CN118095106B - Method, system and equipment for predicting residual fuel value of airplane - Google Patents
Method, system and equipment for predicting residual fuel value of airplane Download PDFInfo
- Publication number
- CN118095106B CN118095106B CN202410488258.4A CN202410488258A CN118095106B CN 118095106 B CN118095106 B CN 118095106B CN 202410488258 A CN202410488258 A CN 202410488258A CN 118095106 B CN118095106 B CN 118095106B
- Authority
- CN
- China
- Prior art keywords
- layer
- data
- aircraft
- bert
- prediction
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 55
- 239000010763 heavy fuel oil Substances 0.000 title claims abstract description 38
- 239000000446 fuel Substances 0.000 claims abstract description 51
- 230000001133 acceleration Effects 0.000 claims abstract description 12
- 230000000694 effects Effects 0.000 claims abstract description 7
- 239000013598 vector Substances 0.000 claims description 65
- 238000012549 training Methods 0.000 claims description 31
- 230000006870 function Effects 0.000 claims description 28
- 230000015654 memory Effects 0.000 claims description 21
- 238000010606 normalization Methods 0.000 claims description 20
- 230000004913 activation Effects 0.000 claims description 17
- 238000012360 testing method Methods 0.000 claims description 17
- 238000012795 verification Methods 0.000 claims description 17
- 238000013528 artificial neural network Methods 0.000 claims description 10
- 230000001364 causal effect Effects 0.000 claims description 9
- 230000010354 integration Effects 0.000 claims description 3
- 230000008569 process Effects 0.000 abstract description 13
- 230000008859 change Effects 0.000 abstract description 11
- 238000000605 extraction Methods 0.000 abstract description 3
- 230000035945 sensitivity Effects 0.000 abstract description 3
- 230000009977 dual effect Effects 0.000 abstract description 2
- 238000012545 processing Methods 0.000 description 8
- 238000010586 diagram Methods 0.000 description 4
- 238000013459 approach Methods 0.000 description 3
- 230000010339 dilation Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000006467 substitution reaction Methods 0.000 description 2
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000007787 long-term memory Effects 0.000 description 1
- 230000000306 recurrent effect Effects 0.000 description 1
- 230000006403 short-term memory Effects 0.000 description 1
- 238000012731 temporal analysis Methods 0.000 description 1
- 238000000700 time series analysis Methods 0.000 description 1
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention belongs to the technical field of prediction, in particular relates to a method, a system and equipment for predicting an aircraft residual fuel value, and aims to solve the problem that the existing prediction method is poor in prediction effect when an input sequence with dual characteristics of rapid change and slow evolution is processed. The invention comprises the following steps: acquiring acceleration and speed of an airplane as first class data, taking temperature and altitude as second class data, and respectively encoding; adding the first coded data, the second coded data and the third coded data to obtain first added data; and acquiring a residual fuel value through an aircraft fuel prediction model based on the Bert and the double-flow path network based on the first addition data. The invention can effectively process the input sequence with rapid change and slow change through the double-flow path network, improves the sensitivity to the sequence change, and can enhance the semantic understanding of the feature data by combining the initial feature extraction of the Bert model.
Description
Technical Field
The invention belongs to the technical field of prediction, and particularly relates to a method, a system and equipment for predicting a residual fuel value of an aircraft.
Background
In the field of aviation operation, it is important to accurately predict the residual fuel quantity of an aircraft in the flight process, and the method not only relates to flight operation efficiency and cost control, but also directly influences flight safety and emergency decision. Conventional approaches typically rely on real-time fuel amount data provided by the aircraft fuel management system and a fuel consumption model based on predetermined flight plans, flight conditions (e.g., speed, altitude, air temperature, etc.), and engine performance parameters. However, these methods may have a problem of insufficient prediction accuracy in the face of complex and variable flight environments and nonlinear fuel consumption characteristics.
In recent years, with the development of artificial intelligence technology, particularly the remarkable advantages of a deep learning method in time series analysis and feature understanding, researchers begin to explore the application of the method in the task of predicting the residual fuel of an aircraft. Among them, a single model such as long and short term memory network (LSTM), recurrent Neural Network (RNN), etc. has been tried to process flight parameter sequences of acceleration, speed, temperature, altitude, etc. to capture time-series dependency between data. Nevertheless, such models, when processing input sequences that include both rapidly changing and slowly evolving characteristics, may be limited by their inherent structure's ability to adapt to different time scale changes, resulting in less than ideal predictive results.
Aiming at the limitations of the prior art, the invention provides an innovative method which aims to improve the prediction precision and the robustness of the residual fuel value of the aircraft by fusing the Bert model and the double-flow path network.
Disclosure of Invention
In order to solve the above-mentioned problems in the prior art, that is, the existing prediction method may be limited by the adaptability of the internal structure to different time scale changes when processing the input sequence including the dual characteristics of rapid change and slow evolution, so that the prediction effect is not ideal, the present invention provides a method for predicting the residual fuel value of an aircraft, which includes:
step S1, acquiring acceleration and speed of an airplane as first class data, taking temperature and altitude as second class data, and taking the first class data and the second class data as data to be processed;
Step S2, the data to be processed are respectively encoded through a random initialization vector table to obtain first encoded data, encoding of unique identifiers is respectively distributed to the first type data and the second type data to obtain second encoded data, and position encoding is used to obtain third encoded data;
Step S3, adding the first coded data, the second coded data and the third coded data to obtain first added data;
S4, acquiring a residual fuel value through an aircraft fuel prediction model based on the Bert and the double-flow path network based on the first addition data;
the aircraft fuel prediction model based on the Bert and double-flow path network comprises a Bert layer, a TCN layer, an LSTM layer and a prediction layer; the two output ends of the Bert layer are respectively connected with the input end of the TCN layer and the input end of the LSTM layer, the output end of the TCN layer and the output end of the LSTM layer are connected to the input end of the prediction layer after passing through the first adding unit, and the output end of the prediction layer is the output end of the aircraft fuel prediction model based on the Bert and double-flow path network.
Further, the step S4 specifically includes:
acquiring an initial feature vector through a Bert layer based on the first addition data;
dividing the initial feature vector into a first type initial feature vector and a second type initial feature vector according to the first type data and the second type data;
the first type of initial feature vectors are obtained through a TCN layer;
Obtaining a second type of initial feature vector through an LSTM layer;
the first type of feature vectors and the second type of feature vectors are subjected to first summation unit to obtain summation feature vectors;
And acquiring the residual fuel value of the aircraft through the prediction layer based on the summation feature vector.
Further, the Bert layer specifically includes:
6 sequentially connected transducer encoders;
Each transducer encoder sequentially comprises a multi-head attention layer, a first normalization layer, a feedforward neural network layer and a second normalization layer; the input end of each multi-head attention layer is connected to the first normalization layer in a residual error connection mode, and the input end of the feedforward neural network layer is connected to the second normalization layer in a residual error connection mode;
The output layer of the last transform encoder outputs the initial feature vector.
Further, the TCN layer specifically includes:
3 time residual error modules which are connected in sequence;
Each time residual error module comprises a first expansion causal convolution layer, a first weight normalization layer, a first activation function layer, a first Dropout layer, a second expansion causal convolution layer, a second weight normalization layer, a second activation function layer and a second Dropout layer which are sequentially connected;
the input of the first causal layer is connected to the output of the second Dropout layer by means of a residual connection.
Further, the LSTM layer specifically includes:
2 LSTM units connected in sequence;
the input end of each LSTM unit and the characteristic vector output end of the last LSTM unit are connected to a third adding unit;
The third adding unit is connected to the input end of the first product unit through a forgetting gate, the input end of the second product unit through an input gate and the input end of the third product unit through an output gate through independent Sigmoid activation function gates respectively;
The third adding unit is connected to the input end of the second product unit through a tanh activation function gate;
The input end of the first product unit is also connected with the output end of the memory cell of the last LSTM unit;
The output end of the first product unit and the output end of the second product unit are connected to the input end of the fourth adding unit;
the output end of the fourth adding unit is connected to the input end of the memory cell;
The output end of the memory cell is connected to the next LSTM unit and the third multiplication unit after the function gate is activated by the tanh;
the third multiplication unit is connected to the characteristic vector output end;
the feature vector output end of the last LSTM unit is the output end of the LSTM layer.
Further, the aircraft fuel prediction model based on the Bert and double-flow path network comprises the following training method:
A1, acquiring historical to-be-processed data and corresponding historical aircraft residual fuel value data as training data;
step A2, standardizing and dividing training data into a training set, a verification set and a test set;
Step A3, respectively encoding a training set, a verification set and a test set;
step A4, inputting a training set into an aircraft fuel prediction model to be trained based on the Bert and double-flow path network, and outputting a training set predicted value;
step A5, calculating a mean square error loss function based on the predicted value of the training set;
step A6, inputting the verification set into a to-be-trained aircraft fuel prediction model based on the Bert and double-flow path network, and outputting a prediction value of the verification set;
A7, adjusting model parameters according to the predicted value of the training set through a random gradient descent algorithm, adjusting model super-parameters according to the predicted value of the verification set, and repeating the steps A4 to A6 until the mean square error loss function takes the minimum value and the fitting phenomenon occurs;
Step A8, inputting a test set into an aircraft fuel prediction model to be trained based on the Bert and double-flow path network, and outputting a test set predicted value;
And A9, obtaining a trained aircraft fuel prediction model based on the Bert and double-flow path network when the predicted value of the test set meets the expected final effect.
Further, the second encoded data specifically includes:
Each segment is assigned a unique identifier, with 0 representing the segment of the first type of data and 1 representing the segment of the second type of data.
In another aspect of the present invention, an aircraft residual fuel value prediction system is provided, the system comprising:
the data acquisition module is used for acquiring acceleration and speed of the aircraft as first class data, taking temperature and altitude as second class data and taking the first class data and the second class data as data to be processed;
The coding module is used for coding the data to be processed through a random initialization vector table to obtain first coded data, respectively distributing codes of unique identifiers for the first type data and the second type data to obtain second coded data, and obtaining third coded data through position coding;
The data integration module sums the first coded data, the second coded data and the third coded data to obtain first added data;
The residual fuel value prediction module is used for acquiring the residual fuel value through an aircraft fuel prediction model based on the Bert and the double-flow path network based on the first addition data;
the aircraft fuel prediction model based on the Bert and double-flow path network comprises a Bert layer, a TCN layer, an LSTM layer and a prediction layer; the two output ends of the Bert layer are respectively connected with the input end of the TCN layer and the input end of the LSTM layer, the output end of the TCN layer and the output end of the LSTM layer are connected to the input end of the prediction layer after passing through the first adding unit, and the output end of the prediction layer is the output end of the aircraft fuel prediction model based on the Bert and double-flow path network.
In a third aspect of the present invention, an electronic device is provided, including:
At least one processor; and
A memory communicatively coupled to at least one of the processors; wherein,
The memory stores instructions executable by the processor for execution by the processor to implement an aircraft residual fuel value prediction method as described above.
In a fourth aspect of the present invention, a computer readable storage medium is provided, the computer readable storage medium storing computer instructions for execution by the computer to implement an aircraft residual fuel value prediction method as described above.
The invention has the beneficial effects that:
(1) According to the method, the acceleration, the speed, the temperature and the altitude of the aircraft are predicted by carrying out feature extraction processing on the acceleration, the speed, the temperature and the altitude of the aircraft by combining the Bert model and the double-flow path network.
Drawings
Other features, objects and advantages of the present application will become more apparent upon reading of the detailed description of non-limiting embodiments, made with reference to the accompanying drawings in which:
FIG. 1 is a flow chart of a method for predicting the fuel remaining in an aircraft according to an embodiment of the invention;
FIG. 2 is a schematic structural diagram of an aircraft fuel prediction model based on Bert and dual-path networks in an embodiment of the invention;
FIG. 3 is a schematic diagram of a transducer encoder in the Bert layer in an embodiment of the present invention;
fig. 4 is a schematic structural diagram of a time residual module in a TCN layer according to an embodiment of the present invention;
Fig. 5 is a schematic structural diagram of an LSTM cell in an LSTM layer in an embodiment of the present invention.
Detailed Description
The application is described in further detail below with reference to the drawings and examples. It is to be understood that the specific embodiments described herein are merely illustrative of the application and are not limiting of the application. It should be noted that, for convenience of description, only the portions related to the present application are shown in the drawings.
It should be noted that, without conflict, the embodiments of the present application and features of the embodiments may be combined with each other. The application will be described in detail below with reference to the drawings in connection with embodiments.
The invention provides a method for predicting the residual fuel value of an aircraft, which can effectively process an input sequence with rapid change and slow change through a double-flow path network, improves the sensitivity to sequence change, and can enhance the semantic understanding of feature data by combining a Bert model to perform initial feature extraction.
The invention discloses a method for predicting the residual fuel value of an aircraft, which comprises the following steps:
step S1, acquiring acceleration and speed of an airplane as first class data, taking temperature and altitude as second class data, and taking the first class data and the second class data as data to be processed;
Step S2, the data to be processed are respectively encoded through a random initialization vector table to obtain first encoded data, encoding of unique identifiers is respectively distributed to the first type data and the second type data to obtain second encoded data, and position encoding is used to obtain third encoded data;
Step S3, adding the first coded data, the second coded data and the third coded data to obtain first added data;
S4, acquiring a residual fuel value through an aircraft fuel prediction model based on the Bert and the double-flow path network based on the first addition data;
the aircraft fuel prediction model based on the Bert and double-flow path network comprises a Bert layer, a TCN layer, an LSTM layer and a prediction layer; the two output ends of the Bert layer are respectively connected with the input end of the TCN layer and the input end of the LSTM layer, the output end of the TCN layer and the output end of the LSTM layer are connected to the input end of the prediction layer after passing through the first adding unit, and the output end of the prediction layer is the output end of the aircraft fuel prediction model based on the Bert and double-flow path network.
In order to more clearly describe a method for predicting the remaining fuel value of an aircraft according to the present invention, each step of the embodiment of the present invention will be described in detail with reference to fig. 1.
The method for predicting the residual fuel value of the aircraft comprises the following steps S1 to S4, wherein the steps are described in detail:
step S1, acquiring acceleration and speed of an airplane as first class data, taking temperature and altitude as second class data, and taking the first class data and the second class data as data to be processed;
Step S2, the data to be processed are respectively encoded through a random initialization vector table to obtain first encoded data, encoding of unique identifiers is respectively distributed to the first type data and the second type data to obtain second encoded data, and position encoding is used to obtain third encoded data;
The codes in this embodiment are Init Embeddings, segment Embeddings and Position Embeddings codes, respectively, init Embeddings is used for initial coding, and random initialization vector representation is adopted; segment Embeddings are used to additionally encode the different segments, each segment being assigned a unique identifier; positionEmbeddings for position coding, assigning a position vector to each feature position;
in this embodiment, the second encoded data specifically includes:
Each segment is assigned a unique identifier, with 0 representing the segment of the first type of data and 1 representing the segment of the second type of data.
Step S3, adding the first coded data, the second coded data and the third coded data to obtain first added data;
S4, acquiring a residual fuel value through an aircraft fuel prediction model based on the Bert and the double-flow path network based on the first addition data;
the aircraft fuel prediction model based on the Bert and double-flow path network, as shown in figure 2, comprises a Bert layer, a TCN layer, an LSTM layer and a prediction layer; the two output ends of the Bert layer are respectively connected with the input end of the TCN layer and the input end of the LSTM layer, the output end of the TCN layer and the output end of the LSTM layer are connected to the input end of the prediction layer after passing through the first adding unit, and the output end of the prediction layer is the output end of the aircraft fuel prediction model based on the Bert and double-flow path network.
In this embodiment, the Bert layer specifically includes:
6 sequentially connected transducer encoders;
As shown in fig. 3, each transducer encoder includes, in order, a multi-head attention layer, a first normalization layer, a feedforward neural network layer, and a second normalization layer; the input end of each multi-head attention layer is connected to the first normalization layer in a residual error connection mode, and the input end of the feedforward neural network layer is connected to the second normalization layer in a residual error connection mode;
The output layer of the last transform encoder outputs the initial feature vector.
Each ransformer encoder firstly calculates a multi-head attention characteristic vector through a multi-head attention layer, and performs layer normalization operation to serve as an input characteristic vector of the feedforward neural network; and the feedforward neural network performs residual connection on the input feature vector and the output of the feedforward neural network and performs layer normalization operation to obtain the input feature vector of the next transducer encoder. The output eigenvector of the final transducer encoder after 6 repetitions is noted as the initial eigenvector.
In this embodiment, the TCN layer specifically includes:
3 time residual error modules which are connected in sequence;
As shown in fig. 4, each time residual module includes a first dilation causal convolution layer, a first weight normalization layer, a first activation function layer, a first Dropout layer, a second dilation causal convolution layer, a second weight normalization layer, a second activation function layer, and a second Dropout layer that are sequentially connected;
the input of the first causal layer is connected to the output of the second Dropout layer by means of a residual connection.
In this embodiment, the LSTM layer specifically includes:
2 LSTM units connected in sequence;
As shown in fig. 5, the input end of each LSTM cell and the feature vector output end of the last LSTM cell are connected to a third summing unit; in FIG. 5, the input features of the current LSTM cell are represented as The feature vector of the last LSTM cell is expressed as；
The third adding unit respectively passes through the forgetting gate through independent Sigmoid activation function gatesConnected to the input of the first product unit via an input gateConnected to the input of the second product unit and through the output gateAn input terminal connected to the third multiplying unit; sigmoid activation function gate is expressed as；
The third adding unit is connected to the input end of the second product unit through a tanh activation function gate; in FIG. 5Representing the variable value calculated by the tanh activation function;
The input end of the first product unit is also connected with the output end of the memory cell of the last LSTM unit;
The output end of the first product unit and the output end of the second product unit are connected to the input end of the fourth adding unit;
The output end of the fourth adding unit is connected to the input end of the memory cell; in FIG. 5, the memory cells are shown as ；
The output end of the memory cell is connected to the next LSTM unit and the third multiplication unit after the function gate is activated by the tanh;
the third multiplication unit is connected to the characteristic vector output end;
the feature vector output end of the last LSTM unit is the output end of the LSTM layer.
The LSTM unit processing process comprises the following steps: the first step, the aircraft temperature and altitude feature vector extracted by the Bert layer is used as the input feature vector of the first LSTM unit in the LSTM layer. Second, outputting the characteristic vector of the current unit through the forgetting gate, the input gate and the updating of the memory cellAnd updating the memory cells of the current cell. And thirdly, taking the memory cells and the output characteristic vectors as the input of the next LSTM unit, and finally outputting the characteristic vectors of the temperature and the altitude of the airplane extracted by the LSTM layer.
In this embodiment, the step S4 specifically includes:
acquiring an initial feature vector through a Bert layer based on the first addition data;
dividing the initial feature vector into a first type initial feature vector and a second type initial feature vector according to the first type data and the second type data;
the first type of initial feature vectors are obtained through a TCN layer;
Obtaining a second type of initial feature vector through an LSTM layer;
the first type of feature vectors and the second type of feature vectors are subjected to first summation unit to obtain summation feature vectors;
And acquiring the residual fuel value of the aircraft through the prediction layer based on the summation feature vector.
In this embodiment, the prediction layer is a full connection layer, and is used for predicting the remaining fuel value of the aircraft.
In this embodiment, a dual-path network is constructed to process the feature vector with a fast change speed and the feature vector with a slow change speed respectively. One of the path branches of the dual-flow path network described above will employ the TCN layer for processing the initial eigenvectors of aircraft acceleration and velocity, and the other path branch will employ the LSTM layer for processing the initial eigenvectors of aircraft temperature and altitude. And finally, adding the feature vectors extracted by the two path branches to be used as input vectors of a prediction layer. The method can effectively adapt to input sequences with different change speeds, and has higher sensitivity to prediction.
In this embodiment, the method for training the aircraft fuel prediction model based on the Bert and dual-path network includes:
A1, acquiring historical to-be-processed data and corresponding historical aircraft residual fuel value data as training data;
Step A2, standardizing and dividing training data into a training set, a verification set and a test set; in the embodiment, the proportion of the training set, the verification set and the test set is 8:1:1, and the standardization processing is carried out by a Min-Max method; the verification set is used for adjusting the super parameters, and determining whether training is stopped according to whether the predictive model is over-fitted or not; after the prediction model training is finished, evaluating the final effect of the prediction model on the test set;
Step A3, respectively encoding a training set, a verification set and a test set;
step A4, inputting a training set into an aircraft fuel prediction model to be trained based on the Bert and double-flow path network, and outputting a training set predicted value;
step A5, calculating a mean square error loss function based on the predicted value of the training set;
step A6, inputting the verification set into a to-be-trained aircraft fuel prediction model based on the Bert and double-flow path network, and outputting a prediction value of the verification set;
A7, adjusting model parameters according to the predicted value of the training set through a random gradient descent algorithm, adjusting model super-parameters according to the predicted value of the verification set, and repeating the steps A4 to A6 until the mean square error loss function takes the minimum value and the fitting phenomenon occurs;
Step A8, inputting a test set into an aircraft fuel prediction model to be trained based on the Bert and double-flow path network, and outputting a test set predicted value;
And A9, obtaining a trained aircraft fuel prediction model based on the Bert and double-flow path network when the predicted value of the test set meets the expected final effect.
In the training process of the model, a ReLU activation function is adopted, wherein an LSTM layer adopts a Sigmoid activation function and a tanh activation function, the value of batch size is set to be 32, the value of epoch is set to be 16, dropout is set to be 0.5, the initial learning rate is set to be 2 multiplied by 10 -5, the approach degree of actual output and expected output is judged by adopting a mean square error loss function, the loss value is minimized by adopting a AdamW optimizer to continuously train, and finally, a prediction model based on Bert and a double-flow path network is obtained.
Although the steps are described in the above-described sequential order in the above-described embodiments, it will be appreciated by those skilled in the art that in order to achieve the effects of the present embodiments, the steps need not be performed in such order, and may be performed simultaneously (in parallel) or in reverse order, and such simple variations are within the scope of the present invention.
An aircraft residual fuel value prediction system according to a second embodiment of the present invention includes:
the data acquisition module is used for acquiring acceleration and speed of the aircraft as first class data, taking temperature and altitude as second class data and taking the first class data and the second class data as data to be processed;
The coding module is used for coding the data to be processed through a random initialization vector table to obtain first coded data, respectively distributing codes of unique identifiers for the first type data and the second type data to obtain second coded data, and obtaining third coded data through position coding;
The data integration module sums the first coded data, the second coded data and the third coded data to obtain first added data;
The residual fuel value prediction module is used for acquiring the residual fuel value through an aircraft fuel prediction model based on the Bert and the double-flow path network based on the first addition data;
the aircraft fuel prediction model based on the Bert and double-flow path network comprises a Bert layer, a TCN layer, an LSTM layer and a prediction layer; the two output ends of the Bert layer are respectively connected with the input end of the TCN layer and the input end of the LSTM layer, the output end of the TCN layer and the output end of the LSTM layer are connected to the input end of the prediction layer after passing through the first adding unit, and the output end of the prediction layer is the output end of the aircraft fuel prediction model based on the Bert and double-flow path network.
It will be clear to those skilled in the art that, for convenience and brevity of description, the specific working process of the system described above and the related description may refer to the corresponding process in the foregoing method embodiment, which is not repeated here.
It should be noted that, in the system for predicting the remaining fuel value of the aircraft provided in the foregoing embodiment, only the division of the foregoing functional modules is illustrated, in practical application, the foregoing functional allocation may be performed by different functional modules according to needs, that is, the modules or steps in the foregoing embodiment of the present invention are further decomposed or combined, for example, the modules in the foregoing embodiment may be combined into one module, or may be further split into a plurality of sub-modules, so as to complete all or part of the functions described above. The names of the modules and steps related to the embodiments of the present invention are merely for distinguishing the respective modules or steps, and are not to be construed as unduly limiting the present invention.
An electronic device of a third embodiment of the present invention includes:
At least one processor; and
A memory communicatively coupled to at least one of the processors; wherein,
The memory stores instructions executable by the processor for execution by the processor to implement an aircraft residual fuel value prediction method as described above.
A fourth embodiment of the present invention is a computer-readable storage medium storing computer instructions for execution by the computer to implement an aircraft residual fuel value prediction method as described above.
It will be clearly understood by those skilled in the art that, for convenience and brevity of description, the specific working process of the storage device and the processing device described above and the related description may refer to the corresponding process in the foregoing method embodiment, which is not repeated herein.
Those of skill in the art will appreciate that the various illustrative modules, method steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware, computer software, or combinations of both, and that the program(s) corresponding to the software modules, method steps, may be embodied in Random Access Memory (RAM), memory, read Only Memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium known in the art. To clearly illustrate this interchangeability of electronic hardware and software, various illustrative components and steps have been described above generally in terms of their functionality. Whether such functionality is implemented as electronic hardware or software depends upon the particular application and design constraints imposed on the solution. Those skilled in the art may implement the described functionality using different approaches for each particular application, but such implementation is not intended to be limiting.
The terms "first," "second," and the like, are used for distinguishing between similar objects and not for describing a particular sequential or chronological order.
The terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus/apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus/apparatus.
Thus far, the technical solution of the present invention has been described in connection with the preferred embodiments shown in the drawings, but it is easily understood by those skilled in the art that the scope of protection of the present invention is not limited to these specific embodiments. Equivalent modifications and substitutions for related technical features may be made by those skilled in the art without departing from the principles of the present invention, and such modifications and substitutions will be within the scope of the present invention.
Claims (8)
1. A method for predicting a fuel remaining value of an aircraft, the method comprising:
step S1, acquiring acceleration and speed of an airplane as first class data, taking temperature and altitude as second class data, and taking the first class data and the second class data as data to be processed;
Step S2, the data to be processed are respectively encoded through a random initialization vector table to obtain first encoded data, codes of unique identifiers are respectively distributed to the first type data and the second type data to obtain second encoded data, and position codes are used to obtain third encoded data;
Step S3, adding the first coded data, the second coded data and the third coded data to obtain first added data;
S4, acquiring a residual fuel value through an aircraft fuel prediction model based on the Bert and the double-flow path network based on the first addition data;
the aircraft fuel prediction model based on the Bert and double-flow path network comprises a Bert layer, a TCN layer, an LSTM layer and a prediction layer; the two output ends of the Bert layer are respectively connected with the input end of the TCN layer and the input end of the LSTM layer, the output end of the TCN layer and the output end of the LSTM layer are connected to the input end of the prediction layer after passing through the first adding unit, and the output end of the prediction layer is the output end of the aircraft fuel prediction model based on the Bert and double-flow path network;
the aircraft fuel prediction model based on the Bert and double-flow path network comprises the following training method:
A1, acquiring historical to-be-processed data and corresponding historical aircraft residual fuel value data as training data;
step A2, standardizing and dividing training data into a training set, a verification set and a test set;
Step A3, respectively encoding a training set, a verification set and a test set;
step A4, inputting a training set into an aircraft fuel prediction model to be trained based on the Bert and double-flow path network, and outputting a training set predicted value;
step A5, calculating a mean square error loss function based on the predicted value of the training set;
step A6, inputting the verification set into a to-be-trained aircraft fuel prediction model based on the Bert and double-flow path network, and outputting a prediction value of the verification set;
A7, adjusting model parameters according to the predicted value of the training set through a random gradient descent algorithm, adjusting model super-parameters according to the predicted value of the verification set, and repeating the steps A4 to A6 until the mean square error loss function takes the minimum value and the fitting phenomenon occurs;
Step A8, inputting a test set into an aircraft fuel prediction model to be trained based on the Bert and double-flow path network, and outputting a test set predicted value;
Step A9, when the predicted value of the test set meets the expected final effect, a trained aircraft fuel prediction model based on the Bert and double-flow path network is obtained;
The step S4 specifically includes:
acquiring an initial feature vector through a Bert layer based on the first addition data;
dividing the initial feature vector into a first type initial feature vector and a second type initial feature vector according to the first type data and the second type data;
the first type of initial feature vectors are obtained through a TCN layer;
Obtaining a second type of initial feature vector through an LSTM layer;
the first type of feature vectors and the second type of feature vectors are subjected to first summation unit to obtain summation feature vectors;
And acquiring the residual fuel value of the aircraft through the prediction layer based on the summation feature vector.
2. The method for predicting the remaining fuel value of an aircraft according to claim 1, wherein the Bert layer specifically comprises:
6 sequentially connected transducer encoders;
Each transducer encoder sequentially comprises a multi-head attention layer, a first normalization layer, a feedforward neural network layer and a second normalization layer; the input end of each multi-head attention layer is connected to the first normalization layer in a residual error connection mode, and the input end of the feedforward neural network layer is connected to the second normalization layer in a residual error connection mode;
The output layer of the last transform encoder outputs the initial feature vector.
3. The method for predicting the remaining fuel value of an aircraft according to claim 1, wherein the TCN layer specifically comprises:
3 time residual error modules which are connected in sequence;
Each time residual error module comprises a first expansion causal convolution layer, a first weight normalization layer, a first activation function layer, a first Dropout layer, a second expansion causal convolution layer, a second weight normalization layer, a second activation function layer and a second Dropout layer which are sequentially connected;
the input of the first causal layer is connected to the output of the second Dropout layer by means of a residual connection.
4. The method for predicting the remaining fuel value of an aircraft according to claim 1, wherein the LSTM layer specifically comprises:
2 LSTM units connected in sequence;
the input end of each LSTM unit and the characteristic vector output end of the last LSTM unit are connected to a third adding unit;
The third adding unit is connected to the input end of the first product unit through a forgetting gate, the input end of the second product unit through an input gate and the input end of the third product unit through an output gate through independent Sigmoid activation function gates respectively;
The third adding unit is connected to the input end of the second product unit through a tanh activation function gate;
The input end of the first product unit is also connected with the output end of the memory cell of the last LSTM unit;
The output end of the first product unit and the output end of the second product unit are connected to the input end of the fourth adding unit;
the output end of the fourth adding unit is connected to the input end of the memory cell;
The output end of the memory cell is connected to the next LSTM unit and the third multiplication unit after the function gate is activated by the tanh;
the third multiplication unit is connected to the characteristic vector output end;
the feature vector output end of the last LSTM unit is the output end of the LSTM layer.
5. The method for predicting the remaining fuel value of an aircraft according to claim 1, wherein the second encoded data is specifically:
Each segment is assigned a unique identifier, with 0 representing the segment of the first type of data and 1 representing the segment of the second type of data.
6. An aircraft residual fuel value prediction system for implementing an aircraft residual fuel value prediction method according to any one of claims 1 to 5, said prediction system comprising:
the data acquisition module is used for acquiring acceleration and speed of the aircraft as first class data, taking temperature and altitude as second class data and taking the first class data and the second class data as data to be processed;
The coding module is used for coding the data to be processed through a random initialization vector table to obtain first coded data, respectively distributing codes of unique identifiers for the first type data and the second type data to obtain second coded data, and obtaining third coded data through position coding;
The data integration module sums the first coded data, the second coded data and the third coded data to obtain first added data;
The residual fuel value prediction module is used for acquiring the residual fuel value through an aircraft fuel prediction model based on the Bert and the double-flow path network based on the first addition data;
the aircraft fuel prediction model based on the Bert and double-flow path network comprises a Bert layer, a TCN layer, an LSTM layer and a prediction layer; the two output ends of the Bert layer are respectively connected with the input end of the TCN layer and the input end of the LSTM layer, the output end of the TCN layer and the output end of the LSTM layer are connected to the input end of the prediction layer after passing through the first adding unit, and the output end of the prediction layer is the output end of the aircraft fuel prediction model based on the Bert and double-flow path network.
7. An electronic device, comprising:
At least one processor; and
A memory communicatively coupled to at least one of the processors; wherein,
The memory stores instructions executable by the processor for performing the method of predicting aircraft residual fuel values of any one of claims 1-5.
8. A computer readable storage medium having stored thereon computer instructions for execution by a computer to implement the method of predicting aircraft residual fuel values of any one of claims 1-5.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410488258.4A CN118095106B (en) | 2024-04-23 | 2024-04-23 | Method, system and equipment for predicting residual fuel value of airplane |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410488258.4A CN118095106B (en) | 2024-04-23 | 2024-04-23 | Method, system and equipment for predicting residual fuel value of airplane |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN118095106A CN118095106A (en) | 2024-05-28 |
| CN118095106B true CN118095106B (en) | 2024-06-28 |
Family
ID=91163273
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410488258.4A Active CN118095106B (en) | 2024-04-23 | 2024-04-23 | Method, system and equipment for predicting residual fuel value of airplane |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN118095106B (en) |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108801387A (en) * | 2018-05-21 | 2018-11-13 | 郑州大学 | A kind of fuel tanker Fuel Oil Remaining measuring system and method based on learning model |
| CN111898020A (en) * | 2020-06-18 | 2020-11-06 | 济南浪潮高新科技投资发展有限公司 | Knowledge learning system recommendation method, device and medium based on BERT and LSTM |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP7229496B1 (en) * | 2022-09-12 | 2023-02-28 | １７Ｌｉｖｅ株式会社 | Systems, methods, and computer readable media for anomaly detection |
-
2024
- 2024-04-23 CN CN202410488258.4A patent/CN118095106B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108801387A (en) * | 2018-05-21 | 2018-11-13 | 郑州大学 | A kind of fuel tanker Fuel Oil Remaining measuring system and method based on learning model |
| CN111898020A (en) * | 2020-06-18 | 2020-11-06 | 济南浪潮高新科技投资发展有限公司 | Knowledge learning system recommendation method, device and medium based on BERT and LSTM |
Also Published As
| Publication number | Publication date |
|---|---|
| CN118095106A (en) | 2024-05-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111860982A (en) | Wind power plant short-term wind power prediction method based on VMD-FCM-GRU | |
| CN106338708B (en) | Electric energy metering error analysis method combining deep learning and recurrent neural network | |
| CN108879732B (en) | Transient stability evaluation method and device for power system | |
| CN111814956B (en) | Multi-task learning air quality prediction method based on multi-dimensional secondary feature extraction | |
| Wei et al. | Learning the health index of complex systems using dynamic conditional variational autoencoders | |
| CN111461463A (en) | Short-term load prediction method, system and equipment based on TCN-BP | |
| CN111190349A (en) | Method, system and medium for monitoring state and diagnosing fault of ship engine room equipment | |
| CN111738520A (en) | System load prediction method fusing isolated forest and long-short term memory network | |
| CN113722985A (en) | Method and system for evaluating health state and predicting residual life of aircraft engine | |
| CN111832825A (en) | Wind power prediction method and system integrating long-term and short-term memory network and extreme learning machine | |
| Nasser et al. | A hybrid of convolutional neural network and long short-term memory network approach to predictive maintenance | |
| Zheng et al. | A fault prediction of equipment based on CNN-LSTM network | |
| Wang et al. | A remaining useful life prediction model based on hybrid long-short sequences for engines | |
| CN112734106A (en) | Method and device for predicting energy load | |
| CN114498619A (en) | Wind power prediction method and device | |
| CN114169091A (en) | Method for establishing prediction model of residual life of engineering mechanical part and prediction method | |
| CN112287596A (en) | Engine residual life prediction method based on clustering and LSTM | |
| Xue et al. | Data-driven prognostics method for turbofan engine degradation using hybrid deep neural network | |
| CN115114843A (en) | Battery capacity prediction method, related device, equipment and storage medium | |
| CN113673774A (en) | Aero-engine remaining life prediction method based on self-encoder and time sequence convolution network | |
| CN118095106B (en) | Method, system and equipment for predicting residual fuel value of airplane | |
| CN112232570A (en) | Forward active total electric quantity prediction method and device and readable storage medium | |
| Alberto-Olivares et al. | Remaining useful life prediction for turbofan based on a multilayer perceptron and kalman filter | |
| Wenqiang et al. | Remaining useful life prediction for mechanical equipment based on temporal convolutional network | |
| Danilin et al. | Algorithm for determining optimum operation tolerances of memristor-based artificial neural networks |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant |