1-h / C Tc2 )+C Tc3 h 2 , C △idle is the idle thrust correction factor; T max，cruise =C T,Cr ×T cruise , C T,Cr is the cruise thrust correction factor; T des =C def ×T maxc , C def is the descent thrust correction factor.

8. The method according to claim 1, characterized in that The hyperparameters for setting the aircraft fuel consumption prediction model are obtained by the following steps: S10, set the first counter N1=1; S11, if N1≤N10, set the second counter N2=1, execute S12, if N1>N10, execute S17; N10 is the number of set hyper parameters; S12, if N2≤N20, randomly divide the data corresponding to the characteristic parameter into N30 data subsets of the same size, set the third counter N3=1, and execute S13; if N2>N20, execute S16; N20 is the number of cross-validation settings; S13, if N3≤N30, use the N3th data subset among the N30 data subsets as the validation set, and the remaining N30-1 data subsets as the training set, and execute S14; otherwise, execute S15; N30 is the set cross-validation fold; S14, using the training set to train the N1th set aircraft fuel consumption prediction model, to obtain the trained N1th set aircraft fuel consumption prediction model as the current aircraft fuel consumption prediction model M N1N2 N3 , and use the validation set to test the current aircraft fuel consumption prediction model M N1N2 N3 Verify and get the corresponding verification result RS N1N2 N3 ; Execute S15, set N3 = N3 + 1, and execute S13; wherein, the N1th set hyperparameter of the aircraft fuel consumption prediction model is the N1th set hyperparameter; S15, obtain the average value AV of the verification result corresponding to the N2th cross-validation under the N1th set hyperparameters N1N2 =(RS N1N2 1 +RS N1N2 2 +……+RS N1N2 N3 +……+RS N1N2 N30 ) / N30, set N2=N2+1, and execute S12; S16, obtain the average value AV of the verification results under the N1th set hyperparameters N1 =(AV N11 +AV N12 +……+AV N1N2 +……+AV N1N20 ) / N20, set N1=N1+1, and execute S11; S17, obtain max(AV1, AV2, ..., AV N1 ,……,AV N10 ) is used as the hyperparameter of the aircraft fuel consumption prediction model, and max() represents the maximum value.

9. An electronic device, characterized in that: including processor and memory; The processor is configured to execute the steps of the method according to any one of claims 1 to 8 by calling the program or instructions stored in the memory.

10. A non-transitory computer-readable storage medium, characterized in that The non-transitory computer-readable storage medium is used to store a program or instruction, and the program or instruction enables a computer to execute the steps of the method according to any one of claims 1 to 8.

Images