Optimization method of fuel consumption during aircraft cruising phase

Abstract

The present invention relates to providing a method for optimizing fuel consumption during the cruising phase of an aircraft, wherein the flight parameters and meteorological data of the aircraft are collected in real time through airborne sensors, and the flight parameters and meteorological data include flight altitude, flight speed, outside temperature, wind speed, wind direction and aircraft weight; the collected flight data are transmitted to a ground server at intervals through an air-to-ground data transmission system; the ground server performs data cleaning, data normalization and feature selection on the flight data; a deep learning neural network model is trained based on historical flight data to predict the fuel flow and ground speed under current flight conditions; the unit distance fuel cost is calculated through a mathematical model; an optimization algorithm is used to find the optimal cruising speed with the lowest fuel consumption; and the optimal cruising speed is fed back to the aircraft through the air-to-ground data transmission system to adjust the flight state. The present invention can dynamically adjust the cruising speed according to real-time flight data, significantly improve fuel economy, and has strong adaptability and is easy to operate.

Description

Technical Field

[0001] The invention relates to the field of fuel optimization in the cruising phase of an aircraft, and in particular to a method for optimizing fuel consumption in the cruising phase of an aircraft. Background Art

[0002] During the cruising phase of civil aviation flights, fuel consumption is the most important part of airline operating costs. Traditional cruising speeds are usually calculated based on a fixed cost index in the flight manual or flight performance software. However, these traditional calculation methods have many limitations. For example, they fail to fully consider real-time meteorological conditions and changes in flight parameters, and lack flexibility in response to changes in the flight environment. In addition, the flight manual only provides cruising speeds under fixed conditions, and does not provide detailed calculation methods and models for minimizing dynamic fuel consumption, resulting in a lack of flexibility and pertinence in the selection of cruising speeds, and an inability to guarantee optimal fuel efficiency in a variety of flight environments.

[0003] Existing fuel consumption optimization methods are often based on empirical formulas or traditional mathematical models, and do not fully utilize the advantages of modern artificial intelligence and big data. These methods usually determine the cruising speed through manual calculations before flight or use simple table lookup methods, and cannot be dynamically adjusted according to real-time data during actual flight. Aerodynamic calculation formulas are also used in the use of performance software, and some key data need to rely on querying known performance curves. After the aircraft is put into use, there will be problems such as engine performance degradation, so the initial performance curve will be unsuitable. Therefore, in a complex flight environment, changing meteorological conditions and engine performance degradation, it is difficult for existing methods to minimize fuel consumption.

[0004] With the improvement of civil aviation flight automation and data collection capabilities, QAR (Quick Access Recorder) equipment can provide a large amount of historical data related to flight status, and air-to-ground data transmission systems such as ACARS (Aircraft Communications Addressing and Reporting System) can also transmit real-time data during flight, which provides possibilities for data-driven deep learning models. By using historical flight data to train deep learning neural network models, it is possible to accurately predict fuel flow and ground speed under different conditions, and dynamically calculate the optimal cruising speed in combination with mathematical models, thereby optimizing fuel consumption during flight. Summary of the invention

[0005] In view of the shortcomings of the prior art, the present invention aims to provide a method for optimizing fuel consumption during the cruising phase of an aircraft. By integrating deep learning and data-driven optimization technology, dynamic calculation of the optimal indicated airspeed (IAS) during the cruising phase is achieved, thereby minimizing fuel consumption.

[0006] To achieve the above-mentioned purpose, the present invention provides a method for optimizing fuel consumption during the cruising phase of an aircraft, which mainly includes the following steps:

[0007] S1. During the flight, the aircraft's flight parameters and meteorological data are collected in real time through airborne sensors. The collected data include but are not limited to flight altitude, flight speed, outside temperature, wind speed, wind direction, pitch angle and aircraft weight;

[0008] S2, the collected real-time flight data is transmitted to the ground server at appropriate intervals through the Aircraft Communications Addressing and Reporting System (ACARS), and the ground server receives and stores the data for subsequent processing and optimization calculation. At the same time, the training data of the neural network model comes from the historical QAR (Quick Access Recorder) data;

[0009] S3. After the collected real-time flight data is transmitted to the ground server, data cleaning, data normalization, and feature selection data processing steps are required to ensure the quality of the data and the effectiveness of the model;

[0010] S4. Train a deep learning neural network model based on historical flight data to predict fuel flow and ground speed in real time under current flight conditions. The input of the neural network model includes parameters such as flight altitude, flight speed, weather conditions, pitch angle, and aircraft weight, and the output is the predicted value of fuel flow and ground speed;

[0011] S5. Calculate the fuel cost per unit distance based on a mathematical model, where the fuel cost model depends on the predicted fuel flow and ground speed;

[0012] S6. Using an optimization algorithm based on a fuel cost model, find the optimal cruising speed with the lowest fuel consumption;

[0013] S7. The ground server feeds back the calculated optimal cruising speed to the aircraft through the air-ground communication system for reference and use by the pilot or the automatic flight system to adjust the flight status and achieve optimal control of fuel consumption.

[0014] Wherein, step S1 includes the following sub-steps:

[0015] S11. Continuously collect flight parameters and meteorological data through the onboard sensor system. The data collection frequency is usually set at least once per second to ensure the timeliness and accuracy of the data;

[0016] S12. The collected data will first be transferred to the flight management computer FMC. Before transmitting the data to the ground server, the collected data will be compressed and packaged to reduce the use of transmission bandwidth.

[0017] In step S2, the air-to-ground data transmission method is to transmit real-time flight data to the ground server at appropriate intervals through the air-to-ground communication system (ACARS). Specifically, the ACARS system uses very high frequency (VHF), high frequency (HF) or satellite communication links to transmit the collected flight parameters (such as flight altitude, speed, outside temperature, wind speed, etc.) to the ground station. The ground server receives and processes these data for optimization calculations, and then returns the calculation results to the aircraft through the same communication link. The data transmission process adopts an automatic triggering method to automatically collect and send data at predetermined time intervals or under specific conditions to ensure real-time and data validity.

[0018] Step S3 includes the following sub-steps:

[0019] S31, data cleaning, missing value processing part identifies which parameters are missing, fills the missing values ​​by forward filling or backward filling, and maintains the continuity of the data by utilizing the trend or periodicity of the data; the outlier processing part uses the interquartile range method IQR to detect outliers in the data; calculates the first quartile of the data Q 1 and the third quartile Q 3 , then define , the judgment condition for abnormal values ​​is greater than or less than , for the detected outliers, choose to remove them;

[0020] S32. Use the standardization method to convert each feature data into a standard normal distribution with a mean of 0 and a variance of 1. The standardization formula is:

[0021] ,

[0022] in, is the original feature data, is the mean of the feature, is the standard deviation of the feature, It is the standardized feature data;

[0023] S33. Select the required data: flight altitude, flight speed, outside temperature, wind speed, wind direction and aircraft weight.

[0024] Training the deep learning neural network model based on historical flight data in step S4 includes:

[0025] The input layer contains 6 neurons, and the input data includes flight parameters such as flight altitude, flight speed, external temperature, wind speed, wind direction, and aircraft weight.

[0026] Hidden layer 1: Contains 128 neurons, and the activation function is ReLU (rectified linear unit) to capture the complex nonlinear relationship between input features.

[0027] Hidden layer 2: contains 64 neurons, with ReLU as the activation function, which further extracts features to improve the prediction accuracy of the model.

[0028] The output layer contains 2 neurons, which are used to predict the fuel flow and ground speed respectively. The activation function is a linear activation function to ensure that the output is a continuous real value, which is suitable for regression tasks.

[0029] Model training uses historical QAR data to cover a variety of flight states and environmental conditions to ensure the generalization and adaptability of the model.

[0030] The loss function uses mean squared error (MSE) to measure the difference between the model's predicted value and the actual value:

[0031] in, is the predicted value, is the true value.

[0032] The optimization algorithm uses the Adam optimizer, which combines the advantages of momentum and adaptive learning rate and has good convergence speed and stability. The calculation process of the Adam optimizer involves two steps:

[0033] The first-order moment estimation (momentum) performs an exponentially weighted moving average on the gradient. The purpose of this part is to calculate the mean of the gradient to correct the direction of the gradient.

[0034] ,

[0035] in, is the momentum term, is the exponential decay rate of the first-order moment estimate (usually 0.9), is the current gradient.

[0036] The second moment estimate (squared gradient average) takes an exponentially weighted moving average of the squared gradients to estimate the variance of the gradients.

[0037] ,

[0038] in, is the second-order moment estimate, is the exponential decay rate of the second-order moment estimate (usually 0.999).

[0039] The training process uses the small batch gradient descent method to divide the data set into multiple small batches to improve the efficiency and stability of training. The early stopping strategy is used to avoid overfitting of the model. When the error of the validation set no longer decreases in several rounds of iterations, the training is stopped.

[0040] After training is complete, the validation set and test set are used to evaluate the performance of the model. The mean square error (MSE) and R² coefficient are calculated to ensure that the model has good predictive ability.

[0041] In step S5, the mathematical model in step S5 is: .

[0042] The mathematical model reasoning is:

[0043] During the cruise phase, the fuel cost is: ,in is the total fuel cost, is the unit fuel cost, Fuel consumption for cruising flight, , that is, the total fuel consumption, equal to the fuel flow Multiply by the total time T ;Will Divide by the distance to get the fuel cost per unit distance , expressed as: ,in, is the fuel cost per unit distance, is the unit fuel cost, is the fuel flow rate, The ground speed.

[0044] In step S6, the specific steps of the optimization algorithm are:

[0045] S61, first set an indicated airspeed LAS The initial speed interval is ,in and are the minimum and maximum cruising speeds, respectively, and set a ;

[0046] S62, divide the current speed range according to the golden ratio to obtain two new test points:

[0047] ,

[0048] The golden ratio is used to divide the search range into two sub-intervals, one of which is 61.8% of the current interval;

[0049] S63, using the previously trained neural network model, input speed and , and other flight parameters within the sample, note and is in two samples separately, without changing except Other features except and Modify separately The neural network model outputs the fuel flow rate at a given speed. , and ground speed ;

[0050] S64, calculating the fuel cost per unit distance based on the predicted fuel flow and ground speed;

[0051] '

[0052] S65. Compare costs and narrow the search range. If , then the optimal value is at , now update the right boundary: ;

[0053] Otherwise update the left border: ;

[0054] S66, repeating steps S62 to S65: applying the golden section method again in the updated speed interval to calculate a new segmentation point, predicting the fuel flow rate and ground speed, and comparing the unit fuel cost. When the length of the interval is less than When , the search process stops and the end point of the interval is considered to be the optimal indicated airspeed.

[0055] Preferably, the air-to-ground data transmission system in step S7 is specifically an ACARS ground station that encrypts and packages the calculated data and sends it to the air. The data transmission process may pass through multiple relay stations to ensure that the signal covers the entire flight area. The aircraft communication management unit (CMU) will decode and verify the data to ensure that the data is complete and correct, and then distribute the decoded data to the flight management system (FMS) or directly pass it to the pilot.

[0056] At the same time, the present invention provides a server, which includes a processor and a memory, wherein the memory stores at least one program, and the program is loaded and executed by the processor to implement the method for optimizing fuel consumption during the cruising phase of the aircraft.

[0057] A computer-readable storage medium is also provided, wherein the storage medium has at least one program, and the program is loaded and executed by a processor to implement the method for optimizing fuel consumption during the aircraft cruising phase.

[0058] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0059] (1) By integrating deep learning neural network models and mathematical modeling methods, the optimal cruising speed can be dynamically calculated based on real-time flight data, thereby minimizing fuel consumption under changing flight conditions and significantly improving fuel economy.

[0060] (2) By using the air-to-ground data transmission system (such as ACARS), the real-time collection, transmission and ground optimization calculation of flight data are realized. Compared with the existing methods based on flight manuals or fixed algorithms, the present invention can be flexibly adjusted under different flight conditions and has stronger adaptability.

[0061] (3) The multi-layer perceptron (MLP) neural network model trained with historical flight data can accurately predict fuel flow and ground speed under different conditions, making the optimization of fuel consumption more precise and reducing the errors introduced by the traditional model due to simplified assumptions.

[0062] (4) The golden section search method is used to quickly find the optimal value within the specified cruising speed range. It has high computational efficiency and avoids the computational complexity brought by exhaustive search, which is suitable for real-time optimization needs in practical applications.

[0063] (5) The present invention greatly reduces the pilot's operating burden, improves flight efficiency, and reduces the impact of human factors on fuel optimization through automatic data collection, transmission and optimization calculation. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] Figure 1 It is a method framework diagram of the present invention;

[0065] Figure 2 is a schematic diagram of the neural network structure of the present invention;

[0066] Figure 3 is a schematic diagram of the air-to-ground data transmission process of the present invention;

[0067] Figure 4 It is the optimization algorithm diagram of the present invention. DETAILED DESCRIPTION

[0068] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. 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 ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0069] like Figure 1 , an aircraft cruising phase fuel consumption optimization method based on deep learning and real-time data transmission, comprising the following steps:

[0070] Step S1: collect QAR data, select features, clean and normalize them in the ground server computing center to provide a data set for the deep learning neural network model, and divide it into a training set and a test set;

[0071] The QAR described in step S1 is a flight data recording device that can record various important parameters during the flight, such as flight altitude, speed, outside temperature, wind speed, wind direction, engine status, aircraft weight, etc.; during the flight, the QAR records the flight parameters every second in real time and stores them in the device; after the flight, these data can be exported through physical media (such as a memory card) or through a network connection; data collection usually covers multiple flight cycles in order to obtain comprehensive and diverse flight condition data, ensuring that as many flight conditions as possible are covered when training the neural network model.

[0072] The feature selection mentioned above is that the original QAR data may contain a large number of flight parameters, and not all of these parameters have a significant impact on the prediction of fuel flow and ground speed. Therefore, it is necessary to find the most relevant features through feature selection to simplify the model and improve the prediction performance. After feature selection, the most relevant features for fuel efficiency, such as flight altitude, flight speed, outside temperature, wind speed, wind direction and aircraft weight, are finally retained to ensure the simplicity and effectiveness of model training.

[0073] Data cleaning is an important step to ensure data quality. There may be outliers, missing values ​​or data collection errors during the flight. If these bad data are not processed, it will affect the training and prediction results of the model. The specific steps are: missing value processing, for missing values, use forward fill (Forward Fill) or backward fill (BackwardFill) method, use adjacent valid data to fill; outlier processing uses the interquartile range method (IQR) to identify and remove outliers to ensure data quality and stability of model training. The formula is ,in, are the first and third quartiles of the data respectively. The criteria for determining outliers are:

[0074] ;

[0075] Data normalization is to convert all input features into the same numerical range in order to eliminate the magnitude differences between features and ensure the stability and convergence speed of model training. This paper uses a standardized method to map feature data to a standard normal distribution with a mean of 0 and a variance of 1. The specific formula is: ,in, is the original feature data, is the mean of the feature, is the standard deviation of the feature. After standardization, each feature value will have the same dimension, thereby improving the training efficiency of the model. For flight parameters, if the conditions of different flights vary dramatically, standardization can help improve the generalization ability of the model.

[0076] The method of dividing the training set and the test set into a ratio of 0.8 and 0.2 is mainly to ensure that the model can accurately predict and have good generalization ability in practical applications, and avoid excessive errors on new data. In this way, the performance of the model can be more reliably evaluated, and suitable models and hyperparameter configurations can be found, so as to obtain better prediction results in practical scenarios.

[0077] Step S2: The ground server computing center trains the deep learning neural network model, selects the appropriate model structure, activation function, loss calculation function, and optimization algorithm, and saves the model after training to provide subsequent rapid prediction. ability.

[0078] like Figure 2 , the neural network model is specifically constructed as follows:

[0079] In the model structure design, the input layer contains feature nodes related to flight parameters, such as flight altitude, flight speed, outside temperature, wind speed, wind direction, aircraft weight, etc. These parameters are key factors affecting fuel flow and ground speed; the network includes two hidden layers, hidden layer 1: contains 128 neurons, and the activation function is ReLU (Rectified Linear Unit). The ReLU activation function is often used to introduce nonlinear factors so that the model can handle complex input and output relationships. Hidden layer 2: contains 64 neurons and still uses the ReLU activation function. Compared with the first layer, the second layer has fewer neurons to reduce the complexity of the model and avoid overfitting problems. Output layer: contains two neurons, which output the fuel flow respectively. and ground speed The predicted value of , the activation function is a linear function (used for regression problems).

[0080] Working principle: The working principle of neural network can be divided into two stages: forward propagation and back propagation. In the forward propagation stage, the input data is passed from the input layer to the output layer layer by layer. Each layer will perform calculations and conversions, and the output results are used to predict fuel flow and ground speed. The input layer accepts input features, such as flight altitude, flight speed, outside temperature, wind speed, wind direction, aircraft weight, etc. These features describe the current flight status. In the hidden layer, the data of the input layer will pass through the neurons of each layer in turn. In the hidden layer, each neuron is weighted and summed by weights and bias, and then nonlinearly transformed by an activation function. Each neuron will perform weighted summation on the input. The formula is: ,in, is the weight, is the input feature, is the bias; the calculated Will pass through the activation function and get the output , the activation function is used to introduce nonlinearity so that the neural network can learn complex relationships; the output of the hidden layer is passed to the output layer, where a linear activation function is used to ensure that the output is a continuous real number, which is suitable for the prediction of fuel flow and ground speed; after forward propagation, the neural network generates predicted values ​​(such as fuel flow and ground speed), which are compared with the actual target values ​​to calculate the loss function; the loss function is used to measure the gap between the predicted value and the actual value. The smaller the loss function, the more accurate the model's prediction; backpropagation is a key step in optimizing the neural network, the purpose of which is to minimize the loss function and thus improve the accuracy of the model. Through the chain rule, the derivative (i.e., gradient) of the weight and bias of each neuron relative to the loss function is calculated to indicate how to adjust the weight and bias to reduce the loss; based on the calculated gradient, the weight and bias are updated through an optimization algorithm (such as the Adam optimizer). The optimization algorithm adjusts the weights and biases of each neuron to continuously reduce the value of the loss function, thereby improving the model's predictive ability. Forward propagation and backpropagation are an iterative process, and each iteration is called a "training round (epoch)". Through multiple rounds of training, the neural network gradually learns the relationship between input and output. During the training process, the neural network will learn through the training set and verify through the test set to ensure that the model has good predictive ability for unseen data.

[0081] The activation function selects the ReLU function. The ReLU (Rectified Linear Unit) activation function is one of the most commonly used activation functions in deep learning. It is widely used in many neural network models because of its simple calculation and superior performance. Its formula is: ,

[0082] ,

[0083] It can be seen that ReLU cuts off the part of the input that is below 0. This feature introduces nonlinearity to the network, allowing the neural network to learn complex input and output mappings.

[0084] Preferably, the loss function is the Mean Squared Error (MSE), which evaluates the model error by calculating the square of the difference between the predicted value and the true value, thereby helping to optimize the model parameters. Its mathematical expression is: ,in, is the number of samples, that is, the number of samples in the data set used to calculate the loss, is the true value, i.e. The actual target value of samples, is the predicted value, that is, the model The purpose of the mean square loss function is to make the predicted value of the model as close to the true value as possible, thereby minimizing the prediction error.

[0085] Preferably, the optimization algorithm in the BP neural network is the Adam (Adaptive Moment Estimation) optimization algorithm.

[0086] Step S3: After the BP neural network model is trained and saved in the ground computing server, the data for calculating the optimal cruising speed is obtained in real time through ACARS, and then input into the optimization algorithm for calculation.

[0087] like Figure 4 , where the specific steps of the optimization algorithm are:

[0088] S301, set the initial value of the golden section search method, in this example, set , and then determine the first two endpoints through preliminary search , the formula is: .

[0089] S302, the two endpoints are combined with other features to be divided into two groups of data and input into the BP neural network to predict .

[0090] S303, calculate two different endpoints (that is, The formula for the unit distance fuel cost is: .

[0091] S304, compare the values ​​calculated by different endpoints. , then search to the left, the formula is: , otherwise search to the right.

[0092] S305. After multiple iterations of searching, , thus finding the optimal , stop the calculation.

[0093] Step S4: After the ground center calculates the optimal cruising speed, it feeds back to the CMU system of the aircraft through ACARS.

[0094] like Figure 3 , the entire data transmission loop is:

[0095] S401. After a certain aircraft completes a flight on a certain route, required data is extracted from QAR.

[0096] S402. When an aircraft is flying in real time, it transmits data to a ground computing center via ACARS at certain time intervals.

[0097] S403: After the ground center calculates the optimal cruising speed, it feeds back to the aircraft CMU system through ACARS.

[0098] At the same time, the present invention provides an aircraft cruising phase fuel consumption optimization system based on deep learning and real-time data transmission, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the aircraft cruising phase fuel consumption optimization method based on deep learning and real-time data transmission as described in any of the above items.

Claims

1. A method for optimizing fuel consumption during the cruising phase of an aircraft, characterized in that: The following steps are involved: S1. Collecting the flight parameters and meteorological data of the aircraft in real time through airborne sensors, wherein the flight parameters and meteorological data include flight altitude, flight speed, outside temperature, wind speed, wind direction and aircraft weight; S2, transmitting the collected flight data to the ground server at intervals through the air-to-ground data transmission system; S3. The ground server performs data cleaning, data normalization and feature selection on the flight data to ensure the quality of the data; S4. training a deep learning neural network model based on historical flight data, wherein the model is used to predict fuel flow and ground speed under current flight conditions; Training the deep learning neural network model based on historical flight data in step S4 includes: The input layer contains 6 neurons, and the input data includes flight altitude, flight speed, external temperature, wind speed, wind direction, and aircraft weight flight parameters; Hidden layer 1: contains 128 neurons, and the activation function is the rectified linear unit ReLU, which is used to capture the complex nonlinear relationship between input features; Hidden layer 2: contains 64 neurons, with ReLU as the activation function, which further extracts features to improve the prediction accuracy of the model; The output layer contains 2 neurons, which are used to predict fuel flow and ground speed respectively. The activation function is a linear activation function to ensure that the output is a continuous real value, which is suitable for regression tasks. S5. Calculate the fuel cost per unit distance through a mathematical model; The mathematical model in step S5 is: , Indicated airspeed, is the fuel cost per unit distance, is the unit fuel cost, is the fuel flow rate, is the ground speed; S6. Use an optimization algorithm to find the optimal cruising speed with the lowest fuel consumption; Step S6 includes the following sub-steps: S61, first set an indicated airspeed The initial speed interval is ,in and are the minimum and maximum cruising speeds, and set a ; S62, divide the current speed range according to the golden ratio to obtain two new test points: ， The golden ratio is used to divide the search range into two sub-intervals, one of which is 61.8% of the current interval; S63, using the previously trained neural network model, input speed and , and other flight parameters within the sample, note and is in two samples separately, without changing except Other features except and Modify separately The neural network model outputs the fuel flow rate at a given speed. , and ground speed ; S64, calculating the fuel cost per unit distance based on the predicted fuel flow and ground speed; ， S65. Compare costs and narrow the search range. If , then the optimal value is at , now update the right boundary: ; Otherwise update the left border: ; S66, repeating steps S62 to S65: applying the golden section method again in the updated speed interval to calculate a new segmentation point, predicting the fuel flow rate and ground speed, and comparing the unit fuel cost. When the length of the interval is less than When , the search process is stopped, and the end point of the interval is considered to be the optimal indicated airspeed; S7. Feedback the optimal cruising speed to the aircraft via an air-to-ground data transmission system to adjust the flight status.

2. The method for optimizing fuel consumption during the cruising phase of an aircraft according to claim 1, characterized in that: Step S1 includes the following sub-steps: S11. Continuously collect flight parameters and meteorological data through the onboard sensor system, with the data collection frequency set to at least once per second to ensure the timeliness and accuracy of the data; S12. The collected data will first be transferred to the flight management computer FMC. Before transmitting the data to the ground server, the collected data will be compressed and packaged to reduce the use of transmission bandwidth.

3. The method for optimizing fuel consumption during the cruising phase of an aircraft according to claim 1, characterized in that: Step S3 includes the following sub-steps: S31, data cleaning, missing value processing part identifies which parameters are missing, fills the missing values ​​by forward filling or backward filling, and maintains the continuity of data by utilizing the trend or periodicity of the data; outlier processing part uses the interquartile range method IQR to detect outliers in the data; Calculate the first quartile of the data Q 1 and the third quartile Q 3 , then define , the judgment condition for abnormal values ​​is greater than or less than , for the detected outliers, choose to remove them; S32. Use the standardization method to convert each feature data into a standard normal distribution with a mean of 0 and a variance of 1. The standardization formula is: , in, is the original feature data, is the mean of the feature, is the standard deviation of the feature, It is the standardized feature data; S33. Select the required data: flight altitude, flight speed, outside temperature, wind speed, wind direction and aircraft weight.

4. The method for optimizing fuel consumption during the cruising phase of an aircraft according to claim 1, characterized in that: Step S7 includes the following steps: S71. The calculated optimal cruising speed is transmitted to the data feedback module and fed back to the CMU system of the aircraft through ACARS.

5. A server, comprising a processor and a memory, characterized in that: At least one program is stored in the memory, and the program is loaded and executed by the processor to implement the method for optimizing fuel consumption during the cruise phase of an aircraft as claimed in any one of claims 1 to 4.

6. A computer-readable storage medium, the storage medium having at least one program, characterized in that: The program is loaded and executed by a processor to implement the method for optimizing fuel consumption during the cruise phase of an aircraft as described in any one of claims 1 to 4.

Images