Method for predicting flight fuel consumption of aircraft and model training method and device

Abstract

The invention provides a method for predicting flight fuel consumption of an aircraft and a model training method and device. A method for predicting flight fuel consumption of an aircraft includes: receiving time series data of the aircraft in a flight phase, data points in the time series data including a plurality of parameters associated with the aircraft at respective time points; weighting a parameter in a first subset of a plurality of parameters of data points in the time series data, where a weight for each parameter in the first subset is determined based on a Sharpley value indicating a contribution of the parameter to flight fuel consumption of the aircraft; and inputting the weighted time series data into a trained first model to predict the flight fuel consumption of the aircraft, the first model being a long short-term memory (LSTM) network model.

Description

Technical Field

[0001] This disclosure relates to the field of aviation technology, and more specifically, to a method for predicting the flight fuel consumption of an aircraft, a method for training a model for predicting the flight fuel consumption of an aircraft, a computing device, a computer-readable storage medium, and a computer program product. Background Technology

[0002] With the rapid development of the air transport industry, aircraft fuel consumption has become increasingly important. Fuel consumption directly affects airline operating costs, environmental impact, and flight safety. Therefore, accurately predicting aircraft fuel consumption not only helps reduce operating costs but also provides support for flight planning, route optimization, and environmental protection. Summary of the Invention

[0003] According to a first aspect of this disclosure, a method for predicting the flight fuel consumption of an aircraft is provided, comprising: receiving time-series data of the aircraft during a flight phase, the time-series data including data points comprising a plurality of parameters associated with the aircraft at corresponding time points; weighting parameters in a first subset of the plurality of parameters of the data points in the time-series data, wherein the weight for each parameter in the first subset is determined based on a Shapley value indicating the parameter’s contribution to the aircraft’s flight fuel consumption; and inputting the weighted time-series data into a trained first model to predict the aircraft’s flight fuel consumption, wherein the first model is a Long Short-Term Memory (LSTM) network model.

[0004] In some embodiments, the plurality of parameters includes attribute parameters and flight parameters, wherein the flight parameters include flight environment parameters and flight state parameters, wherein the attribute parameters include one or more of the aircraft type, age and payload, the flight environment parameters include one or more of the wind speed, wind direction and temperature, and the flight state parameters include one or more of the latitude and longitude, altitude, speed and heading.

[0005] In some embodiments, the first subset includes flight parameters, and a second subset of the plurality of parameters includes attribute parameters, wherein the parameters in the second subset are not weighted.

[0006] In some embodiments, the plurality of parameters are composed of the first subset.

[0007] In some embodiments, the data points in the time series data include additional parameters at the corresponding time points, which indicate whether the aircraft is under control.

[0008] In some embodiments, the Shapley value for each parameter in the first subset is determined based on the following formula: ,in, Let be the Shapley value of parameter i in the first subset, N be the first subset, S be any subset of N that does not contain parameter i, |S| be the number of parameters in S, M be the number of parameters in N, f(S) be the flight fuel consumption predicted by the trained second model using only the parameters in S, and f(S∪{i}) be the flight fuel consumption predicted by the trained second model using only the parameters in i and S.

[0009] In some embodiments, the second model is trained using unweighted historical time-series data and actual flight fuel consumption data.

[0010] In some embodiments, the second model is an LSTM model, and the first model is trained based on the trained second model using weighted historical time-series data and actual flight fuel consumption data.

[0011] In some embodiments, the time-series data is first time-series data, and the method further includes: receiving second time-series data of the aircraft during a flight phase, the second time-series data being earlier than the first time-series data; and determining a real-time Shapley value for each parameter in the first subset based on the second time-series data, and determining a real-time weight for each parameter in the first subset based on the real-time Shapley value, thereby weighting the first time-series data.

[0012] In some embodiments, the first model includes a random mask layer in the connection from the hidden state at the previous time point to the current time point located in the LSTM unit, the random mask layer being configured to randomly discard elements in the hidden state at the previous time point.

[0013] According to a second aspect of this disclosure, a method is provided for training a model for predicting the flight fuel consumption of an aircraft, the method comprising: constructing a first training set, the first training set including historical aircraft flight time-series data as samples and actual fuel consumption data as labels, the data points in the historical aircraft flight time-series data including multiple parameters associated with the aircraft at corresponding time points; training a second model for predicting the flight fuel consumption of an aircraft using the first training set to obtain a trained second model; determining a Shapley value for each parameter in a first subset of the multiple parameters using the trained second model, the Shapley value of the parameter indicating the contribution of the parameter to the flight fuel consumption of the aircraft; constructing a second training set, the second training set being obtained by weighting the parameters in the first subset of the multiple parameters in the historical aircraft flight time-series data in the first training set, wherein the weight for each parameter in the first subset is determined based on the Shapley value of the parameter; and training a first model for predicting the flight fuel consumption of an aircraft using the second training set to obtain a trained first model, wherein the first model is an LSTM model.

[0014] In some embodiments, the untrained first model is the trained second model.

[0015] According to a third aspect of this disclosure, a computing device is provided, comprising: one or more processors; and a memory storing computer-executable instructions, which, when executed by the one or more processors, cause the one or more processors to perform the method according to any embodiment of the first or second aspect of this disclosure.

[0016] According to a fourth aspect of this disclosure, a computer-readable storage medium is provided having computer-executable instructions stored thereon, which, when executed by a computer, cause the computer to perform the method described according to any embodiment of the first or second aspect of this disclosure.

[0017] According to a fifth aspect of this disclosure, a computer program product is provided, the computer program product including instructions that, when executed by a processor, implement the method according to any embodiment of the first or second aspect of this disclosure.

[0018] Other features and advantages of this disclosure will become clearer from the following detailed description of exemplary embodiments with reference to the accompanying drawings. Attached Figure Description

[0019] The foregoing and other features and advantages of this disclosure will become clear from the following description of embodiments illustrated in conjunction with the accompanying drawings. The drawings, incorporated herein and forming a part of the specification, are further used to explain the principles of this disclosure and to enable those skilled in the art to make and use it. Wherein:

[0020] Figure 1 This is a flowchart illustrating a method for predicting the flight fuel consumption of an aircraft according to some embodiments of the present disclosure;

[0021] Figure 2 This is a schematic structural diagram illustrating an LSTM model according to some embodiments of the present disclosure;

[0022] Figure 3 This is a flowchart illustrating a method for training a model for predicting the flight fuel consumption of an aircraft, according to some embodiments of the present disclosure;

[0023] Figure 4 This is a schematic block diagram illustrating a computing device according to some embodiments of the present disclosure; and

[0024] Figure 5 This is a schematic block diagram illustrating a computer system on which embodiments of the present disclosure may be implemented.

[0025] Note that in the embodiments described below, the same reference numerals are sometimes used across different figures to denote the same parts or parts having the same function, and repeated descriptions are omitted. In this specification, similar reference numerals and letters are used to denote similar items; therefore, once an item is defined in one figure, it does not need to be discussed further in subsequent figures.

[0026] For ease of understanding, the positions, dimensions, and extents of the structures shown in the accompanying drawings and other materials may not represent actual positions, dimensions, and extents. Therefore, the disclosed invention is not limited to the positions, dimensions, and extents disclosed in the accompanying drawings and other materials. Furthermore, the drawings are not necessarily drawn to scale, and some features may be enlarged to show details of specific components. Detailed Implementation

[0027] Various exemplary embodiments of the present disclosure will now be described in detail with reference to the accompanying drawings. It should be noted that, unless otherwise specifically stated, the relative arrangement, numerical expressions, and values ​​of the components and steps set forth in these embodiments do not limit the scope of the present disclosure.

[0028] The following description of at least one exemplary embodiment is merely illustrative and is in no way intended to limit this disclosure or its application or use. Those skilled in the art will understand that they merely illustrate exemplary ways that can be used to implement this disclosure, and are not exhaustive.

[0029] Techniques, methods, and equipment known to those skilled in the art may not be discussed in detail, but where appropriate, such techniques, methods, and equipment should be considered part of the specification.

[0030] In all examples shown and discussed herein, any specific values ​​should be interpreted as merely exemplary and not as limitations. Therefore, other examples of exemplary embodiments may have different values.

[0031] Currently, aircraft fuel consumption prediction techniques are mainly based on traditional regression models, statistical methods, and numerical simulation methods. A common approach is Multiple Linear Regression (MLR). This method predicts aircraft fuel consumption by establishing the relationship between static parameters (such as aircraft type and payload) and fuel consumption, combined with meteorological data from the takeoff and landing sites. However, MLR models have limitations; they cannot handle constantly changing dynamic factors during flight and neglect complex nonlinear interactions. Furthermore, traditional methods typically only perform fuel consumption analysis after the flight, failing to provide real-time updates and support pre-flight planning.

[0032] This disclosure provides a method for predicting aircraft fuel consumption by combining a Long Short-Term Memory (LSTM) network model with Shapley values. This method effectively learns the complex temporal characteristics of aircraft during flight by introducing an LSTM model, enabling predictions based on dynamically changing weather conditions, air traffic control, and flight status. Simultaneously, Shapley values ​​quantify the impact of various factors on fuel consumption, helping the LSTM model learn the complex interactions between features, thereby providing more accurate and interpretable prediction results.

[0033] The methods for predicting aircraft flight fuel consumption according to various embodiments of the present disclosure will now be described in detail with reference to the accompanying drawings. It will be understood that actual methods may include other steps, which are not shown in the drawings and will not be discussed herein in order to avoid obscuring the essential points of the disclosure.

[0034] Figure 1 A flowchart of a method 100 for predicting the flight fuel consumption of an aircraft according to some embodiments of the present disclosure is shown. Figure 1 As shown, method 100 includes steps S102 to S106.

[0035] In step S102, timing data of the aircraft during the flight phase is received. The data points in the timing data include multiple parameters associated with the aircraft at the corresponding time points.

[0036] The term "aircraft" as used in this article refers to all powered aircraft, including but not limited to airplanes. For simplicity, some descriptions below will use airplanes as examples. Generally, a flight phase can begin at the moment of takeoff and end at the moment of landing. In some cases, a flight phase can also begin at the moment of wheel chock removal and end at the moment of wheel chock engagement.

[0037] In some embodiments, the multiple parameters may include attribute parameters and flight parameters. Attribute parameters may be associated with the aircraft's own static properties (or internal factors). For example, attribute parameters may include one or more of the aircraft's type, age, and payload. Flight parameters may be associated with the aircraft's dynamic properties and / or external environmental factors. In some examples, flight parameters may include flight environment parameters and flight state parameters. For example, flight environment parameters may include one or more of wind speed, wind direction, and temperature, and flight state parameters may include one or more of latitude, longitude, altitude, speed, and heading. Attribute parameters typically remain constant during the same flight phase of the same aircraft, while flight parameters may change dynamically over time.

[0038] For example, attribute parameters can be determined based on manifest data. For instance, the aircraft type, age, and payload can be obtained from the LoadMessage (LDM). Flight environment parameters can be determined based on meteorological data. For example, by structuring a GRIB2 (Grid Interface Binary, a binary file format developed by the World Meteorological Organization (WMO) for exchanging and storing meteorological data) file, meteorological data for the current flight route can be reconstructed to obtain corresponding meteorological data such as wind speed, wind direction, and temperature. Flight status parameters can be determined based on airborne data link systems. Airborne data link systems can include, but are not limited to, Aircraft Communication Addressing and Reporting System (ACARS) and Automatic Dependent Surveillance – Broadcast (ADS-B). ADS-B data is typically second-level data and can include latitude, longitude, altitude, speed, and heading. ACARS data is typically minute-level data and can include remaining fuel, latitude, longitude, altitude, speed, and heading.

[0039] In some embodiments, data points in the time series data may include additional parameters at the corresponding time points, indicating whether the aircraft is under air traffic control. For example, the additional parameter may be a Boolean parameter, with a value of 1 indicating that the aircraft is under air traffic control at that time point, and a value of 0 indicating that the aircraft is not under air traffic control. This additional parameter can prompt the model whether it needs to consider the additional fuel consumption caused by air traffic control, thereby improving the model's prediction accuracy under complex airspace conditions.

[0040] In step S104, the parameters in a first subset of the multiple parameters of the data points in the time series data are weighted, wherein the weight of each parameter in the first subset is determined based on the Shapley value, which indicates the contribution of the parameter to the aircraft's flight fuel consumption.

[0041] In some embodiments, the multiple parameters consist of a first subset. That is, all the data that will be input into the trained first model is weighted.

[0042] In some embodiments, the first subset may include flight parameters, and the second subset of multiple parameters may include attribute parameters, with the parameters in the second subset not weighted. The parameters in the first subset are typically dynamically changing, and there are strong coupling relationships between parameters such as wind speed, altitude, and velocity. Weighting such parameters based on Shapley values ​​can characterize their relative importance in time-series data. Inputting this preprocessed data into the model allows it to prioritize key parameters that have historically proven to contribute more to fuel consumption prediction, effectively improving the accuracy of model predictions. In contrast, the parameters in the second subset, such as aircraft type and age, have values ​​that remain largely constant throughout the flight. Their impact on fuel consumption manifests as a relatively fixed systematic bias, with weak dynamic interaction effects with flight parameters. This discriminative weighting method allows the model to more clearly learn the impact of different types of parameters on fuel consumption.

[0043] In step S106, the weighted time series data is input into the trained first model to predict the aircraft's flight fuel consumption, wherein the first model is a Long Short-Term Memory (LSTM) network model.

[0044] In some embodiments, the LSTM model can be specifically trained for a particular aircraft type. That is, different models can be used for different aircraft types, which can make fuel consumption predictions more accurate.

[0045] The output of the first trained model is consistent with the learning objective set during the training phase. For example, the first model can be trained to directly output the amount of fuel consumed by the aircraft during the flight phase; or it can be trained to output the remaining fuel at the end of the flight phase, which can be combined with the initial fuel amount to calculate the flight fuel consumption.

[0046] Compared to traditional MLR models, this disclosure offers advantages in feature interaction processing. MLR methods neglect many potential nonlinear interaction effects, such as the increased fuel consumption that might result from a combination of headwinds and aircraft aging. In contrast, the LSTM model of this disclosure can automatically learn and model these complex nonlinear interactions and handle dynamic time-series data during flight, enabling the model to reflect the impact of time-varying factors on fuel consumption. Furthermore, this disclosure combines Shapley values ​​with the input data, weighting the input based on the contribution of each factor. This allows the LSTM model to more accurately capture the nonlinear interactions between features, thereby improving the accuracy of fuel consumption prediction. Moreover, existing fuel consumption prediction techniques are typically limited to static analysis after the flight ends, while the LSTM model of this disclosure can support pre-flight planning or real-time in-flight prediction updates, providing pilots and airlines with real-time optimization decision support.

[0047] In some embodiments, the Shapley value for each parameter in the first subset is determined based on the following formula: .in, Let f(S) be the Shapley value of parameter i in the first subset, N be the first subset, S be any subset of N that does not contain parameter i, |S| be the number of parameters in S, M be the number of parameters in N, f(S) be the flight fuel consumption predicted by the trained second model using only the parameters in S, and f(S∪{i}) be the flight fuel consumption predicted by the trained second model using only the parameters in i and S.

[0048] Specifically, the above formula first iterates through all possible subsets S in the first subset N that do not contain parameter i. For each such subset S, two predictions are performed: one inputs the parameters within subset S into the second model to obtain the predicted value f(S); the other inputs the parameters within subset S and parameter i into the second model to obtain the predicted value f(S∪{i}). The difference between the two predictions is the marginal contribution of parameter i in the context of a specific subset S. The final Shapley value φ obtained by the formula is... i That is, the weighted average of the marginal contributions of parameter i across all possible subsets S, with its weights determined by the number of subset parameters |S| and the total number of parameters M. Here, the trained second model serves as a prediction function, providing reliable predictions of f(S) and f(S∪{i}) when traversing all feature combinations.

[0049] In some embodiments, the second model is trained using unweighted historical time-series data and actual flight fuel consumption data. Actual flight fuel consumption data can be obtained from ACARS. Training the second model using unweighted raw data avoids introducing pre-defined weights before calculating feature importance, preventing circular reasoning, and ensuring that the calculated Shapley values ​​more accurately reflect the true contribution of each parameter to fuel consumption prediction.

[0050] The second model can be any machine learning model capable of performing regression prediction tasks. For example, the second model can be a linear regression model, a decision tree model, a random forest model, or a gradient boosting decision tree model. In some embodiments, the second model can be an LSTM model. In such embodiments, the first model can be trained on top of the trained second model using weighted historical time-series data and actual flight fuel consumption data. Training the second model first using unweighted historical time-series data and actual flight fuel consumption data provides a reliable initial model. The first model is then fine-tuned using weighted data, which not only optimizes fuel consumption prediction accuracy but also avoids the computational overhead of training from scratch, accelerating the training process.

[0051] In some embodiments, the Shapley value can be calculated and fixed based on historical time-series data. For a set of historical time-series data, the Shapley value of each parameter in the first subset can be obtained by averaging the Shapley values ​​calculated from each data point in that set of time-series data. Alternatively, the Shapley value of each parameter in the first subset can be calculated separately for each set of historical time-series data in multiple sets, and then the average of the Shapley values ​​obtained from all sets of historical time-series data can be used as the final Shapley value of the corresponding parameter.

[0052] In some embodiments, the Shapley value can be updated in real time based on current time-series data (fuel consumption to be predicted). For example, the time-series data of the aircraft during the flight phase received in step S102 can be first time-series data, and method 100 can further include: receiving second time-series data of the aircraft during the flight phase, the second time-series data being earlier than the first time-series data; and determining a real-time Shapley value for each parameter in a first subset based on the second time-series data, and determining a real-time weight for each parameter in the first subset based on the real-time Shapley values, thereby weighting the first time-series data. Similarly, determining the real-time Shapley value for each parameter in the first subset based on the second time-series data can be obtained by averaging the Shapley values ​​calculated from each data point in the second time-series data.

[0053] For purposes of non-restrictive description, Figure 2 This is a schematic structural diagram illustrating an LSTM model according to some embodiments of the present disclosure. Figure 2 As shown, the LSTM model contains multiple sequentially connected LSTM units, representing operations at time points t-1 (202), t (204), and t+1 (206), respectively. It should be noted that... Figure 2 This is a schematic expansion of the working principle of LSTM to describe the transmission and evolution of hidden states and cell states in a time series. In practice, an LSTM model can be implemented by connecting a single LSTM unit through recurrent connections. Figure 2 Multiple LSTM cells in a given context are equivalent to the instantaneous states of a single LSTM cell at different times. For example... Figure 2 As shown, x represents the input, y represents the output, c represents the cell state, h represents the hidden state, and z represents the cell state. f z i z o These represent the gating states of the forget gate, input gate, and output gate, respectively. The tanh activation function transforms the data at the corresponding time point to the (-1, 1) interval. Taking time point 204 as an example, its output y... t The parameter matrix W' can be used in conjunction with c t and h t The product of the augmented matrices is obtained by applying the sigmoid activation function, and the calculation formula is as follows:

[0054] , where the symbol ⊙ represents the element-wise multiplication operation of the matrix.

[0055] If time point 204 is considered the current time point, then time point 202 is the previous time point and time point 206 is the next time point. The time series data (one data point) at the current time point 204 can be represented as d. t (Not shown in the figure, but can be represented in vector form), the time series data (one data point) at the previous time point 202 can be represented as d t-1 (Not shown in the diagram, but can be represented as a vector). Let d t For the first time series data, d t-1 Let's take the second time series data as an example for explanation. We can use the second time series data d... t-1 Determine the Shapley value S to be used for the current time point 204. t (Not shown in the figure, but can be in vector form), and then based on the Shapley value S t and the first time series data d t Determine the input x at the current time point 204 t (For example, x) t =dt ·S t Similarly, for time point 206, its input x t+1 It can be based on d t Determined Shapley value S t+1 For time series data d t+1 The value is obtained through weighting. While this explanation illustrates how to determine the Shapley value for the current time point based on parameters from the previous time point, it is merely illustrative and not restrictive. Considering that flight parameters typically do not change very frequently, the Shapley value for the current time period can be determined based on time-series data from a period preceding the current time period. That is, the Shapley value is updated at regular intervals. The length of these intervals can be set as appropriate.

[0056] During flight, especially long-haul flights, key parameters affecting fuel consumption change with the flight phase. If the weights are fixed, the model might treat data from all other phases (e.g., cruise) with the importance of a particular phase (e.g., climb), or it might use a compromise derived from past data to balance the importance of each phase. Real-time updated Shapley values ​​allow the LSTM model to dynamically respond to these changes in dominant fuel consumption factors, more keenly focusing on which flight parameters are more important in the current phase, resulting in more accurate fuel consumption predictions.

[0057] In some embodiments, the first model includes a random mask layer in the connection from the hidden state at the previous time point to the current time point located in the LSTM unit, the random mask layer being configured to randomly discard elements in the hidden state at the previous time point.

[0058] refer to Figure 2 The hidden state of the previous time point 202 is represented as h. t-1 Before being input to the current time point 204, it can first pass through a random mask layer. This random mask layer includes a number of hidden states h. t-1 The mask contains the same number of elements as the given mask. The random mask has a certain probability of being 0. If the random mask is 0, then the hidden state h corresponding to that random mask is... t-1 Elements in the hidden state h will be discarded; if the random mask is 1, the hidden state h will be... t-1 The elements in the time step, when multiplied by the time step, can still be correctly entered into the current time point 204. The specific value of this probability can be set according to the needs of the model; for example, the probability can be 15%. In this way, time step dropout is implemented, so that the model does not depend on the data at a specific time and prevents the model from overfitting.

[0059] Understandably, the temporal sparsity of multiple parameters in time-series data may vary. Typically, flight environment parameters are sourced from meteorological bureaus, with relatively long time intervals between data points (e.g., half an hour). Flight status parameters are sourced from airborne data link systems, where ACARS messages have shorter time intervals (e.g., 10-15 minutes), and ADS-B data has even shorter time periods (e.g., 5-10 seconds). Therefore, in time-series data, there may be instances where only partial data is available at a given point in time (e.g., only latitude, longitude, altitude, and speed data are available, but not wind speed, wind direction, and temperature data). To address this incomplete data at specific points in time, the data for each parameter can be aggregated, ensuring that the data input to the model is time-aligned.

[0060] For example, since ADS-B data typically has the shortest time period, it can be used as a benchmark to align other parameter data (downward aggregation). As a non-limiting example, assuming a specific point in ADS-B data is taken as the current time point, the latitude, longitude, altitude, speed, and heading at that point in the ADS-B data can be directly taken as the flight status parameters for the current time point. Additionally, the wind speed, wind direction, and temperature at the nearest point in meteorological data can be used as the flight environment parameters for the current time point (nearest-neighbor assignment). Alternatively, multiple flight environment parameter samples within a certain window before and after the current time point can be obtained, and the flight environment parameters for the current time point can be calculated using linear interpolation or step interpolation algorithms (interpolation method). This is advantageous for short-haul flights, ensuring sufficient data points for the time series data.

[0061] For long-haul flights, ADS-B data may be too high-frequency, resulting in an excessive amount of data input to the model, which is detrimental to model training and prediction. Therefore, ADS-B data can be upsampled (downsampled), for example, by aggregating ADS-B data into intervals of 1-5 minutes. Similar methods can be used to align the upsampled ADS-B data with other time-density data, such as meteorological data.

[0062] In some embodiments, for example, Figure 1The flight fuel consumption obtained in step S106 can also be analyzed and explained using Shapley values. Specifically, time-series data and the predicted flight fuel consumption can be input into the Shapley value interpretation module to calculate the contribution of each parameter to the final predicted fuel consumption under the specific circumstances of the flight. This Shapley value interpretation module, located after the LSTM model, can analyze the main factors affecting fuel consumption in this flight, enabling the model to not only output predicted values ​​but also have high interpretability. This analysis provides direct and quantitative data support for airlines to conduct fuel efficiency analysis and flight quality monitoring.

[0063] Understandably, the Shapley value used for post-hoc interpretation here serves a different purpose than the Shapley value used for feature weighting mentioned earlier. The weighted Shapley value is a global weight, located before the model input, aiming to improve the overall performance of all predictions. The interpretative Shapley value, located after the model output, is an analysis of a single prediction result, aiming to understand the reasons behind a particular prediction. Both can be calculated based on different models and data to meet their different needs in fuel consumption prediction. Of course, the Shapley value calculated by the Shapley value interpretation module for this flight can also be used for parameter weighting in future flight fuel consumption predictions.

[0064] On the other hand, this disclosure also provides a method for training a model for predicting the flight fuel consumption of an aircraft. Figure 3 A flowchart of a method 300 for training a model to predict the flight fuel consumption of an aircraft, according to some embodiments of the present disclosure, is shown. Figure 3 As shown, method 300 includes steps S302 to S310.

[0065] In step S302, a first training set is constructed. The first training set includes historical aircraft flight time series data as samples and actual fuel consumption data as labels. The data points in the historical aircraft flight time series data include multiple parameters associated with the aircraft at the corresponding time points.

[0066] Similar to method 100 described above, in method 300, the multiple parameters may include attribute parameters and flight parameters. Flight parameters may include flight environment parameters and flight status parameters. Attribute parameters include one or more of the aircraft type, age, and payload; flight environment parameters include one or more of wind speed, wind direction, and temperature; and flight status parameters include one or more of latitude, longitude, altitude, speed, and heading. In some embodiments, data points in the time-series data may include additional parameters at the corresponding time points, indicating whether the aircraft is under control.

[0067] In step S304, the second model for predicting the flight fuel consumption of an aircraft is trained using the first training set to obtain the trained second model.

[0068] In step S306, the trained second model is used to determine the Shapley value of each parameter in a first subset of multiple parameters, the Shapley value of which indicates the parameter’s contribution to the aircraft’s flight fuel consumption.

[0069] In some embodiments, the multiple parameters may consist of a first subset. In other embodiments, the first subset may include flight parameters, and a second subset of the multiple parameters may include attribute parameters, wherein the parameters in the second subset are not weighted.

[0070] In some embodiments, the Shapley value for each parameter in the first subset is determined based on the following formula: ,in, Let f(S) be the Shapley value of parameter i in the first subset, N be the first subset, S be any subset of N that does not contain parameter i, |S| be the number of parameters in S, M be the number of parameters in N, f(S) be the flight fuel consumption predicted by the trained second model using only the parameters in S, and f(S∪{i}) be the flight fuel consumption predicted by the trained second model using only the parameters in i and S.

[0071] In step S308, a second training set is constructed. The second training set is obtained by weighting the parameters in a first subset of multiple parameters of the data points in the historical aircraft flight time series data of the first training set. The weight of each parameter in the first subset is determined based on the Shapley value of that parameter.

[0072] In step S310, the first model for predicting the flight fuel consumption of an aircraft is trained using the second training set to obtain a trained first model, wherein the first model is an LSTM model.

[0073] In some embodiments, the untrained first model is a trained second model. In other words, the LSTM model for predicting aircraft flight fuel consumption can be trained in stages: the first stage is trained using unweighted historical aircraft flight time-series data and actual flight fuel consumption data, and the second stage is trained based on the first stage using weighted historical aircraft flight time-series data and actual flight fuel consumption data.

[0074] This disclosure also provides a computing device that may include one or more processors and a memory storing computer-executable instructions, which, when executed by the one or more processors, cause the one or more processors to perform the methods described according to any of the foregoing embodiments of this disclosure. Figure 4 As shown, computing device 400 may include one or more processors 402 and memory 404 storing computer-executable instructions that, when executed by the one or more processors 402, cause the one or more processors 402 to perform the methods described according to any of the foregoing embodiments of this disclosure. The one or more processors 402 may be, for example, a central processing unit (CPU) of computing device 400. The one or more processors 402 may be any type of general-purpose processor, or may be a processor specifically designed for predicting aircraft fuel consumption and / or training models for predicting aircraft fuel consumption, such as an application-specific integrated circuit (“ASIC”). Memory 404 may be coupled to the one or more processors 402 and may include various computer-readable media accessible by the one or more processors 402. In various embodiments, memory 404 described herein may include volatile and non-volatile media, removable and non-removable media. For example, memory 404 may include any combination of the following: random access memory (“RAM”), dynamic RAM (“DRAM”), static RAM (“SRAM”), read-only memory (“ROM”), flash memory, cache memory, and / or any other type of non-transitory computer-readable medium. Memory 404 may store instructions that, when executed by processor 402, cause processor 402 to perform the methods described according to any of the foregoing embodiments of this disclosure.

[0075] This disclosure also provides a computer-readable storage medium having computer-executable instructions stored thereon, which, when executed by a computer, cause the computer to perform the methods described according to any of the foregoing embodiments of this disclosure.

[0076] This disclosure also provides a computer program product that may include instructions that, when executed by a processor, can implement the methods described according to any of the foregoing embodiments of this disclosure. The instructions may be any set of instructions that will be executed directly by one or more processors, such as machine code, or any set of instructions that will be executed indirectly, such as a script. The instructions may be stored in an object code format for direct processing by one or more processors, or stored in any other computer language, including scripts or sets of independent source code modules that are interpreted on demand or compiled in advance.

[0077] Figure 5This is a schematic block diagram illustrating a computer system 500 on which embodiments of the present disclosure may be implemented. The computer system 500 includes a bus 502 or other communication mechanism for transmitting information, and a processing means 504 coupled to the bus 502 for processing information. The computer system 500 also includes a memory 506 coupled to the bus 502 for storing instructions to be executed by the processing means 504. The memory 506 may be random access memory (RAM) or other dynamic storage device. The memory 506 may also be used to store temporary variables or other intermediate information during the execution of instructions to be executed by the processing means 504. The computer system 500 also includes a read-only memory (ROM) 508 or other static storage device coupled to the bus 502 for storing static information and instructions for the processing means 504. A storage device 510, such as a magnetic disk or optical disk, is provided and coupled to the bus 502 for storing information and instructions. Computer system 500 may be coupled via bus 502 to output device 512 for providing output to a user, such as, but not limited to, a display (such as a cathode ray tube (CRT) or liquid crystal display (LCD)), speakers, etc. Input device 514, such as a keyboard, mouse, microphone, etc., is coupled to bus 502 for transmitting information and command selections to processing device 504. Computer system 500 may execute embodiments of the present disclosure. Consistent with certain implementations of the present disclosure, results are provided by computer system 500 in response to processing device 504 executing one or more sequences of one or more instructions contained in memory 506. Such instructions may be read into memory 506 from another computer-readable medium, such as storage device 510. Execution of the sequence of instructions contained in memory 506 causes processing device 504 to perform the methods described herein. Alternatively, the teachings may be implemented using hard-wired circuitry in place of or in combination with software instructions. Therefore, implementations of the present disclosure are not limited to any particular combination of hardware circuitry and software. In various embodiments, computer system 500 can be connected across a network to one or more other computer systems, such as computer system 500, to form a networked system via network interface 516. This network may include a private network or a public network such as the Internet. In a networked system, one or more computer systems can store data and supply data to other computer systems. As used herein, the term "computer-readable medium" refers to any medium that participates in providing instructions to processing device 504 for execution. Such media can take many forms, including but not limited to non-volatile media, volatile media, and transmission media. Non-volatile media include, for example, optical discs or magnetic disks such as storage device 510. Volatile media include dynamic memory such as memory 506. Transmission media include coaxial cables, copper wires, and optical fibers, including wiring containing bus 502.Common forms of computer-readable media or computer program products include, for example, floppy disks, flexible disks, hard disks, magnetic tapes, or any other magnetic media, CD-ROMs, digital video discs (DVDs), Blu-ray discs, any other optical media, thumb drives, memory cards, RAM, PROMs and EPROMs, fast EPROMs, any other memory chips or cartridges, or any other tangible media from which a computer can read. Various forms of computer-readable media may be involved when carrying one or more sequences of one or more instructions to processing device 504 for execution. For example, instructions may initially be carried on a disk of a remote computer. The remote computer may load the instructions into its dynamic memory and transmit the instructions over a telephone line using a modem. A modem local to computer system 500 may receive data over a telephone line and convert the data into an infrared signal using an infrared transmitter. An infrared detector coupled to bus 502 may receive the data carried in the infrared signal and place the data on bus 502. Bus 502 carries the data to memory 506, from which processing device 504 retrieves and executes the instructions. Optionally, the instructions received by the memory 506 may be stored on the storage device 510 before or after execution by the processing device 504.

[0078] According to various embodiments, instructions configured to be executed by a processing device to perform a method are stored on a computer-readable medium. The computer-readable medium may be a device for storing digital information. For example, a computer-readable medium includes a compact disc read-only memory (CD-ROM) as known in the art for storing software. The computer-readable medium is accessed by a processor adapted to execute the instructions configured to be executed.

[0079] The foregoing has described one or more exemplary embodiments of this disclosure. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than that shown in the embodiments and may still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require the specific or sequential order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0080] The systems, devices, modules, or units described in the above embodiments can be implemented by computer chips or physical entities, or by products with certain functions. A typical implementation device is a server system. Of course, this disclosure does not exclude the possibility that, with the future development of computer technology, the computer implementing the functions of the above embodiments may be, for example, a personal computer, a laptop computer, an in-vehicle human-machine interaction device, a cellular phone, a camera phone, a smartphone, a personal digital assistant, a media player, a game console, a tablet computer, a wearable device, or any combination thereof.

[0081] The terms "comprising," "including," or any other variations thereof are intended to cover a non-exclusive inclusion, such that a process, method, product, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, product, or apparatus. Without further limitation, the presence of other identical or equivalent elements in the process, method, product, or apparatus that includes said elements is not excluded. For example, the use of terms such as "first" or "second" to denote names does not indicate any particular order.

[0082] For ease of description, the above devices are described in terms of function, divided into various modules. Of course, when implementing one or more embodiments of this disclosure, the functions of each module can be implemented in one or more software and / or hardware, or a module that performs the same function can be implemented by a combination of multiple sub-modules or sub-units. The device embodiments described above are merely illustrative. For example, the division of units is only a logical functional division; in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual couplings, direct couplings, or communication connections may be through some interfaces; indirect couplings or communication connections between devices or units may be electrical, mechanical, or other forms.

[0083] This disclosure is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this disclosure. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in one or more blocks of the flowchart illustrations and / or one or more blocks of the block diagrams.

[0084] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means that implement the functions specified in one or more flowcharts and / or one or more block diagrams. These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process, such that the instructions, which execute on the computer or other programmable apparatus, provide steps for implementing the functions specified in one or more flowcharts and / or one or more block diagrams.

[0085] Those skilled in the art will understand that one or more embodiments of this disclosure may take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, one or more embodiments of this disclosure may take the form of a computer program product implemented on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0086] One or more embodiments of this disclosure can be described in the general context of computer-executable instructions, such as program modules, that are executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform a particular task or implement a particular abstract data type. One or more embodiments of this disclosure can also be practiced in distributed computing environments where tasks are performed by remote processing devices connected via a communication network. In a distributed computing environment, program modules can reside in local and remote computer storage media, including storage devices.

[0087] The same or similar parts between the various embodiments of this disclosure can be referred to mutually, and each embodiment focuses on describing the differences from other embodiments. In particular, for the apparatus embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and relevant parts can be referred to the description of the method embodiments. In the description of this disclosure, the reference to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," "exemplary," etc., means that the specific feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of this disclosure. In this disclosure, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described can be combined in a suitable manner in any one or more embodiments or examples. Furthermore, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this disclosure and the features of different embodiments or examples.

[0088] Additionally, when used in this disclosure, the terms “here,” “above,” “below,” “below,” “in the following,” “overall,” and similar terms should refer to the entirety of this disclosure and not any particular part thereof. Furthermore, unless expressly stated otherwise or otherwise understood in the context in which they are used, conditional language used herein, such as “may,” “possibly,” “for example,” “like,” etc., is generally intended to express that certain embodiments include, while other embodiments do not, certain features, elements, and / or states. Therefore, such conditional language is not generally intended to imply that one or more embodiments require features, elements, and / or states in any way, or whether such features, elements, and / or states are included or performed in any particular embodiment.

[0089] The above description is merely an embodiment of one or more embodiments of this disclosure and is not intended to limit the scope of the one or more embodiments of this disclosure. Various modifications and variations can be made to the one or more embodiments of this disclosure by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this disclosure should be included within the scope of the claims.

Claims

1. A method for predicting the flight fuel consumption of an aircraft, comprising: Receive timing data of an aircraft during its flight phase, wherein the data points in the timing data include multiple parameters associated with the aircraft at the corresponding time points; The parameters in a first subset of the plurality of parameters for the data points in the time-series data are weighted, wherein the weight for each parameter in the first subset is determined based on a Shapley value indicating the parameter's contribution to the aircraft's flight fuel consumption; and The weighted time-series data is input into a trained first model to predict the aircraft's flight fuel consumption, wherein the first model is a Long Short-Term Memory (LSTM) network model.

2. The method according to claim 1, wherein, The multiple parameters include attribute parameters and flight parameters, wherein the flight parameters include flight environment parameters and flight state parameters. The attribute parameters include one or more of the aircraft type, age, and payload; the flight environment parameters include one or more of the wind speed, wind direction, and temperature; and the flight status parameters include one or more of the latitude, longitude, altitude, speed, and heading.

3. The method according to claim 2, wherein, The first subset includes flight parameters, and the second subset of the plurality of parameters includes attribute parameters, wherein the parameters in the second subset are not weighted.

4. The method according to claim 1, wherein, The multiple parameters are composed of the first subset.

5. The method according to claim 1, wherein, The data points in the time series data include additional parameters at the corresponding time points, which indicate whether the aircraft is under control.

6. The method according to claim 1, wherein, The Shapley value for each parameter in the first subset is determined based on the following formula: ， Where, φ i Let be the Shapley value of parameter i in the first subset, N be the first subset, S be any subset of N that does not contain parameter i, |S| be the number of parameters in S, M be the number of parameters in N, f(S) be the flight fuel consumption predicted by the trained second model using only the parameters in S, and f(S∪{i}) be the flight fuel consumption predicted by the trained second model using only the parameters in i and S.

7. The method according to claim 6, wherein, The second model was trained using unweighted historical time-series data and actual flight fuel consumption data.

8. The method according to claim 7, wherein, The second model is an LSTM model, and the first model is trained based on the trained second model using weighted historical time-series data and actual flight fuel consumption data.

9. The method according to claim 1, wherein, The time series data is first time series data, and the method further includes: Receive second timing data of the aircraft during its flight phase, the second timing data being earlier than the first timing data; and Based on the second time series data, the real-time Shapley value of each parameter in the first subset is determined, and the real-time weight for each parameter in the first subset is determined based on the real-time Shapley value, thereby weighting the first time series data.

10. The method according to claim 1, wherein, The first model includes a random mask layer in the LSTM unit that connects the hidden state from the previous time point to the current time point, the random mask layer being configured to randomly discard elements in the hidden state from the previous time point.

11. A method for training a model to predict the flight fuel consumption of an aircraft, the method comprising: Construct a first training set, which includes historical aircraft flight time series data as samples and actual fuel consumption data as labels. The data points in the historical aircraft flight time series data include multiple parameters associated with the aircraft at the corresponding time points. The first training set is used to train a second model for predicting aircraft flight fuel consumption to obtain a trained second model. The trained second model is used to determine the Shapley value of each parameter in a first subset of the plurality of parameters, the Shapley value of which indicates the parameter’s contribution to the aircraft’s flight fuel consumption; A second training set is constructed by weighting the parameters of a first subset of the plurality of parameters from the historical aircraft flight time-series data in the first training set, wherein the weight for each parameter in the first subset is determined based on the Shapley value of that parameter; and The first model for predicting aircraft flight fuel consumption is trained using the second training set to obtain a trained first model, wherein the first model is a Long Short-Term Memory (LSTM) network model.

12. The method according to claim 11, wherein, The untrained first model is the trained second model.

13. A computing device, comprising: One or more processors; as well as A memory storing computer-executable instructions, which, when executed by the one or more processors, cause the one or more processors to perform the method according to any one of claims 1 to 12.

14. A computer-readable storage medium having computer-executable instructions stored thereon, the computer-executable instructions causing a computer to perform the method according to any one of claims 1 to 12 when executed by the computer.

15. A computer program product comprising instructions that, when executed by a processor, implement the method according to any one of claims 1 to 12.

Images